{ "status": "completed", "completed_count": 58, "applicable_count": 58, "failed_count": 0, "schema_version": 13, "config_hash": "b3ff58c7ebfb16e9", "data": { "scenes": [ { "scene_index": 1, "heading": "S#1. 숲속 / 해 질 무렵", "beat_title": "겁에 질린 흙 묻은 얼굴", "representative_moment": "C06이 겁에 질려 눈을 크게 뜨고 카메라를 응시하는 정면 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-to-face frontal close shot, handheld tremor blur, low-key dusk light", "t2i_prompt": "Photorealistic cinematic still. C06O08 in a worn black jacket, an adult Korean man, occupies the middle-center foreground, his terrified face dominating the frame. He is facing the camera, frozen mid-run, with intensely widened eyes locked directly on the lens. C06O08's face and upper shoulder are visibly smeared with dirt and mud, the skin texture mixed with sweat. Subtle, subdued traces of red-brown injuries and fresh scratches mark his cheeks and upper shoulder area. Captured from a chest-to-face height observer angle, his shoulders extend toward the lower frame edge. Low-key dusk light in muted blue-gray wraps his muddy features. The immediate space behind his head is a [: wet, dim, and gloomy blue-gray blur of forest mist and dark foliage], maintaining a cold and desolate atmosphere shrouded in soft peripheral motion blur.", "outfit_assignments": [ { "character_id": "C06", "outlook_id": "O08" } ], "source_facts": [ "겁에 질려 눈을 크게 뜬 한국인 남성의 흙 묻은 얼굴 정면", "Low-key dusk light in muted blue-gray wraps the muddy face", "wet blue-gray blur at the edges" ], "visual_inferences": [ "The dirt is mixed with sweat, creating a grimy texture on his skin.", "The injuries on his upper body look like fresh scratches from bushes." ], "creative_decisions": [ "Blood and wounds are kept subdued rather than overly graphic to focus on the psychological terror.", "Using muted blue-gray tones exclusively in the background blur to enhance the cold, isolating feeling." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "122ec9175b8aea1d", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "3563641eef5717fa", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "Not directly continuing a previous scene's exact layout or action flow.", "_shot_index": 2, "visible_entities": [ "C06" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 1, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera inherits the end of the close_observation flow: chest-to-face height directly in front of the fleeing 한국인 남성, backing away in a tight retreat as his dirt-smeared face occupies the central 75% of frame. It is a direct, near-frontal close shot with slight handheld tremor; the forest is reduced to wet blue-gray blur at the edges, while the widened eyes lock past the lens just before the whip toward danger.", "lighting_mood": "Low-key dusk light in muted blue-gray wraps the muddy face, with tiny red-brown accents in the wounds kept subdued rather than graphic.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "한국인 남성 terrified face", "screen_zone": "middle_center", "target_id": "C06", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C06" ], "allowed_outlook_pairs": [ { "character_id": "C06", "outlook_id": "O08" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": null, "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "man_dirt_smeared_face", "element_type": "character_state", "character_name": "Korean man", "description": "The Korean man's face and upper shoulder are visibly smeared with dirt and mud. He maintains a terrified expression with intensely widened eyes. Subtle, subdued traces of red-brown injuries and scratches are present on his facial and shoulder areas.", "applies_to_shots": [ 2, 3 ], "element_scope": "close", "source_facts": [ "겁에 질려 눈을 크게 뜬 한국인 남성의 흙 묻은 얼굴 정면", "dirt-smeared face occupies the central 75% of frame", "tiny red-brown accents in the wounds", "recoiling shoulder and cheek" ], "visual_inferences": [ "The dirt is mixed with sweat, creating a grimy texture on his skin.", "The injuries on his upper body look like fresh scratches from bushes." ], "creative_decisions": [ "Blood and wounds are kept subdued rather than overly graphic to focus on the psychological terror." ], "confidence": "high" }, { "element_id": "dusk_forest_environment", "element_type": "environment_state", "character_name": "", "description": "The dense forest is shrouded in a cold, low-key blue-gray dusk light. The surroundings feature a dim, blurry tree line in the background, establishing a wet, gloomy, and desolate atmosphere.", "applies_to_shots": [ 2, 3 ], "element_scope": "full", "source_facts": [ "숲속 / 해 질 무렵", "wet blue-gray blur at the edges", "Low-key dusk light in muted blue-gray", "dim tree line", "cold dusk backwash silhouettes the attacker" ], "visual_inferences": [ "The foliage is thick but fades rapidly into the shadowy dusk lighting, with moisture reflecting the cold ambient light." ], "creative_decisions": [ "Using muted blue-gray tones exclusively in the background to enhance the cold, isolating feeling of the chase." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "b1f65145985cfc1b" }, { "scene_index": 1, "heading": "S#1. 숲속 / 해 질 무렵", "beat_title": "검은 인영의 습격", "representative_moment": "A fast dark shape lunges through the air toward the terrified, injured man in the dusky forest.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low observer perspective angled upward, wide framing capturing the airborne attacker and the recoiling victim below", "t2i_prompt": "Photorealistic cinematic still. A fast dark shape, appearing as a black silhouette with a humanlike outline, hangs in the upper-center midground, moving diagonally toward a severely injured Korean man below frame right in the same dense forest clearing they both occupy. Captured from a low observer perspective slightly below eye level and angled upward in a wide framing, the dark shape is frozen mid-air at a three-quarter left angle, executing a forceful airborne lunge with both feet off the ground, momentum preserved with its outline trailing backward, body volume preserved against the dim ambient fill, face completely obscured with no discernible details, its unseen gaze fixed on the target. The severely injured Korean man is caught from an over-the-shoulder angle in the lower-right foreground, staggering backward, his visible face and upper shoulder visibly smeared with dirt and sweat, featuring intensely widened eyes and subtle traces of red-brown scratches, gazing up at the attacker. A cold, low-key dusk backwash silhouettes the attacker through thick, blue-gray drifting fog, while wet, nearly black tree trunks form a dim vertical backdrop behind the figure. A faint rim of gray-blue light traces the dark shape's shoulders, with a restrained rusty-red glint highlighting the recoiling man in the wet, gloomy atmosphere.", "outfit_assignments": [], "source_facts": [ "겁에 질려 눈을 크게 뜬 한국인 남성의 흙 묻은 얼굴 정면", "어디선가 쏜살같이 달려온 검은 인영이 남자를 덮친다", "숲속 / 해 질 무렵" ], "visual_inferences": [ "fixed_element body_full_pose: airborne lunge mid-air", "The dirt is mixed with sweat, creating a grimy texture on his skin.", "The injuries on his upper body look like fresh scratches from bushes.", "The foliage is thick but fades rapidly into the shadowy dusk lighting, with moisture reflecting the cold ambient light." ], "creative_decisions": [ "Positioning the attacker high in the upper-center frame and the victim low on the right to emphasize the role asymmetry and forceful attack.", "Muted blue-gray tones exclusively in the background to enhance the cold, isolating feeling of the chase.", "Restrained rusty-red glint near the victim to highlight the injuries without being overly graphic." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "ebd025a010ce0e1d", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "31c537cf24d5f818", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 1, "dependency_reason": "exact_background continuity from the previous forest shot", "_shot_index": 3, "visible_entities": [ "C12" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 1, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The shot completes the attack_reveal pan from the man’s widened gaze, stopping slightly below eye level and angled upward into the dim tree line. 검은 인영 hangs in the upper middle of the frame with both feet off the ground, lunging diagonally toward the lower-right edge where 한국인 남성 is partially caught as a recoiling shoulder and cheek, preserving the same frantic motion but shifting emphasis to the attacker’s position.", "lighting_mood": "A cold dusk backwash silhouettes the attacker, with only a faint rim of gray-blue light and a restrained rusty-red glint near the injured man.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "한국인 남성 below frame right", "label": "검은 인영 airborne attacker", "screen_zone": "upper_center", "target_id": "C12", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C12" ], "allowed_outlook_pairs": [ { "character_id": "C12", "outlook_id": "O13" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C12", "policy_type": "identity_reference", "policy": "generic_descriptor_allowed", "reason": "검은 인영 is meant to read as an unreadable occult silhouette rather than an identifiable person; a generic dark figure descriptor preserves the reveal." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "man_dirt_smeared_face", "element_type": "character_state", "character_name": "Korean man", "description": "The Korean man's face and upper shoulder are visibly smeared with dirt and mud. He maintains a terrified expression with intensely widened eyes. Subtle, subdued traces of red-brown injuries and scratches are present on his facial and shoulder areas.", "applies_to_shots": [ 2, 3 ], "element_scope": "close", "source_facts": [ "겁에 질려 눈을 크게 뜬 한국인 남성의 흙 묻은 얼굴 정면", "dirt-smeared face occupies the central 75% of frame", "tiny red-brown accents in the wounds", "recoiling shoulder and cheek" ], "visual_inferences": [ "The dirt is mixed with sweat, creating a grimy texture on his skin.", "The injuries on his upper body look like fresh scratches from bushes." ], "creative_decisions": [ "Blood and wounds are kept subdued rather than overly graphic to focus on the psychological terror." ], "confidence": "high" }, { "element_id": "dusk_forest_environment", "element_type": "environment_state", "character_name": "", "description": "The dense forest is shrouded in a cold, low-key blue-gray dusk light. The surroundings feature a dim, blurry tree line in the background, establishing a wet, gloomy, and desolate atmosphere.", "applies_to_shots": [ 2, 3 ], "element_scope": "full", "source_facts": [ "숲속 / 해 질 무렵", "wet blue-gray blur at the edges", "Low-key dusk light in muted blue-gray", "dim tree line", "cold dusk backwash silhouettes the attacker" ], "visual_inferences": [ "The foliage is thick but fades rapidly into the shadowy dusk lighting, with moisture reflecting the cold ambient light." ], "creative_decisions": [ "Using muted blue-gray tones exclusively in the background to enhance the cold, isolating feeling of the chase." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 1, "shot_index": 2, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [], "forbidden_refs": [], "readiness_policy": "not_applicable", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "ac34cc8af85ea99a" }, { "scene_index": 2, "heading": "S#2. 도시 뒷골목 / N", "beat_title": "안개 낀 뒷골목 가로등 아래, 목을 물어뜯는 남자의 붉은 눈", "representative_moment": "짙은 안개 속 가로등 불빛 아래 전봇대에서 C06이 붉은 눈을 번뜩이며 C09의 목을 물어뜯는 스틸컷", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low waist-height observer wide shot, deep focus, low-key sickly gold sodium lighting cutting through dense damp blue-gray fog", "t2i_prompt": "Photorealistic cinematic still. The camera sits at waist height looking straight down a damp corridor, placing the wet alley pavement in the lower frame and an old damp utility pole extending into the upper frame. A damp, blue-gray fog envelops the 12-meter distance, illuminated by a dim, sickly gold sodium streetlight casting heavy, low-key shadows. Sharing the dark vertical surface of the utility pole in the middle distance, C06O08 in a worn dark jumper, a Korean adult man, stands at three-quarter angle from the right. C06O08 appears as a rim-lit dark silhouette pinning the woman against the pole, his body volume preserved against the streetlight bloom, eyes fixed on C09O11 with a faint red glint in the deep shadow, his teeth pressing into her neck. Pinned against the same utility pole, C09O11 in a burgundy coat, a Korean adult woman, stands at three-quarter angle from the left, her body arched back with her neck exposed, looking off-screen with an unconscious slack expression. The utility pole rises behind them as a dark vertical spine. The dim yellow-white streetlight head points downward toward the couple and slightly toward the camera, causing a dirty-gold flare in the thick blue-gray mist. The slick, rain-darkened alley pavement reflects the streetlight bloom and the couple's dark legs in broken, shallow streaks across the lower frame.", "outfit_assignments": [ { "character_id": "C06", "outlook_id": "O08" }, { "character_id": "C09", "outlook_id": "O11" } ], "source_facts": [ "S#2 action: '전봇대에 뒤엉켜있는 두 사람... 남자의 붉은 눈동자, 목의 살점을 뜯는 이빨... 푹 힘없이 쓰러지는 여자.'", "Shot 1 description: Dense fog, two figures pressed together at a utility pole." ], "visual_inferences": [ "fixed_element foggy_streetlight_atmosphere: A damp, blue-gray fog envelops the area, illuminated by a dim, sickly gold sodium streetlight casting heavy, low-key shadows.", "Wide waist-height framing from 12 meters away places the wet alley pavement reflections in the lower foreground and the streetlight bloom in the upper frame." ], "creative_decisions": [ "Figures rendered as rim-lit compressed dark silhouettes in the middle distance to emphasize isolation and horror, preserving volume via rim light.", "Red glint on the man's eye preserved despite the deep shadow to highlight the creature dread.", "Both subjects share the explicit vertical surface of the pole to maintain spatial consistency." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "4d9959cd09e2b1c5", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "f9f5b669e6801efc", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 1, "visible_entities": [ "C06", "C09", "L02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 2, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the establishing start of the alley flow, the camera sits low at waist height at the foggy alley mouth, looking straight down a long, damp corridor toward a utility pole about 12 meters away. 한국인 남성 and 한국인 여성 are held as compressed dark silhouettes beneath the streetlight bloom in the middle distance, their bodies pressed together while the surrounding mist and wet pavement occupy most of the frame.", "lighting_mood": "Sodium streetlight diffuses into damp blue-gray fog, with the couple nearly black and a faint dirty-gold halo around the pole.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C06", "C09", "L02" ], "allowed_outlook_pairs": [ { "character_id": "C06", "outlook_id": "O08" }, { "character_id": "C09", "outlook_id": "O11" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "foggy_streetlight_atmosphere", "element_type": "environment_state", "character_name": "", "description": "A damp, blue-gray fog envelops the area, illuminated by a dim, sickly gold sodium streetlight. This specific lighting condition casts heavy, low-key shadows that wrap around the subjects and the environment.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "Sodium streetlight diffuses into damp blue-gray fog", "Low-key blue-gray shadow is cut by a sickly gold streetlight edge" ], "visual_inferences": [ "The ambient fog and specific gold streetlight color persist across different camera distances to maintain the uncanny, atmospheric mood of the alleyway." ], "creative_decisions": [ "Using the sickly gold edge to cut through the heavy blue-gray shadows highlights the visceral nature of the scene while maintaining a dark, mysterious tone." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [ { "scene_index": 2, "shot_index": 3, "description": "한국인 남성의 이빨이 한국인 여성의 목 피부를 강하게 파고들며 물어뜯는 순간, 여성의 고개가 무력하게 뒤로 젖혀진 클로즈업.", "keep_elements": [ { "label": "wet dark vertical surface of the utility pole", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "dense blue-gray fog illuminated by a sickly gold streetlight", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C06O08", "policy": "required" }, { "kind": "character_outlook", "id": "C09O11", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "ddaa15cf74aab44c" }, { "scene_index": 2, "heading": "S#2. 도시 뒷골목 / N", "beat_title": "치명적인 물림", "representative_moment": "C06 bites the neck of C09 under a dim streetlight, her head falling back.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "visceral close-up, shallow depth of field, low-key blue-gray shadow cut by sickly gold edge", "t2i_prompt": "Photorealistic cinematic still. A visceral close-up focusing on the neck and lower face of the same figures from the prior shot. C09O11 in a burgundy coat, an adult Korean woman, is framed in right profile, her head falling backward through the upper right of the frame, her eyes closed, visibly wounded with a pained expression. C06O08 in a worn black jumper, an adult Korean man with short dark hair exerting force, is partially visible in left profile, his jaw pressed forward, looking down, his mouth applying firm contact against her neck. A narrow, sharp red accent marks the contact point on her damp skin. The same damp, blue-gray fog envelops the immediate space, illuminated by the same dim, sickly gold sodium streetlight edge falling from above-left. Behind their neck area, [: a dark, damp, out-of-focus side edge of a utility pole] serves as a soft vertical frame-edge surface. Cold alley mist drifts through the tight space, catching shallow highlights on the skin and breath vapor, all wrapped in heavy, low-key blue-gray shadows.", "outfit_assignments": [ { "character_id": "C06", "outlook_id": "O08" }, { "character_id": "C09", "outlook_id": "O11" } ], "source_facts": [ "S#2 action: '남자의 붉은 눈동자, 목의 살점을 뜯는 이빨… 푹 힘없이 쓰러지는 여자.'", "shot description: male's teeth pressing into female's neck, female's head slackening backward" ], "visual_inferences": [ "fixed_element: The ambient fog and specific gold streetlight color persist across different camera distances to maintain the uncanny mood.", "utility pole edge positioned as a soft background anchor behind the subjects to establish location" ], "creative_decisions": [ "tight close-up framing to emphasize the visceral bite without graphic gore, using shadow and a narrow red accent", "changed location tag to L02 per modification request", "applied force-initiator wording for C06 to clarify the power imbalance in the close-up" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "539b634e5e7ad84d", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "8eaff9ee105f8bbf", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 2, "dependency_reason": "zoom_in_detail from previous shot in the same scene", "_shot_index": 3, "visible_entities": [ "C06", "C09" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 2, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At the mid close_observation point, the camera is less than half a meter from the couple at neck height, angled from 한국인 여성’s side so her head slackens backward through the upper right of frame. The composition is a visceral close-up of 한국인 남성’s mouth and her neck, but the bite is staged with shadow, damp skin texture, and a sharp red accent rather than graphic gore.", "lighting_mood": "Low-key blue-gray shadow is cut by a sickly gold streetlight edge and a narrow red accent at the bite point, restrained and uncanny.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C06", "C09" ], "allowed_outlook_pairs": [ { "character_id": "C06", "outlook_id": "O08" }, { "character_id": "C09", "outlook_id": "O11" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": null, "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "foggy_streetlight_atmosphere", "element_type": "environment_state", "character_name": "", "description": "A damp, blue-gray fog envelops the area, illuminated by a dim, sickly gold sodium streetlight. This specific lighting condition casts heavy, low-key shadows that wrap around the subjects and the environment.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "Sodium streetlight diffuses into damp blue-gray fog", "Low-key blue-gray shadow is cut by a sickly gold streetlight edge" ], "visual_inferences": [ "The ambient fog and specific gold streetlight color persist across different camera distances to maintain the uncanny, atmospheric mood of the alleyway." ], "creative_decisions": [ "Using the sickly gold edge to cut through the heavy blue-gray shadows highlights the visceral nature of the scene while maintaining a dark, mysterious tone." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 2, "shot_index": 1, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C06O08", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "ea074ff362ad6046" }, { "scene_index": 3, "heading": "S#3. 현장 / D", "beat_title": "뉴스에 보도되는 연쇄 변사체 수사 현장", "representative_moment": "짙은 안개 속, 수많은 형사 무리 사이에서 날카로운 눈빛으로 앞을 주시하며 사건 현장을 지휘하는 C02의 상반신 모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "observer slightly below eye level, medium-close upper-body frame, cool blue-gray tone", "t2i_prompt": "Photorealistic cinematic still. C02O03 in a dark field jacket, a Korean woman in her early 40s, commanding focus in the middle-center midground, looking toward the crime scene activity ahead of her. Captured from an observer slightly below eye level, the framing places her waist near the lower frame edge and her head in the upper frame. C02O03 is seen at a three-quarter angle from the right, leaning into command, looking off-screen with sharp eyes, her composed adult face focused. Korean officers in dark coats, middle-aged adults, occupy the left and right foreground of the shared damp field, their blurred shoulder edges facing the camera and forming a narrow visual corridor toward C02O03. In the background, a stretched, wet police line tape features its printed warning side facing the camera in broken yellow strips through the haze. Fog-soft daylight casts a cold blue-gray ambient tone, with small red evidence markers cutting through the dense mist. The entire visible face of the image carries a faint scanline-like broadcast texture.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O03" } ], "source_facts": [ "S#3. 현장 / D", "지역 뉴스 화면. 형사무리 속에 사건을 지휘하는 혜수의 모습이 보인다.", "짙은 안개 속, 무리 사이에서 날카로운 눈빛으로 앞을 주시하는 혜수의 상반신." ], "visual_inferences": [ "fog-soft daylight graded cold blue-gray applied from S1 atmosphere_reference", "Korean officers blurred in the foreground to emphasize C02's focus", "police tape placed in background to establish crime scene depth" ], "creative_decisions": [ "faint scanline texture to imply news footage without showing the device bezel", "low camera angle to empower C02's commanding presence", "color contrast between blue-gray fog and yellow/red crime scene markers" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "59cc9d60493d19fa", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "9964231b20e33895", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 1, "dependency_reason": "atmosphere_reference from S1_Shot3 to maintain the fog-soft daylight and cold blue-gray tone of the island.", "_shot_index": 3, "visible_entities": [ "C02", "C11" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 3, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera follows the close_observation end stage among detectives in the fog, slightly below eye level and pushing in to a medium-close upper-body frame of 혜수. Soft shoulders and dark coats of other investigators blur across the left and right foreground, while 혜수 stays centered, leaning into command with her eyes fixed sharply ahead rather than addressing the lens.", "lighting_mood": "Fog-soft daylight is graded cold blue-gray, with muted yellow police tape and small red evidence markers cutting through the damp field palette.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "looks_to_anchor", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "looks_toward", "gesture_target_label": "crime scene activity ahead of her", "label": "혜수 commanding focus", "screen_zone": "middle_center", "target_id": "C02", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02", "C11" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O03" }, { "character_id": "C11", "outlook_id": "O12" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C11", "policy_type": "identity_reference", "policy": "generic_descriptor_allowed", "reason": "visible_shot_count=1 — selected-shot 저빈도 주변 subject" } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [ { "scene_index": 1, "shot_index": 3, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C02O03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "2915f4f5362b7692" }, { "scene_index": 4, "heading": "S#4. 옥탑방 / D", "beat_title": "고립과 불안의 전경", "representative_moment": "어스름한 저녁 무렵, 인천 변두리 낡은 빌라의 옥탑방이 내려다보이는 스산한 부감 전경.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Steep bird's-eye view, wide establishing shot with cold blue-gray dusk lighting.", "t2i_prompt": "Photorealistic cinematic still. Deviating from the eye-level background reference framing to a steep bird's-eye view high above an outer-city residential district. Looking steeply downward, a weather-stained rooftop room sits slightly off-center on a flat terrace enclosed by a parapet wall. A water tank and an empty drying platform stand nearby, while the worn floor of the terrace and a stair landing are visible from above. The surrounding aging low-rise villa buildings and narrow, damp residential streets recede diagonally under the steep overhead angle. Dim dusk washes the entire district in cold blue-gray shadow, with thin evening haze settling over the rooftops and a few weak amber window lights puncturing the deep gloom.", "outfit_assignments": [], "source_facts": [ "S#4. 옥탑방 / D", "인천 변두리 부감, 오래된 빌라 옥탑방 전경." ], "visual_inferences": [ "dim dusk setting with cold blue-gray shadow as specified by staging", "narrow residential streets and low-rise villas surrounding the rooftop room" ], "creative_decisions": [ "deviated from the eye-level background reference framing to correctly apply the steeply downward bird's-eye view", "added weak amber window lights contrasting with the blue-gray dusk to emphasize neighborhood unease" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "anchor", "source_phrase": "A water tank and an empty drying platform stand nearby" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "the worn floor of the terrace" }, { "owned_token": "parapet wall", "usage_kind": "anchor", "source_phrase": "enclosed by a parapet wall" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "a weather-stained rooftop room sits slightly off-center" }, { "owned_token": "stair landing", "usage_kind": "anchor", "source_phrase": "and a stair landing are visible from above" }, { "owned_token": "terrace", "usage_kind": "anchor", "source_phrase": "on a flat terrace" }, { "owned_token": "utility post", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "anchor", "source_phrase": "A water tank and an empty drying platform stand nearby" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "59a7ee32c6e8e77a", "owned_hash": "e55ba06c295ec177", "camera_direction_hash": "0690e4b1807a18c4", "owned_usage_hash": "38e3726f6ed7fd9c", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "drying platform", "violating_phrase": "an empty drying platform stand nearby", "reason": "The prompt references the drying platform's position to establish the spatial layout in the high-angle shot, which aligns with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "the worn floor of the terrace", "reason": "The prompt mentions the floor of the terrace as visible from above to guide the overhead perspective, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "parapet wall", "violating_phrase": "enclosed by a parapet wall", "reason": "The prompt mentions the parapet wall as a structural boundary of the terrace to anchor the scene layout, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "rooftop room", "violating_phrase": "a weather-stained rooftop room sits slightly off-center", "reason": "The prompt references the rooftop room as the central subject of the new high-angle framing, which serves as a layout anchor, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "stair landing", "violating_phrase": "and a stair landing are visible from above", "reason": "The prompt references the stair landing to confirm its visibility from the high bird's-eye view, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "terrace", "violating_phrase": "on a flat terrace", "reason": "The prompt references the terrace as the base surface for the rooftop room, acting as a spatial anchor, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "water tank", "violating_phrase": "A water tank and an empty drying platform stand nearby", "reason": "The prompt mentions the water tank as a nearby structural reference in the composition, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "a few weak amber window lights", "reason": "The prompt references amber lights emanating from windows to define the dusk atmosphere. Although the producer declared this absent, it functions as an environmental anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "voiceover", "dependent_scene_index": -1, "dependency_reason": "독립적인 시공간의 외부 전경 묘사", "_shot_index": 1, "visible_entities": [ "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 4, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera begins very high above an outer-city residential district at dusk, looking steeply downward in a wide bird’s-eye view. Several aging low-rise villas and narrow lanes remain visible, while the worn rooftop room sits slightly off center as the crane begins its slow descent, gathering neighborhood unease toward that one roof.", "lighting_mood": "Dim dusk washes the district in cold blue-gray shadow, with a few weak amber window lights puncturing the gloom like ritual embers.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L03" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L03B02", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "door", "drying platform", "floor", "parapet wall", "rooftop room", "stair landing", "terrace", "utility post", "water tank", "window" ], "camera_reference": { "camera_position": "near number 3 (threshold stair landing zone), facing diagonally across number 1 (open rooftop terrace) toward number 2 (rooftop room footprint), with number 5 (iron entry door), number 6 (small facade window), and number 7 (water tank) in view", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm wide angle", "framing_notes": "Wide establishing composition of the old rooftop exterior; include the roof perimeter and utility fixtures, but keep all intrusion and dawn-only transient cues absent." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L03B02", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "d991c2a25d7f7fc7" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "심각하게 뉴스를 보는 민숙", "representative_moment": "C04가 어두운 표정으로 빛나는 TV 모니터를 뚫어지게 응시하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close portrait from slightly below eye level, tight focus on facial expression with high-contrast screen glow", "t2i_prompt": "Photorealistic cinematic still. Shot from slightly below eye level, C04O05 in an apron over gray knit homewear, a 40-year-old Korean woman with a mature face and dark hair, occupies the right-center of the frame in a close portrait. She is positioned at a three-quarter angle from the left, leaning forward, her worried face serious, eyes fixed on the television. On the left frame edge, the active screen face of the TV monitor is partially visible, angled toward her and partly toward the camera, glowing with unseen cold blue news light without readable detail. The cold TV blue carves the left contour of her cheek against the dim room, while [: a shallow, damp domestic blur of a closed window with faint condensation and wavering warm steam] fills the out-of-focus background, blending faint gray daylight with pervasive blue-gray shadows.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O05" } ], "source_facts": [ "민숙(수리영 엄마. 40세)", "앞치마를 두른 채 TV 뉴스를 뚫어지게 바라보는", "표정이 어둡다" ], "visual_inferences": [ "dim, damp atmosphere with blue-gray shadows", "cold blue light from TV illuminating her face", "out-of-focus background due to close framing" ], "creative_decisions": [ "placing the TV monitor edge on the extreme left to anchor the light source as directed", "close framing to emphasize Minsuk's serious, worried expression", "using cool gray daylight and muted gold warmth in the blurred background to contrast the harsh TV blue" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cooktop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photo frame", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pot", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "26b0ff37b7d2a5cb", "owned_hash": "82c97b402677f10a", "camera_direction_hash": "7b882b78b7323919", "owned_usage_hash": "3821487005ab2c0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "N/A", "_shot_index": 3, "visible_entities": [ "C04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Inside the rooftop room, the camera inherits the living-area close_observation flow from slightly below eye level, angled between the TV monitor and 민숙’s face. The frame tightens to a close portrait: her serious face occupies the right-center, the TV glow skims her cheek from the left, and the rest of the room falls into shallow, damp domestic blur.", "lighting_mood": "Cold TV blue carves 민숙’s worried face against the dim room, with faint gray daylight and muted gold kitchen warmth barely surviving in the shadows.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C04" ], "allowed_outlook_pairs": [ { "character_id": "C04", "outlook_id": "O05" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L04B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "minsuk_appearance", "element_type": "character_state", "character_name": "민숙", "description": "A 40-year-old Korean woman wearing an apron over casual indoor clothing. Her facial features, hairstyle, and overall upper body attire remain consistent as she transitions from sitting in the living area to standing near the entryway.", "applies_to_shots": [ 3, 11 ], "element_scope": "full", "source_facts": [ "민숙(수리영 엄마. 40세)", "앞치마를 두른 채", "한국인 중년 여성" ], "visual_inferences": [ "casual domestic outfit worn under the apron", "consistent hairstyle suitable for a middle-aged mother doing housework" ], "creative_decisions": [ "maintaining the apron to emphasize her domestic routine before her afternoon shift" ], "confidence": "high" }, { "element_id": "rooftop_room_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The interior of the rooftop room maintains a slightly dim, damp atmosphere. Cool, faint gray daylight from a small window blends with pervasive blue-gray shadows and subtle hints of warm ambient light from the kitchen area.", "applies_to_shots": [ 3, 11, 12 ], "element_scope": "full", "source_facts": [ "거실에 난 작은 창으로 햇살에 가늘게 비친다", "김이 피어오르고" ], "visual_inferences": [ "dampness in the air due to boiling water and steam", "limited natural light creating predominantly cool, shadowed areas" ], "creative_decisions": [ "using a muted blue-gray and gray color palette for the shadows to reflect the cramped, somewhat uneasy domestic mood" ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C04O05", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "9fdd73c48861fc6a" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "수리영의 외출 준비와 다가감", "representative_moment": "C01O01 leans in with her eyes closed to smell C04O05's shoulder near the cramped entryway.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Handheld camera at shoulder height, medium framing with muted blue-gray softness and a faint warm-gold rim light", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a casual hooded jacket, a Korean young woman, appears as 수리영 leaning close in the middle-left foreground, reaching for 민숙 shoulder, positioned in right profile with eyes closed as her nose hovers near the fabric. Sharing the cramped entryway threshold, C04O05 in a gray knit sleeve fabric under an apron, a Korean middle-aged woman, stands as 민숙 shoulder and profile in the middle-center foreground, angled at a three-quarter angle from the left with her shoulder held still, her eyes fixed on C01O01. The inner side of the scuffed door and dim shadow near the doorway bracket the pair vertically, placing their torsos and heads in the upper-mid frame with the top surfaces of shoes visible obliquely below them in the lower edge. Behind them, muted blue-gray softness from the hazy window light and distant steam fills the damp room, settling over the kitchen counter and sink, while a faint warm-gold edge light traces both women's shoulders.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C04", "outlook_id": "O05" } ], "source_facts": [ "수리영(21. 애완동물미용학과 2년)이 방문을 열고 기지개를 켜며 나오자", "가벼운 차림새. 수리영 신발을 신으며", "(엄마의 냄새를 맡으며) 아, 좋다.", "Shot 11: 수리영이 한국인 중년 여성의 어깨 부근에 코를 가까이 댄 옆모습" ], "visual_inferences": [ "C04 wearing an apron over casual indoor clothing, maintaining continuity from the prior beat", "dim, damp rooftop room atmosphere with cool gray daylight and blue-gray shadows", "C04 holding her shoulder still to let C01 lean in", "gray knit sleeve fabric exposed clearly on C04 for the next shot's forward-zoom continuity" ], "creative_decisions": [ "Handheld camera at shoulder height for an intimate but constrained perspective on their side profiles", "Muted blue-gray softness with a warm-gold rim light to reflect restrained warmth and underlying unease", "Framing the two women tightly against the scuffed door to compress them into domestic closeness" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cooktop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "anchor", "source_phrase": "kitchen counter" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "scuffed door and dim shadow" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photo frame", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pot", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "anchor", "source_phrase": "and sink" }, { "owned_token": "stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "hazy window light" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d7d8907baabe4fcb", "owned_hash": "82c97b402677f10a", "camera_direction_hash": "aa8de64acad56a28", "owned_usage_hash": "1e3fb5a828840d57", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "The inner side of the scuffed door and dim shadow near the doorway bracket the pair vertically", "reason": "The scuffed door is used strictly as a vertical framing boundary for composition, which matches the producer's declaration as an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "counter", "violating_phrase": "settling over the kitchen counter", "reason": "The kitchen counter is referenced as an environmental surface where background steam settles, verifying the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "sink", "violating_phrase": "settling over the kitchen counter and sink", "reason": "The sink is used as a spatial reference for atmospheric steam settling in the background, confirming its usage as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "hazy window light", "reason": "The window is mentioned solely as the source of background light, aligning with the anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "atmosphere_reference (Shot 3) to maintain the dim, damp rooftop room lighting tone", "_shot_index": 11, "visible_entities": [ "C01", "C04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 11 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Near the doorway at shoulder height, the handheld camera stays close on the side profiles of 수리영 and 민숙, preserving the intimate_departure mid-flow. 수리영 leans in from frame left with her nose almost touching 민숙’s shoulder at frame center, while 민숙 remains slightly turned away with restrained warmth, the cramped entryway compressing them into a quiet domestic closeness.", "lighting_mood": "Natural window light and kitchen steam are muted into blue-gray softness, with a faint warm-gold edge on their shoulders to keep the moment tender but uneasy.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "shared_space_relation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "reaches_for", "gesture_target_label": "민숙 shoulder", "label": "수리영 leaning close", "screen_zone": "middle_left", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "민숙 shoulder and profile", "screen_zone": "middle_center", "target_id": "C04", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "C04" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C04", "outlook_id": "O05" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B01", "reference_usage": "exact_background", "owned_objects": [ "TV", "chair", "cooktop", "counter", "door", "faucet", "lamp", "photo frame", "pot", "sink", "stand", "table", "window" ], "camera_reference": { "camera_position": "near number 1 (entrance threshold zone), facing diagonally across number 2 (shared living dining zone) toward number 14 (sink counter) and number 18 (television).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "Include numbers 14, 16, 18, and 28 clearly; overlay use includes number 29 on the resident and number 30 as the active television state." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "minsuk_appearance", "element_type": "character_state", "character_name": "민숙", "description": "A 40-year-old Korean woman wearing an apron over casual indoor clothing. Her facial features, hairstyle, and overall upper body attire remain consistent as she transitions from sitting in the living area to standing near the entryway.", "applies_to_shots": [ 3, 11 ], "element_scope": "full", "source_facts": [ "민숙(수리영 엄마. 40세)", "앞치마를 두른 채", "한국인 중년 여성" ], "visual_inferences": [ "casual domestic outfit worn under the apron", "consistent hairstyle suitable for a middle-aged mother doing housework" ], "creative_decisions": [ "maintaining the apron to emphasize her domestic routine before her afternoon shift" ], "confidence": "high" }, { "element_id": "rooftop_room_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The interior of the rooftop room maintains a slightly dim, damp atmosphere. Cool, faint gray daylight from a small window blends with pervasive blue-gray shadows and subtle hints of warm ambient light from the kitchen area.", "applies_to_shots": [ 3, 11, 12 ], "element_scope": "full", "source_facts": [ "거실에 난 작은 창으로 햇살에 가늘게 비친다", "김이 피어오르고" ], "visual_inferences": [ "dampness in the air due to boiling water and steam", "limited natural light creating predominantly cool, shadowed areas" ], "creative_decisions": [ "using a muted blue-gray and gray color palette for the shadows to reflect the cramped, somewhat uneasy domestic mood" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 3, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [ { "scene_index": 5, "shot_index": 12, "description": "수리영의 시선 끝에 걸린, 한국인 중년 여성의 손목 피부에 그려진 작은 원 모양의 표식 클로즈업.", "keep_elements": [ { "label": "the scuffed doorframe and dim shadow near the doorway", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "the gray knit sleeve fabric", "kind": "static_prop", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "character_outlook", "id": "C04O05", "policy": "required" }, { "kind": "background", "id": "L04B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "8c1b0e64018986b0" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "손목의 표식", "representative_moment": "수리영의 시선 끝에 머문, 민숙(C04)의 손목 피부에 선명하게 드러난 작은 원형 표식(P08) 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-height insert framing with sharp focal point, cool daylight edge and muted blue-gray shadows", "t2i_prompt": "Photorealistic cinematic still. At chest height beside the door in the same dim, damp rooftop room, the camera settles in a tight detail view isolating a wrist relaxed downward in the center-left of the frame, with the torso and head implied beyond the upper edges. P08, a sharp, dark, circle-shaped mark, is unusually clear on the skin surface facing the camera directly. The wrist belongs to C04 in a soft, slightly worn sleeve cuff, an Asian woman in her 40s. A narrow cool daylight edge reveals P08, while the surrounding wall shadow falls into a damp, scuffed blue-gray, creating a dark negative space behind the wrist. A faint ominous red-brown undertone is visible in the skin texture. The cuff edge faces the camera diagonally, framing the wrist without covering the mark. The flat side wall recedes behind the wrist at a shallow angle, keeping the same background softly compressed around the focal point.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O05" } ], "source_facts": [ "Shot description: 수리영의 시선 끝에 걸린, 한국인 중년 여성의 손목 피부에 그려진 작은 원 모양의 표식 클로즈업.", "씬 원문: 민숙 손목에 보이는 작은 원(이하 표식)을 본다. 오늘따라 유난히 또렷해 보이지만" ], "visual_inferences": [ "fixed_element rooftop_room_atmosphere: dim, damp atmosphere with cool gray daylight and blue-gray shadows", "chest-height camera angle focusing on the wrist, implying the rest of the body is outside the frame" ], "creative_decisions": [ "insert framing scale to sharply isolate the mark and the wrist", "damp blue-gray shadows with a red-brown skin undertone to emphasize the eerie, occult mood of the mark", "shallow angle for the receding wall to provide dark negative space without distracting from P08" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "prop", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cooktop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "beside the door" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photo frame", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pot", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "0479bd5c3b07f4af", "owned_hash": "82c97b402677f10a", "camera_direction_hash": "33479441e1a7290b", "owned_usage_hash": "0dcc5685c8c494c7", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "beside the door", "reason": "The prompt references 'door' solely as a spatial anchor point ('beside the door') for the camera height, matching the producer's declaration as an anchor with no intention to redraw.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "zoom_in_detail from previous shot in the same scene to preserve the exact rooftop room atmosphere and spatial anchor", "_shot_index": 12, "visible_entities": [ "C04", "P08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 12 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera completes the reveal tilt from 민숙’s shoulder down to her wrist, settling just below chest level in a tight detail view from 수리영’s side of the doorway. The small circle-shaped mark on the wrist is the sharp focal point in the center-left of frame, while skin, sleeve edge, and entryway shadow remain minimal and softly compressed around it.", "lighting_mood": "A narrow cool daylight edge reveals the wrist mark, while surrounding shadows fall into damp blue-gray with a faint ominous red-brown undertone in the skin texture.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C04", "P08" ], "allowed_outlook_pairs": [ { "character_id": "C04", "outlook_id": "O05" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C04", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The frame isolates 민숙’s wrist and mark rather than her full face or outfit; base identity is enough and an outlook reference could overconstrain the insert." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B01", "reference_usage": "exact_background", "owned_objects": [ "TV", "chair", "cooktop", "counter", "door", "faucet", "lamp", "photo frame", "pot", "sink", "stand", "table", "window" ], "camera_reference": { "camera_position": "near number 1 (entrance threshold zone), facing diagonally across number 2 (shared living dining zone) toward number 14 (sink counter) and number 18 (television).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "Include numbers 14, 16, 18, and 28 clearly; overlay use includes number 29 on the resident and number 30 as the active television state." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "rooftop_room_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The interior of the rooftop room maintains a slightly dim, damp atmosphere. Cool, faint gray daylight from a small window blends with pervasive blue-gray shadows and subtle hints of warm ambient light from the kitchen area.", "applies_to_shots": [ 3, 11, 12 ], "element_scope": "full", "source_facts": [ "거실에 난 작은 창으로 햇살에 가늘게 비친다", "김이 피어오르고" ], "visual_inferences": [ "dampness in the air due to boiling water and steam", "limited natural light creating predominantly cool, shadowed areas" ], "creative_decisions": [ "using a muted blue-gray and gray color palette for the shadows to reflect the cramped, somewhat uneasy domestic mood" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 11, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character", "id": "C04", "policy": "required" }, { "kind": "background", "id": "L04B01", "policy": "required" }, { "kind": "prop", "id": "P08", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [ { "contract_id": "rc_001", "scope": { "scene_index": 5, "shot_indices": [ 12 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P08", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] } ], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "757e036908984236" }, { "scene_index": 6, "heading": "S#6. 해안가 도로 / D", "beat_title": "해안가 질주와 생기", "representative_moment": "수리영이 바닷바람을 맞으며 해안가 도로를 자전거로 활기차게 달리고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium tracking shot from front-left at upper-torso height, bright coastal daylight graded into cool blue-gray freshness", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a blue-gray hooded jacket, a young Korean woman in her early 20s, occupies the middle-center midground riding forward, moving toward the camera along the coastal road. She is at three-quarter angle from the right, looking off-screen, her youthful oval face and soft facial features visible. Both hands are firmly gripping the curved front side of the bicycle handlebars in the lower third of the frame, the bicycle frame and a slice of the front wheel visible below. Captured in mid-pedal forward motion, her long natural black hair flows back in the direction opposite to motion. Her upper body is centered, leaving deliberate space behind her. [L05: A deep blue coastal road, dry and open, the road surface running from the lower frame toward the horizon behind her, filled with coastal air and slightly misted blue sea haze]. Bright coastal daylight graded into cool blue-gray freshness illuminates the scene, casting soft gold highlights on her hair and skin.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "shot description: 짙은 푸른빛이 도는 해안가 도로, 자전거 핸들을 쥐고 앞으로 나아가는 수리영의 상반신, 바람에 머리카락이 뒤로 휘날리고 있다.", "staging character_angles: 수리영 three_quarter_right, gripping bicycle handlebars, gaze=off_screen, state=alive", "entity traits: C01 youthful oval face, long natural black hair, soft facial features" ], "visual_inferences": [ "bicycle frame and partial front wheel visible below to ground the moving subject without floating effect", "mid-pedal posture to preserve forward motion direction" ], "creative_decisions": [ "deliberate space left behind the flowing hair to emphasize the coastal wind and forward speed", "soft gold highlights applied to hair and skin to contrast with the cool blue-gray environment" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "868ba83093b9887c", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "d913f30a2bb541bd", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "독립적인 외부 장면", "_shot_index": 1, "visible_entities": [ "C01", "L05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 6, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The tracking camera rides about 1.5 meters in front of 수리영 at upper-torso height, slightly off the bicycle’s front-left, matching her speed along the coastal road. The frame is a medium moving shot with both hands on the handlebars in the lower third, her upper body centered, and deliberate space behind her wind-blown hair so the road’s deep blue strip and sea air remain visible.", "lighting_mood": "Bright coastal daylight is graded into cool blue-gray freshness, with soft gold highlights on hair and skin and no harsh horror contrast yet.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "camera along coastal road", "label": "수리영 riding forward", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L05" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "8d876b56da36a0d8" }, { "scene_index": 7, "heading": "S#7. 유기견 보호원 / D", "beat_title": "사나운 맹견을 제압하는 수리영의 무표정한 응시", "representative_moment": "C01이 차가운 철창 너머로 무표정하게 맹견을 응시하는 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level tight close frame in shallow focus, muted blue-gray tones", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a slightly damp casual hoodie, a Korean woman in her early 20s with a youthful oval face, long natural black hair, and soft facial features, positioned at three-quarter angle from the left with her chin slightly lowered. She sits slightly right of center in a tight close frame, set just beyond the foreground obstruction, wearing a pale, stoic, and unreadable expression. Her blank stare is aimed off-screen, eyes fixed on an unseen subject in the shadowed kennel interior. [: Damp metal cage bars cutting vertical black lines across the near foreground between the lens and her face, featuring faint rust-red stains and dull metallic highlights, while the wet-floored concrete corridor recedes in soft shallow background focus]. Cold, low-key daylight leaks through the space, bathing her face in muted blue-gray tones that enhance the stark contrast of her pale complexion against the heavy shadows, amplifying a subtle occult-horror pressure.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Shot description: 무표정한 얼굴로 창살 너머를 빤히 응시하는 수리영의 얼굴", "S#7 action: 맹견의 눈을 노려보는 수리영" ], "visual_inferences": [ "fixed_element suriyoung_shelter_appearance: Pale, stoic, unreadable expression, damp casual work clothes", "fixed_element kennel_corridor_environment: Dimly lit damp shelter corridor, metal cage bars with rust-red stains" ], "creative_decisions": [ "Tight close framing with foreground bars to create a trapped, ritual-like grid restricting her face", "Muted blue-gray lighting to emphasize occult-horror pressure and contrast with her pale skin", "Replaced forbidden L01 ID with L02 for the required frame-edge surface description block" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "77ee7c4910799346", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "43835f54d7d9e5f7", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct scene dependency.", "_shot_index": 2, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 7, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At Suriyoung's eye level from the corridor side, the camera holds completely still in a tight close frame, with damp cage bars cutting vertical black lines across the near foreground and her face set about two feet beyond them. The composition is an observer's direct view, not her POV; Suriyoung sits slightly right of center in shallow focus, her blank stare aimed past the bars toward the kennel interior.", "lighting_mood": "Low-key daylight leaks through the shelter in muted blue-gray tones, with faint rust-red stains and dull metal highlights sharpening the occult-horror pressure.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": null, "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "suriyoung_shelter_appearance", "element_type": "character_state", "character_name": "Suriyoung", "description": "A young Korean woman with a pale, stoic, and unreadable expression. She is wearing practical, slightly damp casual work clothes suitable for physical shelter chores like bathing dogs. Her hair is styled practically to keep it out of her face.", "applies_to_shots": [ 2, 6 ], "element_scope": "close", "source_facts": [ "무표정한 얼굴로 창살 너머를 빤히 응시하는 수리영의 얼굴", "수리영 (character in dog shelter)" ], "visual_inferences": [ "Wears practical work attire due to upcoming dog grooming and bathing tasks", "Hair tied back or kept neat for manual labor" ], "creative_decisions": [ "Pale complexion to enhance the stark contrast with the cold blue-gray kennel lighting" ], "confidence": "medium" }, { "element_id": "kennel_corridor_environment", "element_type": "environment_state", "character_name": "", "description": "A dimly lit, damp dog shelter corridor lined with rows of metal cages. The environment is bathed in cold, low-key daylight that creates heavy, muted blue-gray shadows. The metal cage bars feature distinct red-brown rust stains and dull metallic highlights.", "applies_to_shots": [ 2, 6 ], "element_scope": "full", "source_facts": [ "damp cage bars cutting vertical black lines", "Low-key daylight leaks through the shelter in muted blue-gray tones", "faint rust-red stains and dull metal highlights", "rows of caged dogs recede behind her", "damp corridor" ], "visual_inferences": [ "Surfaces appear slightly wet and grimy due to ongoing shelter cleaning activities", "Concrete or industrial flooring typical of animal shelters" ], "creative_decisions": [ "Emphasizing the contrast between the cold blue-gray ambient light and the red-brown rust to evoke a subtle occult-horror pressure" ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "1df7cffb61358621" }, { "scene_index": 7, "heading": "S#7. 유기견 보호원 / D", "beat_title": "불안한 침묵 속 통화 연결음", "representative_moment": "수리영(C01)이 보호소 복도에서 서늘한 표정으로 귀에 핸드폰을 댄 채 응답 없는 전화를 기다린다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Chest-to-eye height observer composition, cold blue-gray kennel lighting with dull metallic and rust-red contrast", "t2i_prompt": "Photorealistic cinematic still. Viewed from chest-to-eye height, C01O01 in slightly damp practical work clothes, a Korean woman in her early 20s with a youthful oval face and soft facial features, stands at a three-quarter angle from the right. Her hair is tied back practically, a mobile phone pressed tightly to her ear, the phone arm occupying the upper side of her framed torso. Her pale, stoic face gazes into the distance. Behind her, the dimly lit, damp dog shelter corridor stretches away, forming a narrowing tunnel. Rows of closed metal cage bars with red-brown rust stains and dull metallic highlights recede in diminishing perspective along the side walls. Cold, low-key kennel daylight pools unevenly on her face, casting heavy, muted blue-gray shadows across the quiet, anxious negative space of the corridor.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 민숙에게 연락을 취해보지만, 받지 않는다.", "shot description: 수리영이 핸드폰을 귀에 바짝 댄 채 앞을 응시하는 정지 구도." ], "visual_inferences": [ "fixed_element suriyoung_shelter_appearance: pale, stoic expression, damp casual work clothes, hair tied back", "fixed_element kennel_corridor_environment: dimly lit damp corridor, metal cages, cold low-key daylight, blue-gray shadows, rust stains", "standing at a three-quarter angle from the right, distant gaze" ], "creative_decisions": [ "Chest-to-eye height framing to clearly capture her torso and phone gesture", "Cage bars receding in diminishing perspective to emphasize the psychological isolation of her unanswered call", "Uneven pooling of cold light on her face to contrast with the dark, heavy blue-gray shadows of the environment" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1d7cd1d8e1653070", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "21ce6b5877e8f2c0", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 7, "dependency_reason": "동일 씬 내 이전 샷(Shot 2)의 보호소 공간 톤(차가운 조명, 철창 등) 연속성 유지", "_shot_index": 6, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 7, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Chest-to-eye height in the kennel corridor, about four feet in front of Suriyoung, the camera stays locked in a slightly tighter centered medium frame than the wider corridor setup. Her phone hand occupies the upper side of her body while rows of caged dogs recede behind her on both sides, turning the unanswered call into the only still point in the damp corridor.", "lighting_mood": "Cold kennel light pools unevenly on her face, with blue-gray shadows and small red-brown rust accents in the cage hardware.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "suriyoung_shelter_appearance", "element_type": "character_state", "character_name": "Suriyoung", "description": "A young Korean woman with a pale, stoic, and unreadable expression. She is wearing practical, slightly damp casual work clothes suitable for physical shelter chores like bathing dogs. Her hair is styled practically to keep it out of her face.", "applies_to_shots": [ 2, 6 ], "element_scope": "close", "source_facts": [ "무표정한 얼굴로 창살 너머를 빤히 응시하는 수리영의 얼굴", "수리영 (character in dog shelter)" ], "visual_inferences": [ "Wears practical work attire due to upcoming dog grooming and bathing tasks", "Hair tied back or kept neat for manual labor" ], "creative_decisions": [ "Pale complexion to enhance the stark contrast with the cold blue-gray kennel lighting" ], "confidence": "medium" }, { "element_id": "kennel_corridor_environment", "element_type": "environment_state", "character_name": "", "description": "A dimly lit, damp dog shelter corridor lined with rows of metal cages. The environment is bathed in cold, low-key daylight that creates heavy, muted blue-gray shadows. The metal cage bars feature distinct red-brown rust stains and dull metallic highlights.", "applies_to_shots": [ 2, 6 ], "element_scope": "full", "source_facts": [ "damp cage bars cutting vertical black lines", "Low-key daylight leaks through the shelter in muted blue-gray tones", "faint rust-red stains and dull metal highlights", "rows of caged dogs recede behind her", "damp corridor" ], "visual_inferences": [ "Surfaces appear slightly wet and grimy due to ongoing shelter cleaning activities", "Concrete or industrial flooring typical of animal shelters" ], "creative_decisions": [ "Emphasizing the contrast between the cold blue-gray ambient light and the red-brown rust to evoke a subtle occult-horror pressure" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 7, "shot_index": 2, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "6037fe444b6691ab" }, { "scene_index": 8, "heading": "S#8. 번화가·대형 마트 / 해질 무렵", "beat_title": "엇갈린 만남과 의문", "representative_moment": "Suriyoung stands across from a female market employee in the aisle, asking about her mother with a questioning expression.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Shoulder-height perpendicular side two-shot, medium framing, sickly blue-gray fluorescent lighting", "t2i_prompt": "Photorealistic cinematic still. At shoulder-height beside the supermarket aisle, a perpendicular side two-shot captures C01O01 in a blue-gray casual hoodie, a young Korean woman, and a Korean middle-aged woman in a worker uniform sharing the same narrow aisle floor across from each other. C01O01 stands in right profile, leaning toward the employee, her eyes fixed on the woman, her face clearly visible holding a questioning expression with slightly wide eyes. The Korean middle-aged woman stands in left profile, one hand resting on the horizontal gondola shelf, her eyes fixed on C01O01. The background consists of the gondola shelf stocked with various product units featuring noticeable red and gold packaging, forming stacked bands behind the characters. Sickly fluorescent ceiling light strips overhead flatten the scene into blue-gray realism, the dull floor beneath their feet reflecting a weak green-blue cast.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영이 마트 직원 복장의 한국인 중년 여성을 마주보고 서 있는 측면 구도", "수리영: profile_right, leaning toward employee, state=alive", "마트 여성 직원: profile_left, hand resting on shelf, gaze=looks_at_character→C01", "Sickly fluorescent light flattens the market into blue-gray realism", "front-facing product labels and price tags are visible to camera in stacked bands behind the characters" ], "visual_inferences": [ "fixed_element: The interior of a modern supermarket aisle illuminated by a cold, blue-gray fluorescent wash", "fixed_element: Suriyoung wears casual, everyday clothing suitable for riding a bicycle", "Shared anchor on the floor establishes the tight spatial tension between the two characters in the aisle" ], "creative_decisions": [ "Positioned both figures in strict left/right profile to emphasize the perpendicular two-shot constraint", "Highlighted the red and gold packaging accents to provide contrast against the predominantly blue-gray and cold palette", "Ensured Suriyoung's face is clearly visible with wide eyes to satisfy the forward zoom detail target" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "CCTV camera", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ceiling light", "usage_kind": "anchor", "source_phrase": "Sickly fluorescent ceiling light strips overhead" }, { "owned_token": "checkout counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "display island", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "sharing the same narrow aisle floor across from each other" }, { "owned_token": "gondola shelf", "usage_kind": "anchor", "source_phrase": "one hand resting on the horizontal gondola shelf" }, { "owned_token": "product", "usage_kind": "anchor", "source_phrase": "stocked with various product units" }, { "owned_token": "refrigerated case", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e0b35c8f5f5efbe0", "owned_hash": "29fdf3a42031ea40", "camera_direction_hash": "ea17407082e91fe2", "owned_usage_hash": "939efcb82adba272", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "ceiling light", "violating_phrase": "Sickly fluorescent ceiling light strips overhead", "reason": "The prompt references the overhead ceiling lights to describe the fluorescent lighting style of the scene, acting as an anchor in alignment with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "sharing the same narrow aisle floor", "reason": "The prompt mentions the aisle floor to position the characters and describe lighting reflection, acting as an anchor consistent with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "gondola shelf", "violating_phrase": "one hand resting on the horizontal gondola shelf", "reason": "The prompt references the gondola shelf as a physical anchor for the character's hand and to describe the background, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "product", "violating_phrase": "stocked with various product units", "reason": "The products on the shelves are mentioned to describe the color and detail of the background elements, which is an anchoring reference aligning with the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 3, "visible_entities": [ "C01", "L08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 8, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Shoulder-height at the side of the aisle, roughly two meters away, the camera holds a clean perpendicular side two-shot of Suriyoung and the middle-aged female market employee facing each other. Product shelves remain steady as horizontal layers behind them, while the frame tightens just enough to make their profile exchange feel boxed in by fluorescent retail order.", "lighting_mood": "Sickly fluorescent light flattens the market into blue-gray realism, with small red and gold package accents cutting through the dusk-toned palette.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L08" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L08B01", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "ceiling light", "checkout counter", "display island", "door", "floor", "gondola shelf", "product", "refrigerated case", "shelf", "sign", "window" ], "camera_reference": { "camera_position": "near number 6 (front automatic entrance), facing diagonally across number 2 (entrance threshold) toward number 12 (central gondola shelf bank), number 15 (aisle end display island), and number 10 (checkout counter row)", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "Show the busy front-to-central aisle relationship; include number 18 as the temporary stocking boxes overlay beside the shelves while keeping the fixed CCTV position at number 16 readable if possible." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "supermarket_aisle_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of a modern supermarket aisle illuminated by a cold, blue-gray fluorescent wash. The background consists of horizontal product shelves stocked with various retail items that feature noticeable red and gold packaging.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Sickly fluorescent light flattens the market into blue-gray realism", "small red and gold package accents", "blurred wall of products", "cold fluorescent wash", "blurred red-gold packaging" ], "visual_inferences": [ "Standard metallic retail shelving and layout typical of modern Korean supermarkets." ], "creative_decisions": [ "Using the red and gold packaging as subtly flickering warning marks in the background to visually support the growing tension." ], "confidence": "high" }, { "element_id": "suriyoung_general_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman standing inside the brightly lit supermarket. Her skin appears somewhat pale under the harsh, cold fluorescent lighting, and she wears casual, everyday clothing suitable for riding a bicycle through the neighborhood.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "수리영", "능숙하게 자전거를 모는 수리영", "skin pale against muted blue-gray shadows" ], "visual_inferences": [ "Her attire is practical and casual, fitting a quick late-afternoon bike ride to a local store." ], "creative_decisions": [ "Highlighting her pale complexion against the cold environment to emphasize her sudden confusion and subtle vulnerability." ], "confidence": "medium" } ], "previous_shot_refs": [], "forward_zoom_targets": [ { "scene_index": 8, "shot_index": 5, "description": "눈을 동그랗게 뜨고 의아한 표정으로 멈춘 수리영의 얼굴.", "keep_elements": [ { "label": "supermarket product shelves stocked with red and gold labeled packages", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "aisle floor showing dull reflection of blue-gray fluorescent light", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "background", "id": "L08B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "3de90b53251e13ea" }, { "scene_index": 8, "heading": "S#8. 번화가·대형 마트 / 해질 무렵", "beat_title": "엇갈린 만남과 당혹감", "representative_moment": "직원의 말을 듣고 의아한 표정으로 눈을 동그랗게 뜬 수리영(C01)의 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level close framing with a cold, blue-gray fluorescent wash", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a casual blue-gray hooded jacket, a Korean woman in her early 20s, occupies the middle-center foreground. The close framing isolates the same figure's face and shoulders at a front three-quarter angle from the left, her shoulders pulled back. Her widened eyes are fixed on an unseen figure just outside the frame. The same room from the prior shot is preserved under a cold overhead fluorescent wash, leaving her skin pale with muted blue-gray shadows beneath her eyes. [: The background directly behind her is a completely blurred wall of horizontal product shelves, where out-of-focus red and gold packaging appears as soft colored bands. The narrow side face of a shelf endcap is visible along the frame edge.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영", "(의아) 네? 누구랑요?", "cold fluorescent wash", "blurred red-gold packaging" ], "visual_inferences": [ "fixed_element suriyoung_general_appearance: casual clothing, pale skin under harsh lighting", "fixed_element supermarket_aisle_environment: blue-gray fluorescent wash, horizontal product shelves" ], "creative_decisions": [ "Tight close-up to strictly isolate her widened eyes and reaction", "Completely blurred background elements to draw visual focus exclusively to the character's face" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "CCTV camera", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ceiling light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "checkout counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "display island", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gondola shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "product", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "refrigerated case", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "20de70817908e3cc", "owned_hash": "29fdf3a42031ea40", "camera_direction_hash": "e236762e4d13c00b", "owned_usage_hash": "081e2fa282d2d646", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 8, "dependency_reason": "Shot 3의 대형 마트 내부 공간 및 차가운 형광등 조명 연속성 유지 (zoom_in_detail)", "_shot_index": 5, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 8, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Eye-level about 0.8 meters from Suriyoung near the aisle edge, the camera has already landed and now holds still on a close front three-quarter view of her widened eyes. She sits slightly off-center against a blurred wall of products, with the unseen male employee's line pulling her gaze just outside the frame rather than into the lens.", "lighting_mood": "A cold fluorescent wash leaves her skin pale against muted blue-gray shadows, while blurred red-gold packaging flickers like warning marks behind her.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L08B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "supermarket_aisle_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of a modern supermarket aisle illuminated by a cold, blue-gray fluorescent wash. The background consists of horizontal product shelves stocked with various retail items that feature noticeable red and gold packaging.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Sickly fluorescent light flattens the market into blue-gray realism", "small red and gold package accents", "blurred wall of products", "cold fluorescent wash", "blurred red-gold packaging" ], "visual_inferences": [ "Standard metallic retail shelving and layout typical of modern Korean supermarkets." ], "creative_decisions": [ "Using the red and gold packaging as subtly flickering warning marks in the background to visually support the growing tension." ], "confidence": "high" }, { "element_id": "suriyoung_general_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman standing inside the brightly lit supermarket. Her skin appears somewhat pale under the harsh, cold fluorescent lighting, and she wears casual, everyday clothing suitable for riding a bicycle through the neighborhood.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "수리영", "능숙하게 자전거를 모는 수리영", "skin pale against muted blue-gray shadows" ], "visual_inferences": [ "Her attire is practical and casual, fitting a quick late-afternoon bike ride to a local store." ], "creative_decisions": [ "Highlighting her pale complexion against the cold environment to emphasize her sudden confusion and subtle vulnerability." ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 8, "shot_index": 3, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "1a1b0db01c68222d" }, { "scene_index": 9, "heading": "S#9. 해안가 / N", "beat_title": "밤의 해안도로를 질주하는 수리영", "representative_moment": "짙은 안개 속 자전거를 타고 급히 해안도로를 달려가는 C01의 역동적인 실루엣.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Waist-height tracking wide shot, underexposed blue-gray coastal darkness with sodium-gold rim lighting", "t2i_prompt": "Photorealistic cinematic still. In the middle-center midground, C01O01 in a casual hoodie, a Korean woman in her early 20s on a bicycle, moves toward the foggy coastal road ahead situated in the upper-right background. Captured from a waist-height tracking position, she appears at a three-quarter angle from the right, leaning forward with one bent knee pressing the pedal downward in a mid-action stroke. The bicycle frame and front wheel are visible below her. She is gazing into the distance, her natural black hair flowing back in the direction opposite to motion. The scene is enveloped in underexposed blue-gray coastal darkness, carrying the dim cool ambient tone from the prior scene, with thick sea fog drifting around her. A distant sodium-gold haze provides a rim-lit outline along her back, shoulders, and arms, preserving body volume against the dark mist. A damp dark guardrail and wet asphalt smear laterally in the lower frame, reflecting thin metallic glints.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#9 action: Suriyoung riding her bicycle hastily toward home along the coastal road.", "shot description: thick fog, night, leaning forward on the saddle, mid-action downward pedal stroke." ], "visual_inferences": [ "waist-height camera framing captures the torso, bent knee, and bicycle frame prominently without forcing full head-to-toe to comply with spatial consistency rules", "blurring of the guardrail and road indicates lateral speed in contrast to the tracked subject", "sodium-gold haze contrasts against the blue-gray ambient tone to maintain silhouette depth" ], "creative_decisions": [ "applied rim lighting on back and shoulders to ensure the figure maintains 3D volume instead of becoming a flat cutout", "positioned the road angling toward the upper-right to enhance the sense of movement and geometric depth" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "26e8e1c075d06753", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "272d7e5725f60370", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 6, "dependency_reason": "atmosphere_reference: 이전 씬의 어둡고 푸른빛이 도는 해안가 밤안개 톤(mood/tone)을 연속성 있게 유지하기 위함.", "_shot_index": 1, "visible_entities": [ "C01", "L05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 9, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From a low three-quarter front-side tracking position at about waist height, the camera runs parallel with Suriyoung's bicycle two to three meters away, keeping her body a strong dark silhouette against the coastal fog. The frame favors the mid-action downward pedal stroke: her bent knee and forward lean sit in the center while the road and guardrail smear laterally through mist.", "lighting_mood": "The scene is underexposed into blue-gray coastal darkness, with only thin road glints and distant sodium-gold haze outlining her silhouette.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "foggy coastal road ahead", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "foggy coastal road ahead", "label": "수리영 on bicycle", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L05" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [ { "scene_index": 6, "shot_index": 1, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "28db0a155128ff2e" }, { "scene_index": 10, "heading": "S#10. 옥탑방 / N", "beat_title": "낯선 남자의 침입 시도", "representative_moment": "C12의 손가락이 문틈에 끼워진 전단지를 꾹 밀어넣는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Insert shot at hand height, extreme close-up focus with hard cold blue and dull amber lighting", "t2i_prompt": "Photorealistic cinematic still. C12O13 in a dark sleeve, an Asian man, looking off-screen, appearing as a dark shape with no visible face and obscured details, his shadowed fingers firmly pressing a creased, dirty paper advertising flyer into the metal door seam. The camera is at hand height, extremely close and nearly parallel to the door edge, exposing only a dark vertical gap of the unlit threshold shadow. The compressed flyer edge and the metal door share the insert frame. A torn printed face is partly visible where the flyer bends into the door gap, its text distorted by pressure. Hard cold blue ambient nighttime light grazes the metal surface, leaving faint wet highlights, while a dull amber spill catches the fingertip and stained paper, keeping the rest in damp shadow.", "outfit_assignments": [ { "character_id": "C12", "outlook_id": "O13" } ], "source_facts": [ "Shot description: hand pausing after pushing the paper flyer into the metal door seam", "Hard blue rooftop night grazes the metal edge while a dull amber spill catches the fingertip and stained paper", "metal door seam: closed but flexing, lock under pressure", "paper advertising flyer: creased, jammed, dirty" ], "visual_inferences": [ "fixed_element rooftop_exterior_night_atmosphere: consistent cold blue ambient lighting and damp shadows applied to the door area", "Due to the insert framing, only the hand, flyer, and door edge are visible, obscuring the rest of the character and environment" ], "creative_decisions": [ "Use of extreme macro insert framing to heighten the surgical, illicit feeling of the intrusion", "Contrast of cold blue environment and amber spill on the fingertip to draw immediate focus to the mechanical pressure on the lock" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "concrete floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "metal door seam" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "terrace", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "utility zone", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "f54255f88302a359", "owned_hash": "9cbd332222960422", "camera_direction_hash": "ad0b29f88bf6ab71", "owned_usage_hash": "98e21772e7dbf20e", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "metal door seam", "reason": "The prompt references the door seam, edge, and gap as existing structural elements and spatial anchors where the flyer is being inserted, without requesting to redraw or modify the door itself, matching the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 4, "dependency_reason": "atmosphere_reference (prior shot S4_Shot1 establishes the tone, heavily shrouded in darkness with cold blue ambient nighttime light)", "_shot_index": 4, "visible_entities": [ "C12", "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 10, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At hand height just outside the rooftop room, the lens sits extremely close and nearly parallel to the metal door seam, frozen at the exact pause after the paper flyer has been pushed inward. Only the man's fingers, the compressed flyer edge, and a sliver of dark threshold share the insert frame, making the pressure of the hand feel surgical and illicit.", "lighting_mood": "Hard blue rooftop night grazes the metal edge while a dull amber spill catches the fingertip and stained paper, keeping the rest in damp shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C12", "L03" ], "allowed_outlook_pairs": [ { "character_id": "C12", "outlook_id": "O13" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L03B03", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "concrete floor", "door", "drying platform", "facade", "parapet wall", "rooftop room", "stair landing", "terrace", "utility zone", "water tank", "window" ], "camera_reference": { "camera_position": "near number 5 (iron entry door), facing along the south facade of number 2 (rooftop room footprint) toward number 6 (small facade window), with number 1 (open rooftop terrace) and number 9 (drying platform) partly visible behind", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "Emphasize the door gap area and facade window; depict the flyer intrusion cue, the slightly opened door state, loose dried leaves, and the ominous window silhouette for this night moment." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "rooftop_exterior_night_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The exterior of the rooftop room is heavily shrouded in darkness and illuminated by a strong, cold blue ambient nighttime light. The surrounding surfaces and structures are deeply shadowed and damp, leaving only faint wet highlights visible under the harsh lighting. This consistent blue tint dictates the visual tone of the outdoor environment.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "S#10. 옥탑방 / N", "Hard blue rooftop night grazes the metal edge", "keeping the rest in damp shadow", "A cold blue interior glow", "rooftop falls into muted black with only faint wet highlights" ], "visual_inferences": [ "The cold blue light serves as the primary unifying ambient source (like stylized moonlight or stark city lighting) across the entire rooftop exterior." ], "creative_decisions": [ "Maintaining a high-contrast, heavily blue-tinted dark aesthetic to amplify the eerie, cold, and suspenseful thriller atmosphere." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 4, "shot_index": 1, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C12O13", "policy": "required" }, { "kind": "background", "id": "L03B03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "cef04549c991e67b" }, { "scene_index": 10, "heading": "S#10. 옥탑방 / N", "beat_title": "창문에 비친 기이한 실루엣", "representative_moment": "푸른 창문에 거대한 그림자가 민숙의 그림자를 짓누르는 모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide static shot, slightly low angle, high-contrast cold blue night lighting", "t2i_prompt": "Photorealistic cinematic still. Outside the rooftop room, from a slightly low angle looking up, the camera focuses on the existing window set into the underexposed damp facade. Flattened against its blue-tinted glass from the inside, C12O13 in a black silhouette suit, a massive dark figure with a humanlike outline, stands at a three-quarter left angle, looking down and looming over the body below. Trapped beneath the massive shadow, C04O05 in a gray knit homewear, an East Asian woman in her 40s, lies in right profile with eyes closed, severely injured and pinned against the same window glass. An interior light from inside projects the figures' stark shadow silhouettes against the glass, preserving their shoulder depth and arm silhouettes. The surrounding exterior is heavily shrouded in darkness, illuminated by a cold blue ambient nighttime light, leaving only faint wet highlights on the rough walls.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O05" }, { "character_id": "C12", "outlook_id": "O13" } ], "source_facts": [ "S#10. 옥탑방 / N", "창문에 비친 그림자", "민숙을 덮쳐 어깨를 물어뜯는 듯한 모습" ], "visual_inferences": [ "fixed_element rooftop_exterior_night_atmosphere: strong cold blue ambient nighttime light, heavily shrouded in darkness, faint wet highlights", "Silhouettes projected onto the window glass from the inside", "Attacker looking down at the victim, victim severely injured with eyes closed" ], "creative_decisions": [ "Wide shot from slightly below eye level to emphasize the oppressive, looming scale of the attacker's shadow on the glowing window pane", "High-contrast cold blue lighting to isolate the window screen against the muted black exterior" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "concrete floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "set into the underexposed damp facade" }, { "owned_token": "parapet wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "Outside the rooftop room" }, { "owned_token": "stair landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "terrace", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "utility zone", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "focuses on the existing window set into the underexposed damp facade" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "0019b915a1dc1875", "owned_hash": "9cbd332222960422", "camera_direction_hash": "681dd08fcdd53508", "owned_usage_hash": "7146dd63c8538d5b", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "rooftop room", "violating_phrase": "Outside the rooftop room", "reason": "The rooftop room is used purely as a spatial viewpoint anchor for the camera position, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "facade", "violating_phrase": "set into the underexposed damp facade", "reason": "The facade is mentioned as a static background context containing the window, aligning with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "focuses on the existing window set into the underexposed damp facade", "reason": "The prompt explicitly refers to the 'existing window' as a plane to cast shadow silhouettes on, indicating no intent to recreate or modify it, which confirms the anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 4, "dependency_reason": "푸른빛의 차갑고 어두운 야간 옥탑방 외곽 및 대기 분위기 연속성 유지", "_shot_index": 6, "visible_entities": [ "C04", "C12", "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 10, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Outside the rooftop room several feet from the blue-tinted window, the camera holds static slightly below eye level and angles upward so the glass becomes the main plane of the frame without exceeding natural scale. Two enlarged shadow silhouettes flatten across it: the larger form presses down from the upper side while the smaller form is trapped lower in the blue pane, with the surrounding rooftop underexposed to near-black.", "lighting_mood": "A cold blue interior glow turns the window into a shadow screen, while the rooftop falls into muted black with only faint wet highlights.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C04", "C12", "L03" ], "allowed_outlook_pairs": [ { "character_id": "C04", "outlook_id": "O05" }, { "character_id": "C12", "outlook_id": "O13" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L03B03", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "concrete floor", "door", "drying platform", "facade", "parapet wall", "rooftop room", "stair landing", "terrace", "utility zone", "water tank", "window" ], "camera_reference": { "camera_position": "near number 5 (iron entry door), facing along the south facade of number 2 (rooftop room footprint) toward number 6 (small facade window), with number 1 (open rooftop terrace) and number 9 (drying platform) partly visible behind", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "Emphasize the door gap area and facade window; depict the flyer intrusion cue, the slightly opened door state, loose dried leaves, and the ominous window silhouette for this night moment." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "rooftop_exterior_night_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The exterior of the rooftop room is heavily shrouded in darkness and illuminated by a strong, cold blue ambient nighttime light. The surrounding surfaces and structures are deeply shadowed and damp, leaving only faint wet highlights visible under the harsh lighting. This consistent blue tint dictates the visual tone of the outdoor environment.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "S#10. 옥탑방 / N", "Hard blue rooftop night grazes the metal edge", "keeping the rest in damp shadow", "A cold blue interior glow", "rooftop falls into muted black with only faint wet highlights" ], "visual_inferences": [ "The cold blue light serves as the primary unifying ambient source (like stylized moonlight or stark city lighting) across the entire rooftop exterior." ], "creative_decisions": [ "Maintaining a high-contrast, heavily blue-tinted dark aesthetic to amplify the eerie, cold, and suspenseful thriller atmosphere." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 4, "shot_index": 1, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C04O05", "policy": "required" }, { "kind": "character_outlook", "id": "C12O13", "policy": "required" }, { "kind": "background", "id": "L03B03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "ae2bf820f354c073" }, { "scene_index": 11, "heading": "S#11. 옥탑 방 밖 / N", "beat_title": "불길한 귀가", "representative_moment": "어두운 마당에 자전거를 세운 C01이 역겨운 냄새를 느끼며 옥탑방으로 이어지는 계단을 올려다보는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-height wide angle observer shot, cold low-key lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a dark hooded jacket, an early 20s Korean woman with a youthful oval face and long natural black hair, stands in the lower-center foreground, looking toward the rooftop stair in the upper-right background. Positioned at a three-quarter angle from the right, her head is tilted up, one hand gripping the handlebar of the bicycle resting beside her, the bicycle's frame and wet front tire clearly visible on the ground. The observer camera at chest height captures her full figure planted on the shared yard, the damp stair rising behind her into the upper frame. The scene is bathed in a low-key, dark blue-gray shadow against the dim wall and building facade, accented by a weak, sickly gold light spilling down the steps to carve her silhouette in the eerie nocturnal gloom.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "마당 후미진 곳에 자전거를 대충 세우고, 옥탑방과 연결되는 계단을 급히 오른다.", "어두운 다세대 빌라 마당, 자전거를 옆에 세운 채 계단 위를 올려다보는 수리영의 전신.", "C01: three_quarter_right, one hand on bicycle, gaze=up, state=alive" ], "visual_inferences": [ "fixed_element: eerie_nocturnal_lighting applying a blue-gray shadow wash and weak sickly gold light from above.", "Chest-height observer perspective placing the figure low in the frame to emphasize the vertical ascent of the stairs." ], "creative_decisions": [ "Composed the wide frame to include the bicycle's full body alongside the character to strictly follow vehicle rendering rules.", "Used strict foreground/background placement to adhere to the frame spatial contract, contrasting the dark damp yard with the sickly gold spill from the stairs." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "apron", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bicycle rack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "building facade" }, { "owned_token": "gate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lane", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "manhole", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "rooftop stair in the upper-right background" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "dim wall" }, { "owned_token": "yard", "usage_kind": "anchor", "source_phrase": "planted on the shared yard" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "901c167b9cdd405c", "owned_hash": "32966e993d7d7aed", "camera_direction_hash": "a110976f9a5c276f", "owned_usage_hash": "bf3e566e70ea47d5", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "facade", "violating_phrase": "building facade", "reason": "The prompt references the building facade as part of the background shadow environment, which is an spatial anchor matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "rooftop stair", "reason": "The prompt uses the stair (and steps) as a spatial reference for the character's gaze, lighting direction, and background composition, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "dim wall", "reason": "The prompt mentions the dim wall to establish the background shadow, which acts as a scene anchor matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "yard", "violating_phrase": "shared yard", "reason": "The prompt mentions the shared yard as the ground upon which the character is standing, serving as a spatial anchor matching the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "N/A", "_shot_index": 2, "visible_entities": [ "C01", "L10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 11, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "In the dark apartment courtyard, the camera is locked at chest height about twelve feet in front of Suriyoung, wide enough to include her full body, the bicycle beside her, and the stairs rising behind. She is held low in the frame with the staircase climbing into the upper background, creating a clear upward pull before the later handheld ascent.", "lighting_mood": "Sparse courtyard light carves her and the bicycle in blue-gray shadow, with a weak gold spill from an unseen stair bulb hinting upward.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "looks_to_anchor", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "rooftop stairs", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "looks_toward", "gesture_target_label": "rooftop stairs", "label": "수리영", "screen_zone": "lower_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L10" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L10B01", "reference_usage": "exact_background", "owned_objects": [ "apron", "bicycle rack", "door", "drain", "facade", "gate", "lane", "manhole", "stair", "wall", "yard" ], "camera_reference": { "camera_position": "near number 3 (pedestrian gate), facing diagonally across number 1 (shared yard) toward number 6 (stair base) and number 8 (bicycle parking spot), with number 2 (vehicle access lane) on the left side of frame", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm to 35mm wide angle", "framing_notes": "Include the stair base and bicycle-at-stair-base overlay cue prominently; allow the police car arrival and siren overlay cue to occupy the access lane edge." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "eerie_nocturnal_lighting", "element_type": "environment_state", "character_name": "", "description": "The surrounding environment is bathed in a low-key, dark blue-gray shadow. This pervasive darkness is accented by a weak, sickly gold light spilling from a nearby unseen source, creating subtle rims and highlights that evoke a suspenseful and uncomfortable atmosphere.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "어두운 다세대 빌라 마당", "blue-gray shadow", "weak gold spill", "low-key blue-gray wash", "thin sickly gold rim" ], "visual_inferences": [ "The consistent blue-gray and gold lighting motif visually connects the ground-level courtyard and the upper rooftop landing, maintaining the foreboding mood as the character ascends." ], "creative_decisions": [ "Using a contrasting color palette of cold blue and sickly gold to visually amplify the disturbing nature of the unseen threat and the foul smell mentioned in the scene text." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [ { "scene_index": 11, "shot_index": 4, "description": "반쯤 열린 현관 철문을 굳은 표정으로 주시하는 수리영의 얼굴.", "keep_elements": [ { "label": "damp blue-gray wall texture and facade in shadow", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "sickly gold lighting and dark blue-gray shadows", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "background", "id": "L10B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "a6899d089c682799" }, { "scene_index": 11, "heading": "S#11. 옥탑 방 밖 / N", "beat_title": "불길한 예감", "representative_moment": "반쯤 열린 옥탑방 현관 철문을 굳은 표정으로 주시하는 수리영(C01).", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Close framing from front-left at face height, low-key blue-gray wash with sickly gold rim light", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a dark hooded jacket, an East Asian young woman in her 20s, shown in close framing with only her head and shoulders visible. She is facing three-quarter left with her neck craned forward, looking off-screen with a tightened mouth and fixed stare. She is bathed in the same eerie low-key dark blue-gray shadow, with a weak, sickly gold rim light from an unseen source tracing the side of her face. The tight space is defined by [: a soft dark edge of a half-open metal door at the frame side and a damp, stained blue-gray landing wall compressing the background directly behind her], with thin cold mist hovering faintly around her breath.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "반쯤 열린 현관 철문을 굳은 표정으로 주시하는 수리영의 얼굴", "역겨운 냄새에 멈칫한 그녀", "shot description: On the rooftop landing, angled from Suriyoung's front-left side" ], "visual_inferences": [ "fixed_element eerie_nocturnal_lighting: low-key blue-gray shadow and weak gold spill maintained from prior shot", "close framing focuses intensely on her tightened mouth and stare rather than the threat" ], "creative_decisions": [ "Using a shallow depth of field and close framing to isolate her expression and elevate the sense of dread", "Sickly gold rim accentuates the eerie atmosphere and visually represents the unseen foul smell" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "clothesline post", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "roof deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair bulkhead", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e1d299a70fb8d321", "owned_hash": "78dcfca991c6dc5b", "camera_direction_hash": "d0de48a7a9925a12", "owned_usage_hash": "5e764300e222b9e6", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 11, "dependency_reason": "zoom_in_detail from previous shot, maintaining the same space and lighting", "_shot_index": 4, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 11, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "On the rooftop landing, the camera dollies in from three to four feet away at face height, angled from Suriyoung's front-left side. The half-open metal door is kept mostly implied by her eyeline, with only a soft dark edge at frame side, so the shot concentrates on her tightened mouth and fixed stare rather than showing the threat directly.", "lighting_mood": "A low-key blue-gray wash leaves her eyes shadowed, with a thin sickly gold rim from the doorway tracing the side of her face.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L10B02", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "eerie_nocturnal_lighting", "element_type": "environment_state", "character_name": "", "description": "The surrounding environment is bathed in a low-key, dark blue-gray shadow. This pervasive darkness is accented by a weak, sickly gold light spilling from a nearby unseen source, creating subtle rims and highlights that evoke a suspenseful and uncomfortable atmosphere.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "어두운 다세대 빌라 마당", "blue-gray shadow", "weak gold spill", "low-key blue-gray wash", "thin sickly gold rim" ], "visual_inferences": [ "The consistent blue-gray and gold lighting motif visually connects the ground-level courtyard and the upper rooftop landing, maintaining the foreboding mood as the character ascends." ], "creative_decisions": [ "Using a contrasting color palette of cold blue and sickly gold to visually amplify the disturbing nature of the unseen threat and the foul smell mentioned in the scene text." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 11, "shot_index": 2, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "130249b88dfdc4b4" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "처참하게 훼손된 엄마의 시신 발견", "representative_moment": "커튼 쪽에 고개를 떨군 채 앉아있는 피투성이의 C04 정면 샷.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low seated eye-level medium front-facing tableau, 28mm wide angle perspective, cold blue-gray shadows with faint dirty-gold and deep red accents.", "t2i_prompt": "Photorealistic cinematic still. C04O05 in a gray knit homewear, a Korean woman in her early 40s, appears as a seated body in the middle-center midground, against the parted curtain behind the body in the middle-center background. Captured from a low seated eye level in a medium front-facing tableau, the lifeless figure is facing the camera with her head fallen and gaze directed down. Severe torn shoulder and collarbone wounds are visibly exposed on her slumped form. One stiff clenched hand, extending from a gray knit sleeve, tightly holds a single crumpled photograph resting clearly near the edge of the bed. The damp floorboards in the foreground are marked with dark red blood tracks leading toward her. A weak room lamp casts cold, low-key blue-gray shadows across the dim, stained bedside wall, leaving a faint tarnished-gold edge on the hanging folds of the curtain.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O05" } ], "source_facts": [ "앉은 채로 고개를 떨군 채 죽어있는 엄마- 민숙!!!!", "살점이 뜯긴 어깨와 쇄골, 마치 동물에게 물어뜯긴 듯한 거친 상처가 고스란히 드러나있다.", "엄마 손안에 구겨진 사진 한 장이 보인다." ], "visual_inferences": [ "fixed_element: blue_gray_ambient_lighting establishes cold, low-key shadows with faint dirty gold spill.", "forward_zoom_targets: hand holding crumpled photo, edge of bed, damp floorboards, and gray knit sleeve are placed visibly without obstruction in the medium frame.", "The scene demands a front-facing tableau keeping the lifeless posture fixed." ], "creative_decisions": [ "Medium framing from a low seated eye level to observe the seated body directly, satisfying the DP's 'front-facing tableau' instruction.", "Restrained red blood accents and tarnished-gold edges provide stark contrast against the pervasive blue-gray darkness.", "Integrated the 'parted curtain behind body' and 'seated body' frame spatial constraints directly into the opening sequence." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "edge of the bed" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "parted curtain behind the body" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "anchor", "source_phrase": "weak room lamp" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1141db045f47f7ca", "owned_hash": "e92ed9a0157bff4e", "camera_direction_hash": "7bc36991b9cbab08", "owned_usage_hash": "1a3c210096768ed3", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "resting clearly near the edge of the bed", "reason": "The bed is used as a spatial reference for positioning the photograph and hand, with no instructions to redraw or modify the bed itself.", "verdict": "anchor_reference" }, { "owned_object": "curtain", "violating_phrase": "parted curtain behind the body", "reason": "The curtain acts as a background anchor and lighting reference, aligning with the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "lamp", "violating_phrase": "A weak room lamp casts cold, low-key blue-gray shadows", "reason": "The lamp is mentioned only as a light source casting shadows, not as an object to be redrawn or newly generated.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "atmosphere_reference를 통해 S5_Shot11의 음산한 분위기와 푸른 회색조 조명 톤을 연속적으로 적용", "_shot_index": 8, "visible_entities": [ "C04", "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From low seated eye level directly in front of the curtain area, about six feet away, the camera begins the reveal but holds a beat longer on Minsuk's body as a front-facing tableau. The frame keeps her seated form centered behind the parted curtain, with the blood-marked floor leading toward her but the composition restrained and still rather than graphic.", "lighting_mood": "A weak room lamp and spill from the living room create low-key blue-gray darkness, with restrained red blood accents and a faint tarnished-gold edge on the curtain folds.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "parted curtain behind body", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "민숙 seated body", "screen_zone": "middle_center", "target_id": "C04", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C04", "L04" ], "allowed_outlook_pairs": [ { "character_id": "C04", "outlook_id": "O05" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B05", "reference_usage": "exact_background", "owned_objects": [ "TV stand", "bed", "chair", "curtain", "curtain track", "door", "lamp", "shelf", "table", "wardrobe", "window" ], "camera_reference": { "camera_position": "at number 10 (young bedroom door), looking into number 5 (young adult bedroom) toward number 24 (curtain) and number 22 (young bed), with part of number 2 (shared living dining zone) behind the camera side.", "camera_height": "eye-level standing ~1.55m", "lens_hint": "28mm wide angle for cramped interior", "framing_notes": "Prioritize the line from number 10 to number 24; include number 22 and the wall area above it for overlay number 35. Keep number 16 and number 18 available in adjacent living-area coverage." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "blue_gray_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior environment is dominated by cold, low-key blue-gray shadows, establishing a macabre and chilling atmosphere. A weak ambient spill, occasionally showing faint dirty-gold undertones, provides minimal illumination that flattens the room and severely suppresses vibrant colors.", "applies_to_shots": [ 8, 13, 16 ], "element_scope": "full", "source_facts": [ "weak room lamp and spill", "low-key blue-gray darkness", "faint dirty gold spill", "Cold blue-gray shadow flattens the room" ], "visual_inferences": [ "The dark blue-gray tint is a constant ambient condition that pervades all camera angles, contrasting sharply with any red elements like blood or supernatural marks." ], "creative_decisions": [ "Using a consistent suppressed lighting palette across different focal lengths to visually unify the wide, revealing shots with the extreme close-ups." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 11, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [ { "scene_index": 12, "shot_index": 13, "description": "시신의 굳은 손 안에 구겨진 사진 한 장이 쥐어진 클로즈업.", "keep_elements": [ { "label": "the edge of the bed and damp floorboards", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "the gray knit sleeve of the garment", "kind": "static_prop", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B05", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "31e27680ba6390b6" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "시신의 굳은 손과 사진", "representative_moment": "A close-up of C04's stiff clenched hand holding the crumpled photograph P09.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Extreme floor-level insert angled slightly downward, shallow depth of field.", "t2i_prompt": "Photorealistic cinematic still. Extreme floor-level insert angled slightly downward, lens inches from the subject, shallow focus. C04 in a grey knit sleeve, the same deceased Korean woman in her early 40s from the previous shot, her lifeless, stiff clenched hand occupying the central third of the frame. P09, a small crumpled old black-and-white paper photograph, trapped within the rigid fingers. The photo's image-bearing face turned partly toward the camera, its printed content mostly hidden by deep folds and shadow. A faint dirty gold light spill on the creased paper surface, against a suppressed red undertone near the pale skin. The same cold, low-key blue-gray ambient darkness filling the space, with the out-of-focus dark rumpled texture of the damp bed visible in the soft background.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O05" } ], "source_facts": [ "Shot description: 시신의 굳은 손 안에 구겨진 사진 한 장이 쥐어진 클로즈업.", "state=dead: lying motionless, lifeless.", "P09 visual contract: a small crumpled black-and-white paper photograph." ], "visual_inferences": [ "fixed_element damp_room_bed: bed texture present in the blurred background to anchor the physical space.", "fixed_element blue_gray_ambient_lighting: cold blue-gray shadows heavily suppress the room colors." ], "creative_decisions": [ "Shallow depth of field isolates the stiff fingers and crumpled photo while maintaining the continuity of the damp bed in the soft background.", "Faint dirty gold light spill used to highlight the creases of P09 against the dark environment." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background", "prop" ], "owned_object_usage": [ { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "the out-of-focus dark rumpled texture of the damp bed" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "36d5a7563df95cb3", "owned_hash": "e92ed9a0157bff4e", "camera_direction_hash": "9ad7c4424f47ccce", "owned_usage_hash": "bb1abcfda306c81e", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "the out-of-focus dark rumpled texture of the damp bed", "reason": "The bed is referred to only as an out-of-focus background element to provide depth and context, which aligns with the producer's 'anchor' declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "zoom_in_detail from previous shot of the corpse in the same room, keeping environment lighting and subject state consistent.", "_shot_index": 13, "visible_entities": [ "C04", "P09" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 13 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the flow’s very low close-observation position beside 민숙’s hand, the camera tightens to an extreme floor-level insert angled slightly downward, with the crumpled photograph held in the central third and her rigid fingers forming a cold frame around it. The lens stays inches from the hand, using shallow focus so the photo’s creased surface is sharp while the damp bedding and corpse context fall into blue-gray blur.", "lighting_mood": "Low-key blue-gray room darkness is cut by a faint dirty gold spill on the paper and a suppressed red undertone near the hand.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C04", "P09" ], "allowed_outlook_pairs": [ { "character_id": "C04", "outlook_id": "O05" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C04", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "Only 민숙’s corpse hand is framed; a full outlook reference would not align with the visible body detail." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B05", "reference_usage": "exact_background", "owned_objects": [ "TV stand", "bed", "chair", "curtain", "curtain track", "door", "lamp", "shelf", "table", "wardrobe", "window" ], "camera_reference": { "camera_position": "at number 10 (young bedroom door), looking into number 5 (young adult bedroom) toward number 24 (curtain) and number 22 (young bed), with part of number 2 (shared living dining zone) behind the camera side.", "camera_height": "eye-level standing ~1.55m", "lens_hint": "28mm wide angle for cramped interior", "framing_notes": "Prioritize the line from number 10 to number 24; include number 22 and the wall area above it for overlay number 35. Keep number 16 and number 18 available in adjacent living-area coverage." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "blue_gray_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior environment is dominated by cold, low-key blue-gray shadows, establishing a macabre and chilling atmosphere. A weak ambient spill, occasionally showing faint dirty-gold undertones, provides minimal illumination that flattens the room and severely suppresses vibrant colors.", "applies_to_shots": [ 8, 13, 16 ], "element_scope": "full", "source_facts": [ "weak room lamp and spill", "low-key blue-gray darkness", "faint dirty gold spill", "Cold blue-gray shadow flattens the room" ], "visual_inferences": [ "The dark blue-gray tint is a constant ambient condition that pervades all camera angles, contrasting sharply with any red elements like blood or supernatural marks." ], "creative_decisions": [ "Using a consistent suppressed lighting palette across different focal lengths to visually unify the wide, revealing shots with the extreme close-ups." ], "confidence": "high" }, { "element_id": "damp_room_bed", "element_type": "persistent_prop", "character_name": "", "description": "A bed situated in the room, identifiable by its slightly disturbed and damp bedding. The physical presence and edge of the bed serve as a spatial and scale reference point within the grim interior space.", "applies_to_shots": [ 13, 16 ], "element_scope": "full", "source_facts": [ "damp bedding", "bed edge low in frame as a scale reference" ], "visual_inferences": [ "The bed is located near the center of the supernatural events, visible both as background texture at the floor level and as a foreground framing element when observing the wall." ], "creative_decisions": [ "Maintaining the visible presence of the damp bed edge to ground the horrifying narrative events in the physical, tangible reality of the small room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 8, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character", "id": "C04", "policy": "required" }, { "kind": "background", "id": "L04B05", "policy": "required" }, { "kind": "prop", "id": "P09", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [ { "contract_id": "rc_001", "scope": { "scene_index": 12, "shot_indices": [ 13 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P09", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] } ], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "98ab564ed19d71b4" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "벽면에 나타난 붉은 원형 표식", "representative_moment": "침대 위 벽면에 커다란 붉은 원 표식(P08)이 그려져 있는 정지 구도.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium framing, camera squared flat to the wall, cold blue-gray lighting with a saturated red focal point", "t2i_prompt": "Photorealistic cinematic still. A large, highly saturated red circle, P08, marks the dead center of the pale damp wall. The wall surface is unnervingly perpendicular to the lens. Below the symbol, the foot-side edge of the damp bed anchors the lower frame as a spatial reference, the mattress receding toward the marked wall. Cold, low-key blue-gray shadows flatten the empty environment, while a faint dirty-gold ambient spill highlights the wet-pigment texture of the red anomaly. Focused entirely on the stark contrast between the terrifying red mark and the macabre, suppressed atmosphere of the room.", "outfit_assignments": [], "source_facts": [ "침대 위 벽면에 커다란 붉은 원 표식이 그려져 있는 정지 구도.", "A bed situated in the room... damp bedding.", "Cold blue-gray shadow flattens the room." ], "visual_inferences": [ "fixed_element blue_gray_ambient_lighting: low-key blue-gray darkness pervades the scene, suppressing vibrant colors except for the red mark.", "fixed_element damp_room_bed: the bed edge is positioned low in the frame to ground the supernatural event in physical reality.", "P08 serves as the sole source of highly saturated color, appearing as a hallucination-like wet pigment." ], "creative_decisions": [ "Held the camera completely still and squared flat to the wall to emphasize the unnerving, deliberate nature of the red circle.", "Used stark color contrast between the chilling blue-gray ambient shadows and the intense red to maximize the visceral dread." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "prop", "background" ], "owned_object_usage": [ { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "the foot-side edge of the damp bed anchors the lower frame as a spatial reference" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d5e6384764b7f5b5", "owned_hash": "e92ed9a0157bff4e", "camera_direction_hash": "5a750f212ff2a127", "owned_usage_hash": "33c1142dc4dba3e4", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "the foot-side edge of the damp bed anchors the lower frame as a spatial reference", "reason": "The t2i_prompt refers to the bed as a spatial reference to anchor the lower frame without asking to redraw or alter it, which matches the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 16, "visible_entities": [ "L04", "P08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 16 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At shoulder height from the foot side of the bed, the camera is squared flat to the wall surface and held completely still, inheriting the supernatural reveal stage but removing all motion. The large red circle sits dead center above the bed, with the bed edge low in frame as a scale reference and the wall kept unnervingly perpendicular to the lens.", "lighting_mood": "Cold blue-gray shadow flattens the room while the red circle carries the only saturated color, as if wet pigment is absorbing the light.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L04", "P08" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B05", "reference_usage": "exact_background", "owned_objects": [ "TV stand", "bed", "chair", "curtain", "curtain track", "door", "lamp", "shelf", "table", "wardrobe", "window" ], "camera_reference": { "camera_position": "at number 10 (young bedroom door), looking into number 5 (young adult bedroom) toward number 24 (curtain) and number 22 (young bed), with part of number 2 (shared living dining zone) behind the camera side.", "camera_height": "eye-level standing ~1.55m", "lens_hint": "28mm wide angle for cramped interior", "framing_notes": "Prioritize the line from number 10 to number 24; include number 22 and the wall area above it for overlay number 35. Keep number 16 and number 18 available in adjacent living-area coverage." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "blue_gray_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior environment is dominated by cold, low-key blue-gray shadows, establishing a macabre and chilling atmosphere. A weak ambient spill, occasionally showing faint dirty-gold undertones, provides minimal illumination that flattens the room and severely suppresses vibrant colors.", "applies_to_shots": [ 8, 13, 16 ], "element_scope": "full", "source_facts": [ "weak room lamp and spill", "low-key blue-gray darkness", "faint dirty gold spill", "Cold blue-gray shadow flattens the room" ], "visual_inferences": [ "The dark blue-gray tint is a constant ambient condition that pervades all camera angles, contrasting sharply with any red elements like blood or supernatural marks." ], "creative_decisions": [ "Using a consistent suppressed lighting palette across different focal lengths to visually unify the wide, revealing shots with the extreme close-ups." ], "confidence": "high" }, { "element_id": "damp_room_bed", "element_type": "persistent_prop", "character_name": "", "description": "A bed situated in the room, identifiable by its slightly disturbed and damp bedding. The physical presence and edge of the bed serve as a spatial and scale reference point within the grim interior space.", "applies_to_shots": [ 13, 16 ], "element_scope": "full", "source_facts": [ "damp bedding", "bed edge low in frame as a scale reference" ], "visual_inferences": [ "The bed is located near the center of the supernatural events, visible both as background texture at the floor level and as a foreground framing element when observing the wall." ], "creative_decisions": [ "Maintaining the visible presence of the damp bed edge to ground the horrifying narrative events in the physical, tangible reality of the small room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 8, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B05", "policy": "required" }, { "kind": "prop", "id": "P08", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [ { "contract_id": "rc_001", "scope": { "scene_index": 12, "shot_indices": [ 16 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P08", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] } ], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "497c867a5c2ebcda" }, { "scene_index": 13, "heading": "S#13. 옥탑방 외곽 / N", "beat_title": "경찰 도착과 현장 발견", "representative_moment": "붉은 사이렌 불빛이 젖은 마당과 옥탑방 계단 입구를 비추는 경찰차 전경.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide angle eye-level shot, high-contrast strobing red lighting", "t2i_prompt": "Photorealistic cinematic still. A police car with flashing siren occupies the lower-left foreground, while the rooftop stairwell entrance stands open in the middle-center background. Captured from an eye-level standing height near the pedestrian gate, the camera faces diagonally across the shared yard. The wet car roof and its rhythmic strobing red siren bar anchor the access lane edge, cutting through the dense blue-gray cold night darkness. Stark red pulses cast dramatic wet reflections across the damp yard surface and the building facade. In the distance, the stair base and a bicycle rack sit partially swallowed by mist, faint gold highlights catching on their metal edges.", "outfit_assignments": [], "source_facts": [ "계단 아래 다세대 빌라 마당, 붉은 사이렌 불빛이 번쩍이는 경찰차 전경.", "Strobing red police light cuts through dense blue-gray courtyard darkness (Shot 2)", "Red siren pulses rhythmically over cold night fill (Shot 5)", "wet car roof and wet reflections (Shot 2)" ], "visual_inferences": [ "The ground and ambient surrounding surfaces are damp, amplifying the red reflections.", "The red light serves as the primary practical light source linking the wide establishing angle and the courtyard atmosphere." ], "creative_decisions": [ "Use the high-contrast red pulsing light to visually emphasize the urgency and danger of the scene.", "Position the police car in the foreground to anchor the composition against the misty depth of the background stairwell." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "apron", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bicycle rack", "usage_kind": "anchor", "source_phrase": "bicycle rack" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "building facade" }, { "owned_token": "gate", "usage_kind": "anchor", "source_phrase": "near the pedestrian gate" }, { "owned_token": "lane", "usage_kind": "anchor", "source_phrase": "access lane edge" }, { "owned_token": "manhole", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "stair base" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "yard", "usage_kind": "anchor", "source_phrase": "across the shared yard" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "7690b4cf956ccff5", "owned_hash": "32966e993d7d7aed", "camera_direction_hash": "355e01daca823b13", "owned_usage_hash": "36aceb95029a5f5e", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bicycle rack", "violating_phrase": "a bicycle rack sit partially swallowed by mist", "reason": "The prompt references the existing bicycle rack in the distance being swallowed by mist, acting as a spatial anchor with no redraw instructions, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "facade", "violating_phrase": "the building facade", "reason": "The prompt references the building facade solely as a surface that receives dramatic wet red reflections, serving as an environmental anchor in agreement with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "gate", "violating_phrase": "near the pedestrian gate", "reason": "The pedestrian gate is used exclusively as a camera viewpoint reference, representing a valid anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "lane", "violating_phrase": "access lane edge", "reason": "The access lane edge is mentioned as a compositional anchor point for the foreground police car, with no redraw instructions.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "stair base", "reason": "The stair base is referenced as being partially swallowed by mist in the distance, serving as a structural anchor, which is consistent with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "yard", "violating_phrase": "across the shared yard", "reason": "The yard is referenced to define the camera direction and as a surface reflecting the red siren lights, acting as a stable background anchor without redraw commands.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 11, "dependency_reason": "previous_shot_refs specifies Scene 11 Shot 2 as exact_background", "_shot_index": 2, "visible_entities": [ "L10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 13, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the bottom of the stairs in the villa courtyard, the camera holds a chest-height wide frame with the police car in the foreground and the stairwell behind it, following the flow’s courtyard alarm stage before redirecting toward 수리영. The red siren bar and wet car roof occupy the lower foreground without overwhelming the frame, while the courtyard walls and stair mouth open behind in misty depth.", "lighting_mood": "Strobing red police light cuts through dense blue-gray courtyard darkness, creating hard wet reflections and brief gold highlights on metal edges.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "police car with flashing siren", "screen_zone": "lower_left", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "rooftop stairwell entrance", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L10" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L10B01", "reference_usage": "exact_background", "owned_objects": [ "apron", "bicycle rack", "door", "drain", "facade", "gate", "lane", "manhole", "stair", "wall", "yard" ], "camera_reference": { "camera_position": "near number 3 (pedestrian gate), facing diagonally across number 1 (shared yard) toward number 6 (stair base) and number 8 (bicycle parking spot), with number 2 (vehicle access lane) on the left side of frame", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm to 35mm wide angle", "framing_notes": "Include the stair base and bicycle-at-stair-base overlay cue prominently; allow the police car arrival and siren overlay cue to occupy the access lane edge." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "pulsing_red_siren_lighting", "element_type": "environment_state", "character_name": "", "description": "The environment is bathed in the rhythmic strobing of red police siren lights cutting through the dense, cold night darkness. The stark red pulses cast brief, dramatic highlights and reflections across the wet surfaces of the courtyard and any figures present in the area.", "applies_to_shots": [ 2, 5 ], "element_scope": "full", "source_facts": [ "Strobing red police light cuts through dense blue-gray courtyard darkness (Shot 2)", "Red siren pulses rhythmically over cold night fill (Shot 5)", "wet car roof and wet reflections (Shot 2)" ], "visual_inferences": [ "The ground and ambient surrounding surfaces are damp, amplifying the red reflections.", "The red light serves as the primary practical light source linking the wide establishing angle and the intimate character close-up." ], "creative_decisions": [ "Use the high-contrast red pulsing light to visually emphasize the urgency, danger, and emotional shock of the murder scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 11, "shot_index": 2, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L10B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "d2a64af7de2575a3" }, { "scene_index": 13, "heading": "S#13. 옥탑방 외곽 / N", "beat_title": "엄마의 죽음을 알리는 수리영", "representative_moment": "충격에 빠져 멍하니 서 있던 수리영이 다가온 혜수를 올려다보며 울먹이는 순간", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium over-the-shoulder framing, eye-level angled slightly downward, shallow depth of field on foreground subject, strong red-blue color contrast", "t2i_prompt": "Photorealistic cinematic still. Shot from upper-body height, angling slightly downward over the blurred shoulder of a Korean woman in her 40s in a detective coat, her back to the camera in the left foreground. Across the shared wet yard, C01O01 in a blue-gray hoodie, a Korean woman in her early 20s, fills the center-right. C01O01 stands at a three-quarter angle from the right, her shoulders curled inward, her tearful wet eyes fixed upward on the foreground woman. The dense, cold blue-gray night is pierced by rhythmic red police siren pulses. The stark red light casts striking reflections on C01O01's face and the damp yard, while the foreground woman remains mostly in protective shadow. In the dim background, the wet facade and the dark stair base fade into the chilly darkness.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영(C01)이 혜수(unregistered, 40대 중반 여자 경위)를 올려다보며 울먹임", "초점을 잃고 멍하니 서있는 수리영의 모습", "Shot 5: 혜수를 올려다보며 울먹이는 표정의 수리영 상반신", "Close beside 혜수 at upper-body height, the camera stays about one meter from 수리영 and angles slightly downward" ], "visual_inferences": [ "fixed_element pulsing_red_siren_lighting: stark red pulses cast brief, dramatic highlights across the wet surfaces", "Hye-su is placed in the left foreground with her back to the camera to serve as an OTS anchor", "Su-ri-young's curled posture emphasizes her shock and vulnerability in the wet courtyard" ], "creative_decisions": [ "Medium OTS composition specifically aligns with the camera angle to isolate the emotional connection across the shared yard", "Contrasting the stark red siren highlights on C01O01's face against the deep blue-gray night background to enhance the tragedy" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "apron", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bicycle rack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "the wet facade" }, { "owned_token": "gate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lane", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "manhole", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "the dark stair base" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "yard", "usage_kind": "anchor", "source_phrase": "Across the shared wet yard / the damp yard" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "fa8a5bdc7b737b5b", "owned_hash": "32966e993d7d7aed", "camera_direction_hash": "103f5066218e4502", "owned_usage_hash": "c33745028693ed62", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "yard", "violating_phrase": "Across the shared wet yard", "reason": "The yard is mentioned to describe the shared space between the characters and reflections, which serves as a spatial anchor matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "facade", "violating_phrase": "the wet facade", "reason": "The facade is mentioned as a background element to establish the setting, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "the dark stair base", "reason": "The stair base is used as a dim background anchor to set the scene, aligning with the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 13, "dependency_reason": "exact_background (same location, different angle/moment)", "_shot_index": 5, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 13, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Close beside 혜수 at upper-body height, the camera stays about one meter from 수리영 and angles slightly downward from 혜수’s position, preserving the flow’s intimate close-observation axis. 수리영’s tearful torso fills the center-right, while a blurred edge of 혜수’s shoulder enters the left foreground to make 수리영’s upward gaze emotionally legible.", "lighting_mood": "Red siren pulses rhythmically over cold night fill, catching 수리영’s wet eyes while 혜수 remains partly in protective shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L10B01", "reference_usage": "exact_background", "owned_objects": [ "apron", "bicycle rack", "door", "drain", "facade", "gate", "lane", "manhole", "stair", "wall", "yard" ], "camera_reference": { "camera_position": "near number 3 (pedestrian gate), facing diagonally across number 1 (shared yard) toward number 6 (stair base) and number 8 (bicycle parking spot), with number 2 (vehicle access lane) on the left side of frame", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm to 35mm wide angle", "framing_notes": "Include the stair base and bicycle-at-stair-base overlay cue prominently; allow the police car arrival and siren overlay cue to occupy the access lane edge." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "pulsing_red_siren_lighting", "element_type": "environment_state", "character_name": "", "description": "The environment is bathed in the rhythmic strobing of red police siren lights cutting through the dense, cold night darkness. The stark red pulses cast brief, dramatic highlights and reflections across the wet surfaces of the courtyard and any figures present in the area.", "applies_to_shots": [ 2, 5 ], "element_scope": "full", "source_facts": [ "Strobing red police light cuts through dense blue-gray courtyard darkness (Shot 2)", "Red siren pulses rhythmically over cold night fill (Shot 5)", "wet car roof and wet reflections (Shot 2)" ], "visual_inferences": [ "The ground and ambient surrounding surfaces are damp, amplifying the red reflections.", "The red light serves as the primary practical light source linking the wide establishing angle and the intimate character close-up." ], "creative_decisions": [ "Use the high-contrast red pulsing light to visually emphasize the urgency, danger, and emotional shock of the murder scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 13, "shot_index": 2, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "background", "id": "L10B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "3c05110a027226fa" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "사라진 사건 현장과 완벽하게 깨끗해진 옥탑방", "representative_moment": "모든 핏자국과 시신이 감쪽같이 사라진 채 서늘한 푸른빛이 감도는 비정상적으로 깨끗한 빈 방의 전경.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide eye-level doorway composition with flat, cold blue-gray lighting", "t2i_prompt": "Photorealistic cinematic still. Viewed from the doorway at standing eye level, the unnaturally pristine bedroom interior is completely devoid of any violent marks. The tidy bed edge aligns against the clean unmarked pale bedroom wall. To the side, the thin curtain fabric hangs partly open, revealing a spotless empty space behind it. On the pristine floorboards and corner baseboard, the fabric textures of the backpack and the doll are visible as they sit separated by a small gap. The entire space is washed in a cold, flat blue-gray room light, casting muted shadows and creating a sterile, chilling atmosphere across the overly ordered surfaces.", "outfit_assignments": [], "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다", "수리영 방안에 놓여진 백팩, 인형. 약간씩 위치가 달라진.", "The backpack and doll sit in the lower half of frame, separated by a small but suspicious gap" ], "visual_inferences": [ "fixed_element sterile_bedroom_environment: space consistently washed in cold, flat blue-gray light", "fixed_element displaced_backpack_and_doll: placed on floorboards near wall base, small gap between them" ], "creative_decisions": [ "Using wide eye-level doorway framing to scan the meticulously organized room and emphasize the chilling absence of the crime scene.", "Applying flat blue-gray tint to strip the room of warmth and highlight the sterile, accusatory cleanliness." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "The tidy bed edge" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "the thin curtain fabric" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "On the pristine floorboards" }, { "owned_token": "kitchenette", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "the clean unmarked pale bedroom wall" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "25bc2bb346053764", "owned_hash": "3bf5d875fea06c07", "camera_direction_hash": "720930da3c3b2a76", "owned_usage_hash": "9440c5ba0759dece", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "The tidy bed edge", "reason": "The bed is mentioned only as a layout alignment reference, which matches the producer's declaration as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "curtain", "violating_phrase": "the thin curtain fabric", "reason": "The curtain is described as hanging partly open to establish the spatial composition, serving as an anchor as declared.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "On the pristine floorboards", "reason": "The floorboards are referenced as a surface to position the backpack and doll, aligning with the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "the clean unmarked pale bedroom wall", "reason": "The wall is used as a reference point to show the tidy alignment of the bed, matching the anchor usage.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "Viewed from the doorway", "reason": "The doorway is mentioned as the camera viewpoint rather than being redrawn, acting as an anchor despite being declared absent.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "exact_background: 이전에 사건이 벌어졌던 동일한 수리영의 옥탑방 공간 구조를 유지하되 흔적이 완벽하게 지워진 상태로 묘사.", "_shot_index": 4, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the bedroom doorway at eye level, the camera inherits the room-reveal dolly-in but holds the wide composition longer, about three meters back, so the bed, floor, curtain area, and wall are all visible at once. The absence becomes the subject: no blood on the floor, no body behind the curtain, and no red circle above the bed, with the overly ordered surfaces arranged too neatly across the frame.", "lighting_mood": "Flat, cold blue-gray room light replaces the earlier red horror, making the cleanliness feel sterile and accusatory.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L04" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B06", "reference_usage": "exact_background", "owned_objects": [ "bed", "chair", "curtain", "door", "floor", "kitchenette", "shelf", "table", "track", "wall", "wardrobe", "window" ], "camera_reference": { "camera_position": "just inside number 10 (young bedroom door), facing number 5 (young adult bedroom) and number 24 (curtain), then allowing a view back toward number 16 (dining table).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "32mm moderate wide angle", "framing_notes": "Show the same room geometry as the prior bedroom event but clean; use number 36 for the spotless vanished-body state and number 37 for subtly shifted loose items." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "sterile_bedroom_environment", "element_type": "environment_state", "character_name": "", "description": "The bedroom interior is unnaturally pristine, meticulously organized, and entirely stripped of any bloodstains, red markings, or bodies. The space is consistently washed in a cold, flat blue-gray light, creating a sterile and chilling atmosphere across all visible walls and surfaces.", "applies_to_shots": [ 4, 5, 12 ], "element_scope": "full", "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다", "Flat, cold blue-gray room light replaces the earlier red horror", "clean room remains soft behind her", "Dim blue-gray floor light" ], "visual_inferences": [ "The complete absence of the horror elements and the intensely spotless surfaces are maintained across the wide reveal, the background of the character's reaction, and the floor layout." ], "creative_decisions": [ "Using the pervasive, cold blue-gray tint across all shots to visually emphasize the unnatural cleanliness and the void left by the missing evidence." ], "confidence": "high" }, { "element_id": "displaced_backpack_and_doll", "element_type": "persistent_prop", "character_name": "", "description": "A backpack and a doll are placed on the floorboards near the base of the wall. They sit with a small but suspicious gap between them, appearing slightly shifted from their usual positions but remaining entirely static.", "applies_to_shots": [ 4, 12 ], "element_scope": "full", "source_facts": [ "수리영 방안에 놓여진 백팩, 인형. 약간씩 위치가 달라진.", "bed, floor, curtain area, and wall are all visible at once", "The backpack and doll sit in the lower half of frame, separated by a small but suspicious gap" ], "visual_inferences": [ "Since the floor and overly ordered surfaces are fully visible in the wide room reveal (Shot 4), the specific arrangement of the displaced backpack and doll must be present exactly as detailed in the later insert shot (Shot 12)." ], "creative_decisions": [ "Keeping the small, distinct gap between the two props fixed across the wide spatial context and the tight clue insert to reinforce the unsettling feeling of the displacement." ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 16, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [ { "scene_index": 14, "shot_index": 5, "description": "하얗게 질린 얼굴로 멍하니 멈춘 수리영의 상반신.", "keep_elements": [ { "label": "the clean unmarked pale bedroom wall", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "the tidy bed edge", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "the thin curtain fabric", "kind": "environment", "subject_kind": "non_human_visual_element" } ] }, { "scene_index": 14, "shot_index": 12, "description": "방 한구석, 위치가 미묘하게 틀어진 백팩과 인형이 놓인 바닥 정지 구도.", "keep_elements": [ { "label": "the pristine floorboards and corner baseboard", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "the fabric textures of the backpack and the doll", "kind": "static_prop", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B06", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "64c5d46ebc1c3518" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "missing evidence to stunned reaction", "representative_moment": "하얗게 질린 얼굴로 멍하니 멈춘 수리영의 상반신.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-to-eye height close framing, shallow depth of field, drained blue-gray wash with faint red spill", "t2i_prompt": "Photorealistic cinematic still. Captured from chest-to-eye height, C01O01 in a blue-gray hooded jacket, a Korean woman in her early 20s, stands at a three-quarter angle from the left, upper body and face occupying the central frame. She is frozen mid-breath, gazing into the distance with a stunned expression, her pale face whitened by a drained blue-gray wash. [: The soft, out-of-focus background behind her reveals an unnaturally pristine and sterile bedroom wall and a tidy bed edge, entirely stripped of any bloodstains or bodies, bathed in a flat cold blue-gray light with a faint red spill barely grazing the room's edge.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다", "얼굴이 파리해지는 수리영.", "하얗게 질린 얼굴로 멍하니 멈춘 수리영의 상반신." ], "visual_inferences": [ "fixed_element sterile_bedroom_environment: unnaturally pristine, washed in cold blue-gray light", "Close framing emphasizes her stunned reaction by isolating her pale face against the unnaturally clean, soft background" ], "creative_decisions": [ "Applied a drained blue-gray wash to enhance the character's pale, stunned expression.", "Used a shallow depth of field to render the meticulously clean room as a soft, chilling backdrop without drawing focus away from her face." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "kitchenette", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "19aa7224f88c456e", "owned_hash": "3bf5d875fea06c07", "camera_direction_hash": "54cfa76bb65475b4", "owned_usage_hash": "636d7c56def09752", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "previous_shot_refs specifies zoom_in_detail from Shot 4, preserving the exact same space, lighting, and moment.", "_shot_index": 5, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Inside the bedroom at chest-to-eye height, the camera tightens slightly from the flow’s close-observation end position into 수리영’s pale upper body at a three-quarter frontal angle. Her face and shoulders occupy the central frame, while the clean room remains soft behind her, turning the changed axis from missing evidence to her stunned reaction.", "lighting_mood": "A drained blue-gray wash whitens 수리영’s face while faint red spill from outside barely grazes the room, suggesting dread without proof.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L04B06", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "sterile_bedroom_environment", "element_type": "environment_state", "character_name": "", "description": "The bedroom interior is unnaturally pristine, meticulously organized, and entirely stripped of any bloodstains, red markings, or bodies. The space is consistently washed in a cold, flat blue-gray light, creating a sterile and chilling atmosphere across all visible walls and surfaces.", "applies_to_shots": [ 4, 5, 12 ], "element_scope": "full", "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다", "Flat, cold blue-gray room light replaces the earlier red horror", "clean room remains soft behind her", "Dim blue-gray floor light" ], "visual_inferences": [ "The complete absence of the horror elements and the intensely spotless surfaces are maintained across the wide reveal, the background of the character's reaction, and the floor layout." ], "creative_decisions": [ "Using the pervasive, cold blue-gray tint across all shots to visually emphasize the unnatural cleanliness and the void left by the missing evidence." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 4, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "48c390deb97a9649" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "사라진 시신과 흔적, 깨끗해진 방", "representative_moment": "위치가 미묘하게 틀어진 채 바닥 구석에 놓여 있는 백팩과 인형의 섬뜩한 정지 구도.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "push-in close insert from a low floor-level angle", "t2i_prompt": "Photorealistic cinematic still. The bedroom corner baseboard is visible in the middle-center background, while a slightly shifted P12 backpack occupies the lower-left foreground, and a slightly shifted doll sits in the lower-right foreground. Deviating from the background reference's eye-level wide framing, this shot is a push-in close insert from a low floor-level angle. P12 sits closed on the pristine floor, its front pocket side angled partly toward the camera rather than lying square to the wall. The clean, still doll is positioned beside it, its face and torso turned slightly off-axis toward the low camera, separated from P12 by a small but suspicious gap. The baseboard runs horizontally behind the objects, meeting the completely clean, dry, unmarked wall at a visible corner. The same sterile room environment from the prior view is preserved, unnaturally organized and entirely stripped of any markings. The space is washed in a dim blue-gray floor light, with a faint gold edge catching on the fabric and plastic, highlighting the displacement.", "outfit_assignments": [], "source_facts": [ "수리영 방안에 놓여진 백팩, 인형. 약간씩 위치가 달라진.", "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고" ], "visual_inferences": [ "fixed_element sterile_bedroom_environment: pristine, completely cleaned, washed in dim blue-gray light", "fixed_element displaced_backpack_and_doll: small gap between objects, sitting on the floorboards near the baseboard", "the insert shot deviates from the prior wide room camera to strictly focus on the altered props near the floor" ], "creative_decisions": [ "Using a low floor-level push-in framing to emphasize the subtle shift of the props against the sterile baseboard", "Adding a faint gold edge light catching on the plastic and fabric to make the tiny displacement feel ritualistically arranged" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002", "fsc_003" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "P12 sits closed on the pristine floor" }, { "owned_token": "kitchenette", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "meeting the completely clean, dry, unmarked wall" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "53e3bfc01ef42e06", "owned_hash": "3bf5d875fea06c07", "camera_direction_hash": "1f63c44a425c2571", "owned_usage_hash": "5141e35682b2c254", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "floor", "violating_phrase": "P12 sits closed on the pristine floor", "reason": "The floor is referenced as a placement surface for the backpack, matching the anchor declaration as it is not being redrawn or modified.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "meeting the completely clean, dry, unmarked wall", "reason": "The wall is referenced to define the spatial boundary and background of the scene, matching the anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 12, "visible_entities": [ "L04", "P12" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 12 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From a low corner of the bedroom near floor level, the camera is static and angled slightly downward across the floor, converting the quiet-reveal stage into a still clue insert with no people present. The backpack and doll sit in the lower half of frame, separated by a small but suspicious gap, with the wall base and floorboards giving precise scale and placement.", "lighting_mood": "Dim blue-gray floor light with a faint gold edge on plastic and fabric makes the tiny displacement feel ritualistically arranged.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "bedroom corner baseboard", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "slightly shifted backpack", "screen_zone": "lower_left", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_003", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "slightly shifted doll", "screen_zone": "lower_right", "target_id": "", "target_kind": "background" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L04", "P12" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B06", "reference_usage": "exact_background", "owned_objects": [ "bed", "chair", "curtain", "door", "floor", "kitchenette", "shelf", "table", "track", "wall", "wardrobe", "window" ], "camera_reference": { "camera_position": "just inside number 10 (young bedroom door), facing number 5 (young adult bedroom) and number 24 (curtain), then allowing a view back toward number 16 (dining table).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "32mm moderate wide angle", "framing_notes": "Show the same room geometry as the prior bedroom event but clean; use number 36 for the spotless vanished-body state and number 37 for subtly shifted loose items." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "sterile_bedroom_environment", "element_type": "environment_state", "character_name": "", "description": "The bedroom interior is unnaturally pristine, meticulously organized, and entirely stripped of any bloodstains, red markings, or bodies. The space is consistently washed in a cold, flat blue-gray light, creating a sterile and chilling atmosphere across all visible walls and surfaces.", "applies_to_shots": [ 4, 5, 12 ], "element_scope": "full", "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다", "Flat, cold blue-gray room light replaces the earlier red horror", "clean room remains soft behind her", "Dim blue-gray floor light" ], "visual_inferences": [ "The complete absence of the horror elements and the intensely spotless surfaces are maintained across the wide reveal, the background of the character's reaction, and the floor layout." ], "creative_decisions": [ "Using the pervasive, cold blue-gray tint across all shots to visually emphasize the unnatural cleanliness and the void left by the missing evidence." ], "confidence": "high" }, { "element_id": "displaced_backpack_and_doll", "element_type": "persistent_prop", "character_name": "", "description": "A backpack and a doll are placed on the floorboards near the base of the wall. They sit with a small but suspicious gap between them, appearing slightly shifted from their usual positions but remaining entirely static.", "applies_to_shots": [ 4, 12 ], "element_scope": "full", "source_facts": [ "수리영 방안에 놓여진 백팩, 인형. 약간씩 위치가 달라진.", "bed, floor, curtain area, and wall are all visible at once", "The backpack and doll sit in the lower half of frame, separated by a small but suspicious gap" ], "visual_inferences": [ "Since the floor and overly ordered surfaces are fully visible in the wide room reveal (Shot 4), the specific arrangement of the displaced backpack and doll must be present exactly as detailed in the later insert shot (Shot 12)." ], "creative_decisions": [ "Keeping the small, distinct gap between the two props fixed across the wide spatial context and the tight clue insert to reinforce the unsettling feeling of the displacement." ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 4, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B06", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "6703c93376c7a938" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "망연자실한 수리영과 고립된 밤바다", "representative_moment": "어두운 버스정류장 벤치에 홀로 앉아 밤바다를 멍하니 바라보는 수리영(C01)을 멀리서 관찰하는 시선.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Bench-side height observer POV, establishing side composition showing the subject small against the dark coastal background, strong contrast between golden sodium spill and blue-black mist", "t2i_prompt": "Photorealistic cinematic still. Establishing observer perspective at bench-side height. The wet, empty bus stop bench cuts a horizontal line through the lower third of the frame, its seat line leading toward the right. C01O01 in a blue-gray hoodie, a Korean woman in her early 20s with long natural black hair and soft facial features, sits small but clear in the frame in right profile, slumped on the bench edge, gazing into the distance at the dark coastline. The bus stop shelter's dim, damp side posts and roof edge angle away from the camera into the heavy blue-black negative space and windy sea mist that swallow the deep background void. Cold coastal darkness dominates, with a weak, golden sodium-vapor light emitting from the shelter, casting a dim yellowish spill and thin rim light on the figure against the oppressive darkness of the surrounding roadside.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#15 action: '바다를 바라보며 멍한 채로 앉아있는 수리영'", "shot description: observer at dark coastal bus stop, looking at the figure from the side", "character angle: profile_right, slumped on bench edge, gazing into the distance" ], "visual_inferences": [ "fixed_element coastal_bus_stop_environment: weak golden sodium-vapor light, dim yellowish spill and thin rim light against oppressive darkness", "fixed_element coastal_bus_stop_environment: heavy blue-black negative space and sea mist", "bench cutting a horizontal line in the lower frame to anchor the figure in the wide composition without close-up mixing" ], "creative_decisions": [ "establishing side composition from a short distance to emphasize the subject's isolation within the void of the sea and mist", "cool color temperature for the night mood contrasted with the warm sodium spill to highlight the character's rim-lit silhouette against the dark background" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d15252b002f2abbe", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "163ef5c033018789", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "해당 없음", "_shot_index": 1, "visible_entities": [ "C01", "L11" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At bench-side height about eight feet from the bus stop bench, the camera holds the establishing side composition slightly seaward of 수리영, angled across her profile with the dark coastline behind her. She sits small but clear in the frame, the bench cutting a horizontal line through the lower third while the sea and mist swallow the background.", "lighting_mood": "Cold coastal darkness dominates, with weak sodium-gold bus stop spill barely separating 수리영 from the blue-gray sea haze.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L11" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "coastal_bus_stop_environment", "element_type": "environment_state", "character_name": "", "description": "The dark coastal bus stop area is engulfed in heavy blue-black negative space and sea mist. A weak, golden sodium-vapor light emits from the bus stop, casting a dim yellowish spill and thin rim light against the oppressive darkness of the surrounding roadside.", "applies_to_shots": [ 1, 4, 6 ], "element_scope": "full", "source_facts": [ "weak sodium-gold bus stop spill barely separating the character from the blue-gray sea haze", "thin gold bus-stop backlight against heavy blue-black road shadow", "tiny gold bus-stop light and faint red roadside reflections", "blue-black negative space" ], "visual_inferences": [ "The light source is likely a typical modern Korean streetlamp or a small illuminated bus stop sign isolated in the rural darkness." ], "creative_decisions": [ "Use high contrast between the sparse golden glow and the overwhelming cold blue-black shadows to emphasize isolation and vulnerability." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "c25a0455cbcb30f7" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "결심과 질주", "representative_moment": "C01이 진상을 확인하기 위해 어둠이 깔린 해안도로를 향해 달려가는 뒷모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "low tracking wide shot, high contrast of golden rim light against blue-black shadow", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a blue-gray hooded jacket, a young Korean woman in her early 20s, is captured mid-run in the middle-left midground moving toward the dark road edge in the upper-right background. The camera is positioned low and slightly behind her, placing her airborne foot in the lower frame. With her back completely to the camera and looking off-screen into the distance, her posture leans forward, hair flowing back in the direction opposite to motion. A weak golden sodium-vapor light from the wet bus stop bench and shelter, which are falling out of focus behind her, casts a dim yellowish spill and a thin gold backlight rimming her moving silhouette against the heavy blue-black negative space. Sea mist hangs over the slick, faintly reflective wet road surface, swallowing her figure into the oppressive darkness ahead.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "running away from the bench into the darker road edge", "back_to_camera, one foot airborne, mid-run pulling away", "weak sodium-gold bus stop spill barely separating the character" ], "visual_inferences": [ "fixed_element suriyoung_running_posture: mid-run pulling away with back turned completely", "fixed_element coastal_bus_stop_environment: dark coastal bus stop engulfed in sea mist and blue-black negative space", "hair flowing back indicates the momentum of her sudden sprint" ], "creative_decisions": [ "low camera angle explicitly places the airborne foot at the lower frame edge to emphasize motion and ground level perspective", "placed the character middle-left moving toward upper-right to fulfill the spatial contract and emphasize the trajectory into the unknown darkness" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "f2f8f13078bd628d", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "4e86bb8d5498b371", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 15, "dependency_reason": "ref_usage='exact_background' applies S15_Shot1's established bus stop room/space to this new instantaneous moment of departure", "_shot_index": 4, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Low and slightly behind 수리영, the camera tracks from six to ten feet back along the direction she runs away from the bench into the darker road edge. The frame catches her rear silhouette at the instant one foot hangs in the air, her body placed center-left and moving toward the upper-right darkness while the bus stop falls behind.", "lighting_mood": "A thin gold bus-stop backlight rims her moving silhouette against heavy blue-black road shadow, with no comforting fill ahead.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "dark road edge", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "dark road edge", "label": "수리영 running away", "screen_zone": "middle_left", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "coastal_bus_stop_environment", "element_type": "environment_state", "character_name": "", "description": "The dark coastal bus stop area is engulfed in heavy blue-black negative space and sea mist. A weak, golden sodium-vapor light emits from the bus stop, casting a dim yellowish spill and thin rim light against the oppressive darkness of the surrounding roadside.", "applies_to_shots": [ 1, 4, 6 ], "element_scope": "full", "source_facts": [ "weak sodium-gold bus stop spill barely separating the character from the blue-gray sea haze", "thin gold bus-stop backlight against heavy blue-black road shadow", "tiny gold bus-stop light and faint red roadside reflections", "blue-black negative space" ], "visual_inferences": [ "The light source is likely a typical modern Korean streetlamp or a small illuminated bus stop sign isolated in the rural darkness." ], "creative_decisions": [ "Use high contrast between the sparse golden glow and the overwhelming cold blue-black shadows to emphasize isolation and vulnerability." ], "confidence": "high" }, { "element_id": "suriyoung_running_posture", "element_type": "character_state", "character_name": "Suri-young", "description": "A young Korean woman seen from behind, running away from the glowing bus stop into the deeper darkness. Her back is turned completely toward the camera, and her body is captured in a mid-run posture, pulling away from the foreground.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "back_to_camera", "one foot airborne", "mid-run pulling away", "running away from the bench into the darker road edge" ], "visual_inferences": [ "Her posture leans forward slightly as she builds momentum away from the bus stop." ], "creative_decisions": [ "Keep her silhouette relatively dark but rim-lit by the distant golden light to highlight her hasty departure." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 15, "shot_index": 1, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "17a250d59ba6ce2d" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "어둠 속으로 사라지는 수리영과 지켜보는 시선", "representative_moment": "어두운 전경 너머로 C01이 정류장의 C02를 뒤로하고 밤바다를 향해 달려가는 모습을 감시하듯 멀리서 포착한다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "distant hidden eye-level voyeur POV, wide framing, high contrast between tiny golden light and deep blue-black shadow", "t2i_prompt": "Photorealistic cinematic still. Foreground darkness masking the watcher obscures the middle-center foreground, while C01O01 in a blue-gray hooded jacket, a Korean woman in her early 20s, represents 수리영 leaving frame in the lower-center background, moving toward the dark road beyond the stop, and C02O03 in a dark gray field jacket, a Korean woman in her early 40s, represents 혜수 near bus stop in the lower-left background, looking toward 수리영 leaving the bus stop. [L11: The dark coastal bus stop area is engulfed in heavy blue-black negative space and sea mist.] A weak, golden sodium-vapor light emits from the bus stop shelter, casting a dim yellowish spill and thin rim light against the oppressive darkness. C01O01 is positioned with her back to camera in a mid-run posture, one foot airborne, pulling away from the glowing bus stop into the deeper darkness. C02O03 stands at a three-quarter angle from the right, her eyes fixed on C01O01. The composition is captured from a distant hidden eye-level position across the dark roadside, keeping both figures small within the wide frame, surrounded by unlit misty blue-black void and faint red roadside reflections to heighten the surveillance-like tension.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C02", "outlook_id": "O03" } ], "source_facts": [ "Suri-young running away from the bus stop", "Hye-soo looking after Suri-young and calling out", "back_to_camera, mid-run pulling away", "blue-black negative space", "weak golden sodium-vapor light" ], "visual_inferences": [ "fixed_element coastal_bus_stop_environment: golden light casting yellowish spill and rim light", "fixed_element suriyoung_running_posture: mid-run posture, pulling away from foreground", "voyeuristic hidden vantage point obscured by foreground darkness" ], "creative_decisions": [ "wide framing through irregular dark edges to create surveillance-like gaze", "high contrast between small golden bus stop and overwhelming blue-black shadows" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002", "fsc_003" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d07e3194a071f1aa", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "c98e3fcaa6b35f16", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 15, "dependency_reason": "previous_shot_refs dictates atmosphere_reference to S15_Shot4", "_shot_index": 6, "visible_entities": [ "C01", "C02", "L11" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera hard-cuts to a distant hidden eye-level position across the dark roadside, static and partially masked by foreground blackness, as if an unseen watcher has stopped breathing. 혜수 and 수리영 are kept small within the bus stop area, framed through irregular dark edges so the prior physical tracking energy becomes an unnerving surveillance-like gaze.", "lighting_mood": "The image is mostly blue-black negative space, with tiny gold bus-stop light and faint red roadside reflections making the figures feel watched rather than protected.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "shared_space_relation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "foreground darkness masking the watcher", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "moves_toward", "gesture_target_label": "dark road beyond the stop", "label": "수리영 leaving frame", "screen_zone": "lower_center", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_003", "depth_plane": "background", "gesture_action": "looks_toward", "gesture_target_label": "수리영 leaving the bus stop", "label": "혜수 near bus stop", "screen_zone": "lower_left", "target_id": "C02", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "C02", "L11" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C02", "outlook_id": "O03" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "coastal_bus_stop_environment", "element_type": "environment_state", "character_name": "", "description": "The dark coastal bus stop area is engulfed in heavy blue-black negative space and sea mist. A weak, golden sodium-vapor light emits from the bus stop, casting a dim yellowish spill and thin rim light against the oppressive darkness of the surrounding roadside.", "applies_to_shots": [ 1, 4, 6 ], "element_scope": "full", "source_facts": [ "weak sodium-gold bus stop spill barely separating the character from the blue-gray sea haze", "thin gold bus-stop backlight against heavy blue-black road shadow", "tiny gold bus-stop light and faint red roadside reflections", "blue-black negative space" ], "visual_inferences": [ "The light source is likely a typical modern Korean streetlamp or a small illuminated bus stop sign isolated in the rural darkness." ], "creative_decisions": [ "Use high contrast between the sparse golden glow and the overwhelming cold blue-black shadows to emphasize isolation and vulnerability." ], "confidence": "high" }, { "element_id": "suriyoung_running_posture", "element_type": "character_state", "character_name": "Suri-young", "description": "A young Korean woman seen from behind, running away from the glowing bus stop into the deeper darkness. Her back is turned completely toward the camera, and her body is captured in a mid-run posture, pulling away from the foreground.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "back_to_camera", "one foot airborne", "mid-run pulling away", "running away from the bench into the darker road edge" ], "visual_inferences": [ "Her posture leans forward slightly as she builds momentum away from the bus stop." ], "creative_decisions": [ "Keep her silhouette relatively dark but rim-lit by the distant golden light to highlight her hasty departure." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 15, "shot_index": 4, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "character_outlook", "id": "C02O03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "e67e7e0d1610c2cf" }, { "scene_index": 16, "heading": "S#16. 경찰서 사무실 / N", "beat_title": "불길한 예감에 김형사에게 위치 추적을 요구하는 혜수", "representative_moment": "혜수가 핸드폰 화면을 김형사 눈앞에 들이밀며 다급하게 위치 추적을 요구하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close framing with shallow depth of field from a neutral observer's view, slightly below eye level, focusing tightly on the phone screen", "t2i_prompt": "Photorealistic cinematic still. [: A dimly lit police office interior during a night shift, bathed in a cold fluorescent-blue wash with damp-looking blue-gray shadows, a worn shared desk edge crossing the lower frame]. Captured from slightly below the seated man's eye line, the neutral observer's direct view uses shallow depth of field. Across the shared desk from each other, two figures interact. In the lower-center foreground, occupying about one third of the frame, a hand thrusts P05, a modern smartphone at a realistic handheld scale, close to the lens. The lit front face of P05 is angled slightly upward facing the camera, its pale glow displaying a saved contact interface. P05 and the gripping fingers are the sharp focal point. Just beyond the phone in the soft background, a Korean man in his 30s appears at three-quarter angle from the left, leaning toward the screen with eyes looking down, his face heavily blurred. Leading diagonally back from the foreground hand, the forearm connects to C02O03 in a dark grey field jacket, a Korean woman in her early 40s, visible at three-quarter angle from the right as a tense partial silhouette. C02O03 maintains a rigid, urgent posture, her chin tightened and angled forward, her brow distinctly furrowed, eyes fixed on the seated man.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O03" } ], "source_facts": [ "S#16. 경찰서 사무실 / N", "혜수 (핸드폰에 저장된 민숙번호 보여주며) 이 번호 위치 좀 따봐.", "이상해, 예감이 안 좋아." ], "visual_inferences": [ "fixed_element hyesu_tense_posture: C02 maintaining a rigid, urgent posture with a tightened chin and furrowed brow.", "fixed_element office_night_lighting: The office is bathed in a cold, low-key fluorescent-blue wash casting muted blue-gray shadows.", "The seated man (C05) is kept completely out of focus to emphasize the foreground evidence." ], "creative_decisions": [ "Close framing places the P05 smartphone prominently in the lower-center foreground as the focal point.", "Shallow depth of field isolates the phone and the thrusting hand while blurring the seated detective in the background.", "C02's tense partial silhouette is kept in the diagonal background to ground the spatial relationship across the shared desk." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "copier", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "notice board", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d53cff0fe28ff431", "owned_hash": "82dc57410da5a3ee", "camera_direction_hash": "a7d0d8a909b27244", "owned_usage_hash": "f03a2aa450b38fa8", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No dependent visual continuity from a prior scene applies directly to this specific instantaneous thrusting action.", "_shot_index": 3, "visible_entities": [ "C02", "C05", "P05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 16, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the tight over-the-desk position inherited from the flow, the camera sits slightly below 김형사's eye line, less than two feet from 혜수's extended hand; her fingers and the phone screen occupy the lower-center foreground at about one third of the frame, with 김형사's blurred eye line hovering just beyond it. This is a neutral observer's direct view, compressed through shallow depth of field so the evidence feels pushed into his face while 혜수's forearm leads diagonally back toward her tense partial silhouette.", "lighting_mood": "A cold fluorescent-blue office wash is punctured by the phone's pale glow, with damp-looking shadows and tiny red/gold desk accents kept subdued.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02", "C05", "P05" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O03" }, { "character_id": "C05", "outlook_id": "O07" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C05", "policy_type": "identity_reference", "policy": "generic_descriptor_allowed", "reason": "visible_shot_count=1 — selected-shot 저빈도 주변 subject" } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L12B02", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "hyesu_tense_posture", "element_type": "character_state", "character_name": "혜수", "description": "A tense Korean female police detective maintaining a rigid, urgent posture. Her chin is tightened and angled forward, with a fixed, serious stare directed off-camera towards a seated colleague. Her brow is distinctly furrowed, projecting a grim premonition.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "이상해, 예감이 안 좋아.", "tense partial silhouette", "chin tightened forward", "furrowed brow and fixed stare", "gaze=looks_at_character" ], "visual_inferences": [ "She is a Korean woman in practical detective attire.", "Her body language remains rigidly oriented toward the desk to demand immediate attention from her colleague." ], "creative_decisions": [ "Maintain the furrowed brow and intense gaze across different focal lengths to ground her emotional urgency." ], "confidence": "high" }, { "element_id": "office_night_lighting", "element_type": "environment_state", "character_name": "", "description": "A dimly lit police office interior during a night shift. The space is bathed in a cold, low-key fluorescent-blue wash. This harsh lighting casts muted, damp-looking blue-gray shadows, rendering the background office furniture and architectural details into soft, subdued shapes.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "S#16. 경찰서 사무실 / N", "A cold fluorescent-blue office wash", "Low-key fluorescent light sculpts her brow in muted blue-gray", "police office dissolves into soft blue-gray shapes" ], "visual_inferences": [ "The overall environment is mostly dark, typical of an empty office at night, except for the harsh direct light from specific fixtures." ], "creative_decisions": [ "Emphasize the cold, blue-gray tone of the environment to reflect the unsettling and isolating mood of the late-night realization." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [ { "scene_index": 16, "shot_index": 4, "description": "심각하게 눈썹을 찌푸린 혜수의 얼굴 클로즈업.", "keep_elements": [ { "label": "worn shared desk edge crossing the lower frame", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "dimly lit office interior with cold fluorescent-blue lighting and soft blue-gray shapes", "kind": "environment", "subject_kind": "non_human_visual_element" } ] }, { "scene_index": 21, "shot_index": 5, "description": "사진을 내려다보며 사색이 된 수리영의 얼굴.", "keep_elements": [ { "label": "worn desk edge crossing the lower frame", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "dimly lit police office interior bathed in a cold fluorescent-blue wash with blue-gray shadows", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C02O03", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "099f6ce9793aa458" }, { "scene_index": 16, "heading": "S#16. 경찰서 사무실 / N", "beat_title": "불길한 예감에 사로잡힌 혜수", "representative_moment": "C02가 미간을 찌푸린 채 강렬한 시선으로 오프스크린의 김형사 쪽을 응시하는 장면.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "observer perspective centered on the face, low-key fluorescent blue wash", "t2i_prompt": "Photorealistic cinematic still. C02O03 in a dark gray field jacket, a Korean woman in her early 40s with a composed adult face, occupying the center of the frame with her face and upper neck visible. At a three-quarter angle from the left, her chin is tightened and angled forward, her eyes fixed on an off-screen colleague. Her brow is distinctly furrowed, maintaining a rigid, urgent posture with a tense, serious stare. The visual focus lands squarely on her furrowed brow and intense expression, emphasizing her emotional urgency. Low-key fluorescent light sculpts her brow in muted blue-gray, leaving a narrow warm glint on her cheekbone like an omen. [: A dimly lit night-shift police office dissolves into soft blue-gray shapes behind her head; blurred institutional partitions and a small defocused red-gold paper surface form the soft secondary background, while the top edge of a desk is faintly implied at the bottom frame edge.]", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O03" } ], "source_facts": [ "S#16. 경찰서 사무실 / N", "이상해, 예감이 안 좋아.", "chin tightened forward, gaze=looks_at_character" ], "visual_inferences": [ "fixed_element hyesu_tense_posture: furrowed brow and fixed stare, grim premonition", "fixed_element office_night_lighting: low-key fluorescent-blue wash, soft subdued shapes" ], "creative_decisions": [ "Framed tightly on the face and neck to strictly follow the close framing requirement, eliminating full body parts to avoid view mixing.", "Included a defocused red-gold point and the faint edge of a desk in the background description to anchor the face and prevent it from floating in an empty void." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "copier", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "notice board", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "28e12d0557dc9af9", "owned_hash": "82dc57410da5a3ee", "camera_direction_hash": "7e09ee8c56d4db21", "owned_usage_hash": "f03a2aa450b38fa8", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 16, "dependency_reason": "zoom_in_detail from previous shot (Shot 3) preserving the exact moment, space, and lighting", "_shot_index": 4, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 16, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Following the flow's tilt up the line of 혜수's arm, the camera lands close in front of her face from just off 김형사's side of the desk, about two feet away, with the desk edge only implied below frame. The composition changes only the visual focus from evidence to emotion: her furrowed brow and fixed stare fill the center while the police office dissolves into soft blue-gray shapes behind her.", "lighting_mood": "Low-key fluorescent light sculpts her brow in muted blue-gray, leaving a narrow warm glint on her cheekbone like an omen.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O03" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L12B02", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "hyesu_tense_posture", "element_type": "character_state", "character_name": "혜수", "description": "A tense Korean female police detective maintaining a rigid, urgent posture. Her chin is tightened and angled forward, with a fixed, serious stare directed off-camera towards a seated colleague. Her brow is distinctly furrowed, projecting a grim premonition.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "이상해, 예감이 안 좋아.", "tense partial silhouette", "chin tightened forward", "furrowed brow and fixed stare", "gaze=looks_at_character" ], "visual_inferences": [ "She is a Korean woman in practical detective attire.", "Her body language remains rigidly oriented toward the desk to demand immediate attention from her colleague." ], "creative_decisions": [ "Maintain the furrowed brow and intense gaze across different focal lengths to ground her emotional urgency." ], "confidence": "high" }, { "element_id": "office_night_lighting", "element_type": "environment_state", "character_name": "", "description": "A dimly lit police office interior during a night shift. The space is bathed in a cold, low-key fluorescent-blue wash. This harsh lighting casts muted, damp-looking blue-gray shadows, rendering the background office furniture and architectural details into soft, subdued shapes.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "S#16. 경찰서 사무실 / N", "A cold fluorescent-blue office wash", "Low-key fluorescent light sculpts her brow in muted blue-gray", "police office dissolves into soft blue-gray shapes" ], "visual_inferences": [ "The overall environment is mostly dark, typical of an empty office at night, except for the harsh direct light from specific fixtures." ], "creative_decisions": [ "Emphasize the cold, blue-gray tone of the environment to reflect the unsettling and isolating mood of the late-night realization." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 16, "shot_index": 3, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C02O03", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "3191ec9923d5d1b1" }, { "scene_index": 17, "heading": "S#17. 옥탑방 / 동틀무렵", "beat_title": "새벽녘 옥탑방을 주시하는 검은 고양이", "representative_moment": "동트는 푸른 새벽, 난간 위 검은 고양이가 환하게 불 켜진 옥탑방 창문을 올려다보고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low eye-level wide angle, looking upward from the terrace floor", "t2i_prompt": "Photorealistic cinematic still. A small black cat with pointed ears, a slim face, and a dark fur silhouette sits in the lower-left midground, looking up toward the lit rooftop room window in the upper-right background. Shot from a low eye-level near the concrete floor, the camera is angled upward just below the animal's shoulder line. The cat is seen in right profile, neck craned upward, its raised muzzle pointing through the dense blue-gray dawn mist. The feline rests securely on the damp parapet wall. Beyond the drifting mist, the small window on the facade of the rooftop room leaks a restrained warm gold glow, contrasting sharply with the cold bluish predawn shadows wrapping the exterior.", "outfit_assignments": [], "source_facts": [ "S#17 action: '검은 고양이가 불이 켜져 있는 옥탑 방을 슬쩍 보더니 담을 넘는다'", "Shot 1 description: 동이 트는 푸르스름한 새벽, 불이 켜진 옥탑방 창문을 올려다보는 검은 고양이의 측면.", "Entity traits: black fur, small cat, pointed ears, slim face, dark silhouette" ], "visual_inferences": [ "Cat's state frozen in mid-gaze right before the jump, maintaining right profile and upward neck extension.", "Low camera angle places the parapet wall at the bottom and emphasizes the vertical height of the rooftop room facade.", "Dawn mist visually separates the lower-left foreground subject and the upper-right background light source." ], "creative_decisions": [ "Wide framing carefully distributes the two spatial anchor points (the cat and the window) across the diagonal axis.", "Used heavy blue-gray ambient shadows to isolate the glowing gold window and the cat's dark silhouette.", "Maintained single shutter moment, omitting any implied jumping action to focus purely on the eerie stillness." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "concrete floor", "usage_kind": "anchor", "source_phrase": "near the concrete floor" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "on the facade of the rooftop room" }, { "owned_token": "parapet wall", "usage_kind": "anchor", "source_phrase": "rests securely on the damp parapet wall" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "the lit rooftop room window" }, { "owned_token": "stair landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "terrace", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "the lit rooftop room window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "5ac466c9418ee5c9", "owned_hash": "6e01883df8f5586f", "camera_direction_hash": "7b49566ebb8a1fb3", "owned_usage_hash": "2bc033f5a1aceef4", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "concrete floor", "violating_phrase": "near the concrete floor", "reason": "The concrete floor is used as a camera placement reference, acting as a spatial anchor as declared by the producer.", "verdict": "anchor_reference" }, { "owned_object": "facade", "violating_phrase": "on the facade of the rooftop room", "reason": "The facade is used to define the location of the window, serving as a background reference anchor.", "verdict": "anchor_reference" }, { "owned_object": "parapet wall", "violating_phrase": "rests securely on the damp parapet wall", "reason": "The parapet wall is referenced as the physical support for the cat, acting as a composition anchor.", "verdict": "anchor_reference" }, { "owned_object": "rooftop room", "violating_phrase": "the lit rooftop room window", "reason": "The rooftop room is referenced to specify the location of the window, functioning as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "the small window on the facade of the rooftop room leaks a restrained warm gold glow", "reason": "The window serves as a composition and lighting anchor, rather than being redrawn as a new object.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 10, "dependency_reason": "exact_background (동일한 옥상 공간의 외경이므로 구조적 일관성 유지 필요)", "_shot_index": 1, "visible_entities": [ "C15", "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 17, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At rooftop level beside the 검은 고양이, the camera holds static just below the cat's shoulder line, angled upward so the cat sits crisp in lower-left side profile while the lit rooftop-room window hangs higher and farther in the upper-right background. The direct observer view uses the cat's raised muzzle as an arrow through the damp dawn mist, preserving the flow's upward gaze before any lateral movement begins.", "lighting_mood": "Bluish predawn ambience wraps the rooftop in wet shadows while the window leaks a restrained gold glow through mist.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "looks_to_anchor", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "lit rooftop room window", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "looks_toward", "gesture_target_label": "lit rooftop room window", "label": "black cat", "screen_zone": "lower_left", "target_id": "C15", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C15", "L03" ], "allowed_outlook_pairs": [ { "character_id": "C15", "outlook_id": "O00" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C15", "policy_type": "identity_reference", "policy": "generic_descriptor_allowed", "reason": "visible_shot_count=1 — selected-shot 저빈도 주변 subject" } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L03B01", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "concrete floor", "door", "drying platform", "facade", "parapet wall", "rooftop room", "stair landing", "terrace", "water tank", "window" ], "camera_reference": { "camera_position": "near number 1 (open rooftop terrace), facing toward number 6 (small facade window) and number 10 (perimeter parapet wall), with number 2 (rooftop room footprint) as the main mass", "camera_height": "low eye-level ~1.2m, slightly below parapet line", "lens_hint": "35mm wide angle", "framing_notes": "Blue dawn exterior view; include the lit window state and black cat on the parapet, while keeping all intrusion-night cues absent." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [ { "scene_index": 10, "shot_index": 6, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L03B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "b55923e927b5ba86" }, { "scene_index": 18, "heading": "S#18. 안방 안 / D", "beat_title": "욕실 거울 속 붉은 원 환각", "representative_moment": "A vivid red circle hallucination appears on the reflected bathroom wall in the fogged mirror, behind the soft reflection of Suyeong's shoulder.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Inside the bathroom at sink height, pressed close to the mirror plane, focusing tightly on the reflection with cold blue-gray lighting and a sharp red focal mark.", "t2i_prompt": "Photorealistic cinematic still. Inside the bathroom at sink height, the camera is pressed close to the glass plane, capturing [L04: a bathroom mirror surface with fog-dulled edges and wet specks, anchored by a thin curved wet porcelain sink rim grazing the lower foreground edge]. Within the reflection, a soft, out-of-focus slice of a young East Asian woman's shoulder appears at the left side of the composition, her arm bracing downward, gaze directed off-screen. The reflected space is dominated by the tiled wall behind her, where a vivid, impossible red circular mark sits sharp near the center-right on the clean tiles. A muted, cold blue-gray daylight fills the interior, casting deep blue-gray shadows over the bathroom tile and mirror fog, allowing the red mark to cut through the gloomy atmosphere with ritualistic intensity.", "outfit_assignments": [], "source_facts": [ "S#18 action: '거울에 비친 욕실 벽에 갑자기 또 발생하는 붉은 원.'", "shot description: Inside the bathroom at sink height, the camera is pressed close to the mirror plane and tightens within the reflection." ], "visual_inferences": [ "fixed_element house_eerie_lighting: A muted, cold blue-gray daylight fills the interior of the house, creating a gloomy and unsettling domestic atmosphere.", "The red circle hallucination cuts through the cold blue-gray ambient light with sharp contrast on the clean tiles." ], "creative_decisions": [ "Camera pressed close to the mirror to make the red circle within the reflection the dominant focal point.", "The character is rendered only as a soft shoulder slice at the edge to emphasize the psychological horror of the impossible mark on the wall." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [], "owned_object_usage": [ { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mirror", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photo frame", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "tile", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "toilet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "washbasin", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1c13222a75d6a044", "owned_hash": "44513e404db9eaf7", "camera_direction_hash": "a44f77d3c09ea6b9", "owned_usage_hash": "3726d58abf53961e", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct visual dependency specified.", "_shot_index": 10, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 10 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Inside the bathroom at sink height, the camera is pressed close to the mirror plane and tightens within the reflection so the reflected tiled wall fills most of the frame while the red circle sits sharp near center-right. 수리영 is only a soft reflected edge at the side of the composition, preserving the flow's mirror axis and making the viewer read the apparition as something seen in glass rather than on the real wall.", "lighting_mood": "Cold bathroom daylight turns blue-gray on tile and mirror fog, while the impossible red circle cuts through with ritualistic intensity.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L04" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L04B04", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "house_eerie_lighting", "element_type": "environment_state", "character_name": "", "description": "A muted, cold blue-gray daylight fills the interior of the house, creating a gloomy and unsettling domestic atmosphere. Deep blue-gray shadows dominate the space, unifying the different rooms under the same eerie ambient light.", "applies_to_shots": [ 10, 14 ], "element_scope": "full", "source_facts": [ "Cold bathroom daylight turns blue-gray on tile and mirror fog", "Muted daylight is drained into blue-gray domestic shadow" ], "visual_inferences": [ "The overarching interior lighting of the house maintains a consistent cold tone across different rooms to reflect the character's psychological horror." ], "creative_decisions": [ "Using a cohesive blue-gray color palette as a persistent environmental state to visually bridge the distinct hallucination sequences occurring in the bathroom and living room." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "115162f0690ede23" }, { "scene_index": 18, "heading": "S#18. 안방 안 / D", "beat_title": "식탁 위 사진 속 엄마 얼굴이 녹아내리는 환각", "representative_moment": "식탁 위 액자 속 아시아인 중년 여성의 얼굴이 환각에 의해 기괴하게 흘러내린다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Insert scale from a low table-height perspective, completely isolating the photograph against the dim background.", "t2i_prompt": "Photorealistic cinematic still. From a low table-height perspective, an isolated insert shot focuses entirely on the printed portrait inside the pre-existing photo frame resting on the table. Within this frame, the portrait of an Asian middle-aged woman—facing the camera with a smiling pose and eyes looking at the camera—is actively distorted by a terrifying hallucination. The face seems to sag and liquefy downward within the paper surface, her warm skin tones curdling into sickly red-gold smears. The table, sloping subtly away from the lens, is barely visible at the frame edges, submerged in a muted, cold blue-gray daylight that casts deep domestic shadows, leaving the physical boundaries of the room unnervingly still.", "outfit_assignments": [], "source_facts": [ "Shot 14: 식탁 위에 놓인 사진 속, 환하게 웃는 한국인 중년 여성의 얼굴이 기괴하게 흘러내리는 찰나 클로즈업.", "DP Note: The image is perceived as a hallucination: the face seems to sag and liquefy within the printed photo surface.", "DP Note: Muted daylight is drained into blue-gray domestic shadow, with the photo's warm skin tones curdling into sickly red-gold smears." ], "visual_inferences": [ "fixed_element house_eerie_lighting: A muted, cold blue-gray daylight fills the interior, creating a gloomy and unsettling domestic atmosphere.", "The tabletop and the edges of the photo frame remain sharp and still, emphasizing that the melting effect is an isolated hallucination occurring only within the picture." ], "creative_decisions": [ "Framed as an extreme insert to fill the central field with the melting face, leaving only slivers of the table and frame trim to provide scale and physical grounding.", "Contrasted the warm, curdled red-gold smears of the face against the cold blue-gray ambient shadows of the room to heighten the horror." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mirror", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photo frame", "usage_kind": "anchor", "source_phrase": "inside the pre-existing photo frame" }, { "owned_token": "table", "usage_kind": "anchor", "source_phrase": "resting on the table" }, { "owned_token": "tile", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "toilet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "washbasin", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "00dec9b201abf16b", "owned_hash": "44513e404db9eaf7", "camera_direction_hash": "6be6bf461252c288", "owned_usage_hash": "2c5503bf5f737676", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "photo frame", "violating_phrase": "inside the pre-existing photo frame resting on the table", "reason": "The prompt references the pre-existing photo frame as a spatial anchor containing the distorted portrait, with no instruction to redraw or modify the frame itself, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "table", "violating_phrase": "resting on the table. ... The table, sloping subtly away from the lens, is barely visible", "reason": "The table serves purely as a physical surface for the photo frame and is barely visible at the edges, acting as a spatial anchor, which is consistent with the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "이전 샷(욕실 거울 앞)의 푸른 회색조 조명 톤 유지 및 거실 식탁 위 액자 묘사 연속성", "_shot_index": 14, "visible_entities": [], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 14 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At dining-table height, the camera pushes closer than the base flow until the framed photograph becomes an isolated insert, the mother's smiling face occupying the central field while the frame edge and tabletop barely remain as scale references. The image is perceived as a hallucination: the face seems to sag and liquefy within the printed photo surface, while the surrounding room stays unnervingly still.", "lighting_mood": "Muted daylight is drained into blue-gray domestic shadow, with the photo's warm skin tones curdling into sickly red-gold smears during the hallucination.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B04", "reference_usage": "exact_background", "owned_objects": [ "chair", "counter", "door", "faucet", "mirror", "photo frame", "table", "tile", "toilet", "washbasin" ], "camera_reference": { "camera_position": "inside number 6 (bathroom service zone), near number 25 (washbasin), facing number 27 (bathroom mirror), with a secondary cutaway path back toward number 2 (shared living dining zone) and number 28 (family photo frame).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle, slightly tight for small bathroom", "framing_notes": "For the bathroom view, feature numbers 25 and 27; for the living-area continuation, feature numbers 16 and 28. Overlay use includes numbers 38 and 39 only as subjective visual states." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "house_eerie_lighting", "element_type": "environment_state", "character_name": "", "description": "A muted, cold blue-gray daylight fills the interior of the house, creating a gloomy and unsettling domestic atmosphere. Deep blue-gray shadows dominate the space, unifying the different rooms under the same eerie ambient light.", "applies_to_shots": [ 10, 14 ], "element_scope": "full", "source_facts": [ "Cold bathroom daylight turns blue-gray on tile and mirror fog", "Muted daylight is drained into blue-gray domestic shadow" ], "visual_inferences": [ "The overarching interior lighting of the house maintains a consistent cold tone across different rooms to reflect the character's psychological horror." ], "creative_decisions": [ "Using a cohesive blue-gray color palette as a persistent environmental state to visually bridge the distinct hallucination sequences occurring in the bathroom and living room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 13, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B04", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "64f0f146ddbe20fb" }, { "scene_index": 19, "heading": "S#19. 마트 / D", "beat_title": "CCTV 위치 확인과 절박한 시선", "representative_moment": "수리영(C01)이 처마 밑 어두운 그늘에서 고개를 치켜든 채 오프스크린의 CCTV를 빤히 올려다보고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close frontal framing from slightly below eye level, contrasting cool blue-gray morning shadows on the face with a faint warm background spill", "t2i_prompt": "Photorealistic cinematic still. A low observer position slightly below eye level holds the face and neck in close framing. C01O01 in a dark gray hooded jacket, a Korean woman in her early 20s with a youthful oval face, long natural black hair, and soft facial features, is facing the camera with her chin lifted tense, looking up intently at an off-screen point above. The space around her is tightly compressed by [: the underside of the market awning receding above her head and a weathered pale exterior wall corner forming a hard vertical line beside her]. Cool morning light flattens into deep blue-gray shadow across her skin, contrasting with a faint warm gold light spilling from the out-of-focus smudged shop entrance glass behind her.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Shot description: CCTV 카메라 위치를 확인하다. 찍혔을 만한 위치다.", "Character angles: C01 facing_camera, chin lifted tense, gaze=up, state=alive.", "Lighting mood: Cool morning light under the awning flattens into blue-gray shadow, with a faint gold spill from the market entrance behind her." ], "visual_inferences": [ "Fixed element (market_exterior_awning_environment) confirms the overhead awning blocks ambient daylight, placing her immediate area in deep cool shadow.", "The low camera position naturally emphasizes the tension in her raised chin and her intense upward gaze." ], "creative_decisions": [ "Applied close framing to tightly focus on her face and neck, isolating her emotional intensity from the wider environment.", "Used a [L00: ...] block to describe the immediate frame-edge surfaces without explicitly invoking the omitted background reference image." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "CCTV camera", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ceiling", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "display island", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "refrigerated case", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "2d0acf6768b30108", "owned_hash": "ca0ffcfc4a17ab57", "camera_direction_hash": "bda4f1e46fa7ee49", "owned_usage_hash": "b5c2b0e97bbb6304", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 7, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 19, "shot_index": 7 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Outside under the market awning, the camera remains slightly below 수리영's eye level near the wall corner and frames her more frontally as requested, holding her upturned face in close view while the CCTV stays offscreen above her line of sight. The flow's tilt is felt in the low angle and raised chin, but the composition steadies on her stare, with the awning edge and wall compressing the space around her.", "lighting_mood": "Cool morning light under the awning flattens into blue-gray shadow, with a faint gold spill from the market entrance behind her.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L08B02", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "market_exterior_awning_environment", "element_type": "environment_state", "character_name": "", "description": "The exterior space outside the market entrance, specifically near a wall corner under an awning. The area is enveloped in cool, blue-gray morning shadows created by the overhead awning, while a faint, warm golden light spills out and reflects from the shop's glass entrance behind.", "applies_to_shots": [ 7, 8 ], "element_scope": "full", "source_facts": [ "이른 아침, 마트 밖", "Cool morning light under the awning flattens into blue-gray shadow, with a faint gold spill from the market entrance behind her.", "Soft overcast daylight is suppressed by the awning into tense blue-gray tones, with a small warm reflection from the shop glass." ], "visual_inferences": [ "The awning physically blocks the overcast morning light from above, keeping the immediate wall and entrance area in deep cool shadow.", "The interior market lights are turned on, providing the warm golden contrast seen through and reflecting off the entrance glass." ], "creative_decisions": [ "Contrasting the cool blue-gray exterior shadows with the warm interior spill to highlight the early morning atmosphere and the tense mood of the character standing outside." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "1991991b76a1f014" }, { "scene_index": 19, "heading": "S#19. 마트 / D", "beat_title": "CCTV 확인 후 전화", "representative_moment": "C01O01이 마트 밖에서 P05 핸드폰을 귀에 댄 채 심각한 표정으로 어딘가를 응시하며 혜수에게 전화를 건다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-to-eye height upper-body composition, cool blue-gray morning shadows contrasting with warm interior spill", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a dark blue-gray hooded jacket, a Korean woman in her early 20s with long natural black hair, framed from the chest up at a three-quarter angle from the right. She holds a sleek dark P05 smartphone pressed against her ear, clenching her jaw with a serious expression, gazing into the distance. The scene is enveloped in cool, blue-gray morning shadows cast by the dull, damp overhead awning stretching across the upper background. A faint, warm golden light spills out from the market interior, catching softly along her cheek. The weathered, muted exterior wall runs behind her shoulder, while the smudged glass entrance door and window sit softly behind her at a slight angle, showing a dim counter and shelf inside. Chest-to-eye height observer view, maintaining an overcast morning atmosphere with tense blue-gray tones contrasting the warm reflection.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#19 text: 마트 밖 엄마와 남자 직원 CCTV 위치를 확인하다. 수리영, 혜수에게 전화 걸어 상황을 설명한다.", "Shot extract: 핸드폰을 귀에 댄 채 심각한 표정으로 서 있는 수리영의 상반신.", "DP color/lighting: Soft overcast daylight is suppressed by the awning into tense blue-gray tones, with a small warm reflection from the shop glass along her cheek." ], "visual_inferences": [ "fixed_element market_exterior_awning_environment applied: The exterior space outside the market entrance near a wall corner under an awning. Faint warm light spills out.", "DP background mapping: smudged market entrance glass, dull shadow-casting awning edge, weathered exterior wall." ], "creative_decisions": [ "Positioned the camera at chest-to-eye height to emphasize her upper body tension while leaving the market interior softly visible in the background to anchor the location.", "Used a cool blue-gray color palette with a single warm highlight on her cheek to increase the suspenseful, tense morning mood." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "CCTV camera", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ceiling", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "anchor", "source_phrase": "showing a dim counter" }, { "owned_token": "display island", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "smudged glass entrance door" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "refrigerated case", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "anchor", "source_phrase": "and shelf inside" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "and window sit softly behind her at a slight angle" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e48f71bf5955bef9", "owned_hash": "ca0ffcfc4a17ab57", "camera_direction_hash": "c77f28760e737a68", "owned_usage_hash": "fe60742cb9596604", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "counter", "violating_phrase": "showing a dim counter", "reason": "The prompt mentions the counter as part of the background interior visible through the glass, which acts as an anchor reference and does not request redrawing it.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "smudged glass entrance door", "reason": "The prompt references the door sitting softly behind the character to establish composition, matching the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "shelf", "violating_phrase": "and shelf inside", "reason": "The shelf is mentioned passively inside the market to anchor the background setting, with no instruction to redraw or modify it.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "and window sit softly behind her at a slight angle", "reason": "The window is mentioned as a soft background element for spatial framing, aligning with the anchor reference declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 19, "dependency_reason": "S19_Shot7 배경과 장소가 이어짐 (마트 입구 내외부의 빛, 구조 및 연속적인 시간선)", "_shot_index": 8, "visible_entities": [ "C01", "P05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 19, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera settles from the previous upward implication to chest-to-eye height in front of 수리영, tightening to an upper-body frame while she holds the phone to her ear. Only the camera distance changes from the prior close face shot: the wall and entrance area sit softly behind and beside her as she explains the CCTV possibility with a clenched, serious expression.", "lighting_mood": "Soft overcast daylight is suppressed by the awning into tense blue-gray tones, with a small warm reflection from the shop glass along her cheek.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "P05" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L08B02", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "ceiling", "counter", "display island", "door", "floor", "light", "refrigerated case", "shelf", "window" ], "camera_reference": { "camera_position": "near number 3 (main sales floor shared zone), facing back toward number 16 (overhead CCTV camera) and along number 17 (fixed CCTV coverage axis) toward number 6 (front automatic entrance)", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "Keep the interior sparse and clear; emphasize the line from the front CCTV camera across the sales floor and do not include the temporary stocking boxes." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "market_exterior_awning_environment", "element_type": "environment_state", "character_name": "", "description": "The exterior space outside the market entrance, specifically near a wall corner under an awning. The area is enveloped in cool, blue-gray morning shadows created by the overhead awning, while a faint, warm golden light spills out and reflects from the shop's glass entrance behind.", "applies_to_shots": [ 7, 8 ], "element_scope": "full", "source_facts": [ "이른 아침, 마트 밖", "Cool morning light under the awning flattens into blue-gray shadow, with a faint gold spill from the market entrance behind her.", "Soft overcast daylight is suppressed by the awning into tense blue-gray tones, with a small warm reflection from the shop glass." ], "visual_inferences": [ "The awning physically blocks the overcast morning light from above, keeping the immediate wall and entrance area in deep cool shadow.", "The interior market lights are turned on, providing the warm golden contrast seen through and reflecting off the entrance glass." ], "creative_decisions": [ "Contrasting the cool blue-gray exterior shadows with the warm interior spill to highlight the early morning atmosphere and the tense mood of the character standing outside." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 19, "shot_index": 7, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "background", "id": "L08B02", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "0fe0614c68ee6f32" }, { "scene_index": 20, "heading": "S#20. 마트 총무과 / 실내", "beat_title": "식별 불가능한 영상과 증폭되는 불안", "representative_moment": "어두운 보안실 안, 빛을 발하는 감시 모니터 앞에 나란히 서서 화면을 주시하는 혜수(C02)와 보안 직원의 뒷모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide eye-level observer view, crushed blue-black surveillance shadows, cold blue screen glow", "t2i_prompt": "Photorealistic cinematic still. Wide eye-level observer view from the rear threshold of the dim security office. The environment is heavily shadowed in deep, crushed blue-black tones. In the lower-middle foreground, C02O03 in a dark gray field jacket, a Korean woman in her early 40s, stands with her back to camera, leaning over the shared desk, her gaze directed off-screen. Beside her, a Korean adult man in a security uniform stands with his back to camera, one hand resting on the console, also gazing off-screen. Both figures appear as dark back silhouettes facing the desk monitor, their body volume preserved with cold blue rim light tracing their shoulders against the dim ambient fill. The desk monitor sits prominently ahead; its front screen faces away from the camera but casts visible cold blue light around its thin bezel edges toward the characters, accompanied by faint red indicator lights pricking through the heavy shadow. The workspace stretches forward, with a staff chair and equipment cabinet receding into the damp, unnerving blue-black background.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O03" } ], "source_facts": [ "Shot description: 어두운 보안실 안, 책상 위 모니터 화면을 응시하며 나란히 서 있는 혜수와 한국인 보안 직원의 뒷모습", "C02 Angle: back_to_camera, leaning over desk, gaze=off_screen", "C08 Angle: back_to_camera, hand on console, gaze=off_screen" ], "visual_inferences": [ "fixed_element security_office_lighting: heavily shadowed in deep, crushed blue-black tones", "fixed_element cctv_monitor: monitor acting as the focal point, emitting cold blue glow", "Observer view establishes spatial depth from the threshold to the desk and monitor" ], "creative_decisions": [ "Rim-lighting the back silhouettes with cold blue to preserve body volume and avoid a flat 2D cutout effect", "Emphasizing the contrast between the crushed blue-black shadows of the background and the cold glow of the monitor bezel" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "anchor", "source_phrase": "with a staff chair" }, { "owned_token": "desk", "usage_kind": "anchor", "source_phrase": "leaning over the shared desk" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "equipment cabinet", "usage_kind": "anchor", "source_phrase": "equipment cabinet receding into the damp, unnerving blue-black background" }, { "owned_token": "monitor", "usage_kind": "anchor", "source_phrase": "The desk monitor sits prominently ahead" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "f151d619c3bb2120", "owned_hash": "99dbbb6ec83d3c35", "camera_direction_hash": "d4cfda16b42107f1", "owned_usage_hash": "23c3b46e4ef82cbf", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "chair", "violating_phrase": "with a staff chair", "reason": "The staff chair is mentioned as receding into the background to establish depth, functioning as a spatial anchor, which is consistent with the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "desk", "violating_phrase": "leaning over the shared desk", "reason": "The desk is referenced solely to anchor the character's pose and position in the environment, matching the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "equipment cabinet", "violating_phrase": "equipment cabinet receding into the damp, unnerving blue-black background", "reason": "The equipment cabinet is described as receding into the background for depth and mood, which constitutes an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "monitor", "violating_phrase": "The desk monitor sits prominently ahead", "reason": "The monitor is described for its screen orientation and lighting emission to set the scene's ambient lighting, acting as a spatial and visual anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 1, "visible_entities": [ "C02", "C08", "L14" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 20, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the rear of the dim security office at eye level, the camera holds about eight feet behind the desk, centered on 혜수 and the 한국인 보안 직원 as dark back silhouettes facing the monitor. This direct observer view establishes the room layout: their shoulders occupy the lower middle foreground, the desk and glowing monitor sit ahead, and the surrounding office recedes into damp blue-black corners.", "lighting_mood": "The office is lit almost entirely by the monitor's cold blue glow, with faint red and gold indicator lights pricking through the heavy shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02", "C08", "L14" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O03" }, { "character_id": "C08", "outlook_id": "O10" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C08", "policy_type": "identity_reference", "policy": "generic_descriptor_allowed", "reason": "visible_shot_count=1 — selected-shot 저빈도 주변 subject" } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L14B01", "reference_usage": "exact_background", "owned_objects": [ "cabinet", "chair", "desk", "door", "equipment cabinet", "monitor", "window" ], "camera_reference": { "camera_position": "near number 4 (threshold zone), facing diagonally toward number 7 (security desk), number 8 (CCTV monitor bank), and number 9 (staff chair), with number 3 (visitor viewing zone) in the foreground", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "include the monitor bank prominently for the playback cue number 13; keep the visitor viewing area and staff workstation relationship clear" }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "cctv_monitor", "element_type": "persistent_prop", "character_name": "", "description": "A desk monitor with a visible thin bezel, displaying a paused low-resolution CCTV image. One side of the footage on the screen is swallowed by a distinct, unnaturally heavy dark patch that obscures the detail. The screen acts as the focal point of the room, emitting a prominent cold blue glow.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "그 부분만 유독 어두워 식별이 불가능하다", "정지된 모니터 속 저해상도 이미지", "화면의 한쪽이 짙은 어둠에 가려진 채" ], "visual_inferences": [ "The monitor has a physical bezel framing the screen.", "The dark patch on the display feels like a physical void rather than simple underexposure." ], "creative_decisions": [ "Emphasizing the contrast between the cold blue light of the monitor and the heavy, unnatural dark patch on the screen to build tension." ], "confidence": "high" }, { "element_id": "security_office_lighting", "element_type": "environment_state", "character_name": "", "description": "The security office environment is heavily shadowed in deep, crushed blue-black tones. The primary illumination comes strictly from the glowing monitor screen. This lighting arrangement creates a cold, dim, and unnerving surveillance atmosphere across the visible space.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "어두운 보안실 안", "Cold monitor light and crushed blue-black shadows create a surveillance-horror stillness" ], "visual_inferences": [ "The room lacks any bright overhead lighting, relying almost entirely on the screen's glow." ], "creative_decisions": [ "Using dominant blue-black shadows to emphasize the isolation and eerie mood of the security room." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C02O03", "policy": "required" }, { "kind": "background", "id": "L14B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "dc506872b268aae5" }, { "scene_index": 20, "heading": "S#20. 마트 총무과 / 실내", "beat_title": "식별 불가능한 진실", "representative_moment": "A tight view of the paused CCTV monitor, its screen swallowed by an unnaturally heavy dark patch emitting a cold blue glow.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Insert framing squared to the monitor glass, cold blue screen glow with crushed black shadows", "t2i_prompt": "Photorealistic cinematic still. Looking closely at the pre-existing desk monitor, the screen surface nearly fills the frame, with only the dusty black plastic bezel visible at the edges. On this existing monitor screen, a paused, low-resolution CCTV image is displayed, characterized by heavy digital pixelation. One side of the footage is completely swallowed by an unnaturally heavy, deep dark patch that absorbs all detail like a physical void. On the remaining visible side, vague, unidentifiable pixelated silhouettes of figures stand in conversation. The screen emits a prominent cold blue glow. The smudged glass faintly reflects dim office indicator lights and the crushed blue-black shadows of the surrounding security room environment, anchoring the device in an unnerving stillness.", "outfit_assignments": [], "source_facts": [ "정지된 모니터 속 저해상도 이미지", "화면의 한쪽이 짙은 어둠에 가려진 채", "그 부분만 유독 어두워 식별이 불가능하다" ], "visual_inferences": [ "fixed_element cctv_monitor: visible thin bezel, emitting prominent cold blue glow", "fixed_element security_office_lighting: crushed blue-black tones", "smudged monitor glass faintly reflecting dim office indicator lights" ], "creative_decisions": [ "Explicitly stating the deviation from the wide room camera framing to justify the monitor filling the frame.", "Using the monitor's cold blue glow as the sole light source to emphasize the deep, crushed shadows and unnatural void of the dark patch." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "anchor", "source_phrase": "pre-existing desk monitor" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "equipment cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "anchor", "source_phrase": "Looking closely at the pre-existing desk monitor, the screen surface nearly fills the frame" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "2167c2ba6292f0fd", "owned_hash": "99dbbb6ec83d3c35", "camera_direction_hash": "57f8eff9581f30ff", "owned_usage_hash": "013d79aaac67557a", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "desk", "violating_phrase": "pre-existing desk monitor", "reason": "The desk is mentioned only as a modifier to locate the pre-existing monitor, serving as an spatial anchor with no instructions to redraw the desk itself.", "verdict": "anchor_reference" }, { "owned_object": "monitor", "violating_phrase": "Looking closely at the pre-existing desk monitor, the screen surface nearly fills the frame, with only the dusty black plastic bezel visible at the edges", "reason": "The monitor is referenced as an existing anchor to frame the shot and display a CCTV image on its screen, with no intention of redrawing or altering the monitor structure.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 20, "dependency_reason": "prior shot S20_Shot1 for exact_background continuity of the security office environment and monitor prop", "_shot_index": 3, "visible_entities": [], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 20, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Now directly in front of the desk monitor, the camera squares to the glass surface less than two feet away, cropping tight so the monitor frame nearly fills the shot while still leaving a thin bezel as a physical boundary. The viewer sees the paused low-resolution CCTV image through the device screen, with one side swallowed by unnatural darkness that feels heavier than simple underexposure.", "lighting_mood": "Cold monitor light and crushed blue-black shadows create a surveillance-horror stillness, with the dark patch absorbing nearly all detail.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L14B01", "reference_usage": "exact_background", "owned_objects": [ "cabinet", "chair", "desk", "door", "equipment cabinet", "monitor", "window" ], "camera_reference": { "camera_position": "near number 4 (threshold zone), facing diagonally toward number 7 (security desk), number 8 (CCTV monitor bank), and number 9 (staff chair), with number 3 (visitor viewing zone) in the foreground", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "include the monitor bank prominently for the playback cue number 13; keep the visitor viewing area and staff workstation relationship clear" }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "cctv_monitor", "element_type": "persistent_prop", "character_name": "", "description": "A desk monitor with a visible thin bezel, displaying a paused low-resolution CCTV image. One side of the footage on the screen is swallowed by a distinct, unnaturally heavy dark patch that obscures the detail. The screen acts as the focal point of the room, emitting a prominent cold blue glow.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "그 부분만 유독 어두워 식별이 불가능하다", "정지된 모니터 속 저해상도 이미지", "화면의 한쪽이 짙은 어둠에 가려진 채" ], "visual_inferences": [ "The monitor has a physical bezel framing the screen.", "The dark patch on the display feels like a physical void rather than simple underexposure." ], "creative_decisions": [ "Emphasizing the contrast between the cold blue light of the monitor and the heavy, unnatural dark patch on the screen to build tension." ], "confidence": "high" }, { "element_id": "security_office_lighting", "element_type": "environment_state", "character_name": "", "description": "The security office environment is heavily shadowed in deep, crushed blue-black tones. The primary illumination comes strictly from the glowing monitor screen. This lighting arrangement creates a cold, dim, and unnerving surveillance atmosphere across the visible space.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "어두운 보안실 안", "Cold monitor light and crushed blue-black shadows create a surveillance-horror stillness" ], "visual_inferences": [ "The room lacks any bright overhead lighting, relying almost entirely on the screen's glow." ], "creative_decisions": [ "Using dominant blue-black shadows to emphasize the isolation and eerie mood of the security room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 20, "shot_index": 1, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L14B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "2b893eb3e7a826bb" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "엄마의 위치 단서와 사색이 된 수리영", "representative_moment": "사무실에 홀로 남은 수리영(C01)이 책상 위 살인 사건 사진들을 내려다보며 창백하게 사색이 된 정면.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low close-up over the desk edge with shallow depth of field, cool blue-gray and cyan tones", "t2i_prompt": "Photorealistic cinematic still. A low camera at desk level captures the same figure, C01O01 in a blue-gray hooded jacket, a Korean woman in her early 20s, seated at three-quarter angle from the left and hunched forward over the desk. She is facing the camera, her head tilted down with eyes fixed on the scattered crime scene photographs. Her face appears visibly pale and drained of color, maintaining a tense, shocked expression. In the immediate foreground, the worn near edge of the desk crosses the bottom frame, and a scattered pile of disturbing crime scene photographs sits low on the same surface. The printed image sides of the photographs face upward, appearing deliberately out of focus with muted red flecks bleeding softly. The same room is preserved behind her as a soft, emptied police office bathed in cold, low-key lighting that casts deep blue-gray shadows and an icy cyan glow across the defocused background.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "사색이 된 수리영", "hunched over photos", "disturbing photographs blurred across the lower foreground", "Flat daytime office light is cooled into corpse-like blue-gray" ], "visual_inferences": [ "fixed_element character_state: C01O01 is seated and hunched over a police office desk, maintaining a tense expression", "fixed_element persistent_prop: scattered crime scene photographs spread out on the desk", "fixed_element environment_state: cold, low-key lighting with cyan monitor glows" ], "creative_decisions": [ "Close framing from a low angle incorporates the desk edge and out-of-focus photographs in the foreground, physically connecting the camera to her POV.", "A shallow depth of field isolates her drained, pale expression against the soft, emptied blue-gray background to emphasize her internal shock." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "ceiling", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "copier", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "filing cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "notice board", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "printer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "929196697ceeecbc", "owned_hash": "9f13f326b3d6be40", "camera_direction_hash": "5f0ff7488c346b37", "owned_usage_hash": "ff3a4a9d25a43450", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 16, "dependency_reason": "zoom_in_detail ref_usage requires preserving the same room, surface, and lighting established in the prior shot", "_shot_index": 5, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Low over the desk edge, the camera keeps the disturbing photographs blurred across the lower foreground and dollies in toward 수리영's drained face as she looks down at them. The flow's angle from photographs up to her reaction is preserved, but the frame tightens so her pallor and frozen breath become the emotional center while the police office remains a soft, emptied background.", "lighting_mood": "Flat daytime office light is cooled into corpse-like blue-gray on her face, with muted red flecks from the photographs bleeding softly in the foreground.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L12B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "su_ri_young_pale_hunched_posture", "element_type": "character_state", "character_name": "수리영", "description": "A young East Asian female teenager is seated and hunched over a police office desk. Her face appears visibly pale and drained of color, maintaining a tense, shocked expression under the cold office lighting.", "applies_to_shots": [ 5, 8 ], "element_scope": "close", "source_facts": [ "사색이 된 수리영", "hunched over photos", "hunched toward monitor", "pallor and frozen breath", "pale face" ], "visual_inferences": [ "She is wearing casual teenage clothing suitable for a modern daytime setting." ], "creative_decisions": [ "Keeping her upper body consistently hunched forward emphasizes her intense focus and psychological weight from the disturbing discoveries." ], "confidence": "high" }, { "element_id": "scattered_crime_scene_photos", "element_type": "persistent_prop", "character_name": "", "description": "A scattered pile of disturbing crime scene photographs is spread out on the desk. These photos sit low in the foreground, appearing slightly out of focus and displaying muted red tones.", "applies_to_shots": [ 5, 8 ], "element_scope": "close", "source_facts": [ "혜수 책상위에는 살인사건 확대 사진들이 어지럽게 널려있다", "disturbing photographs blurred across the lower foreground", "muted red flecks from the photographs", "crime-scene photo edges sit low in the foreground" ], "visual_inferences": [ "The photos have standard white borders and contain gruesome visual details that are obscured by the shallow depth of field." ], "creative_decisions": [ "Blurring the photos in the foreground ensures they remain a constant, morbid presence without stealing focus from the character's reactions." ], "confidence": "high" }, { "element_id": "police_office_blue_gray_lighting", "element_type": "environment_state", "character_name": "", "description": "The police station office is bathed in cold, low-key lighting. The ambient illumination casts deep blue-gray shadows and icy cyan glows from nearby monitors across the desk area.", "applies_to_shots": [ 5, 8, 10 ], "element_scope": "full", "source_facts": [ "Flat daytime office light is cooled into corpse-like blue-gray", "Low-key police-office fluorescents mix with icy monitor light", "cyan monitor glow falls off into damp blue-gray shadow" ], "visual_inferences": [ "The light source is primarily artificial overhead fluorescents and computer screens, with minimal warm natural sunlight." ], "creative_decisions": [ "Using dominant blue-gray and cyan tones establishes a chilling, clinical atmosphere that mirrors the grim nature of the crime scene photos and the GPS location discovery." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 16, "shot_index": 3, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [ { "scene_index": 21, "shot_index": 8, "description": "눈을 동그랗게 뜨고 모니터 화면을 응시하는 수리영의 정면.", "keep_elements": [ { "label": "scattered crime scene photographs on the desk", "kind": "static_prop", "subject_kind": "non_human_visual_element" }, { "label": "worn near edge of the desk crossing the bottom frame", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "cold, low-key lighting casting deep blue-gray shadows and icy cyan glow", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "2a88d69b58b11dd0" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "충격적인 엄마의 GPS 좌표 확인", "representative_moment": "사색이 된 수리영(C01)이 널브러진 살인 사건 사진들 너머로 모니터의 바다 GPS 좌표를 확인하며 동공이 흔들리는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level close frontal framing, shallow depth of field, icy cyan monitor light against blue-gray ambient shadows", "t2i_prompt": "Photorealistic cinematic still. The same figure, C01O01 in a gray casual hoodie, an East Asian young woman in her early 20s with long natural black hair, occupies the center two-thirds of the frame, seated and hunched over the same police office desk, facing the camera and gazing off-screen. Her face appears visibly pale and drained of color, maintaining a tense, shocked expression with widened eyes. The same room is bathed in cold, low-key lighting; an unseen monitor just past the camera axis casts an icy cyan spill that shapes her features, keeping its screen hidden from the camera. The desk surface recedes from the lower frame toward the character, holding a scattered pile of disturbing crime scene photographs spread out in the foreground. The printed faces of the photographs lie upward toward the camera but are out of focus via shallow depth of field, their disturbing contents reduced to pale shapes and muted red stains. A stationery pen rests on the same surface beside the blurred paperwork. The ambient illumination casts deep blue-gray shadows under her eyes and across the desk area against [L12B01: a dark, soft-focused office edge surface].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#21 action: '사색이 된 수리영, 필기도구를 찾아 좌표를 메모 한다.'", "S#21 action: '혜수 책상위에는 살인사건 확대 사진들이 어지럽게 널려있다.'", "shot description: 눈을 동그랗게 뜨고 모니터 화면을 응시하는 수리영의 정면." ], "visual_inferences": [ "fixed_element su_ri_young_pale_hunched_posture: seated and hunched over a police office desk, face appears visibly pale and drained of color", "fixed_element scattered_crime_scene_photos: sitting low in the foreground, appearing slightly out of focus and displaying muted red tones", "fixed_element police_office_blue_gray_lighting: cold, low-key lighting casting deep blue-gray shadows and icy cyan glows", "forward target requirement: desk surface with stationery pen must be visible for subsequent zoom-in" ], "creative_decisions": [ "Close framing strictly focuses on Suri-young's reaction while the disturbing crime scene photos remain an out-of-focus psychological anchor in the foreground.", "Using an unseen off-frame monitor for the cyan light logically explains her gaze and perfectly matches the DP's instruction to hide the monitor's content.", "Adhered to 'zoom_in_detail' continuity constraints by emphasizing 'the same figure', 'the same room', and 'the same surface'." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "ceiling", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "copier", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "filing cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "notice board", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "printer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d19c99fe750616bf", "owned_hash": "9f13f326b3d6be40", "camera_direction_hash": "b669ff03fafd4dc9", "owned_usage_hash": "ff3a4a9d25a43450", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No prior scene visual dependency applies globally outside of the fixed_elements inherited from Shot 5.", "_shot_index": 8, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At eye level just past the monitor axis, the camera holds a close frontal view of 수리영 across the desk; the monitor itself stays barely out of frame, but its cold glow shapes her widened eyes and pale face. She occupies the center two-thirds of the frame, shoulders slightly hunched toward the unseen screen, while blurred police paperwork and crime-scene photo edges sit low in the foreground without stealing focus.", "lighting_mood": "Low-key police-office fluorescents mix with icy monitor light, leaving muted blue-gray shadows under her eyes and a faint red alert accent in the blur.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L12B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "su_ri_young_pale_hunched_posture", "element_type": "character_state", "character_name": "수리영", "description": "A young East Asian female teenager is seated and hunched over a police office desk. Her face appears visibly pale and drained of color, maintaining a tense, shocked expression under the cold office lighting.", "applies_to_shots": [ 5, 8 ], "element_scope": "close", "source_facts": [ "사색이 된 수리영", "hunched over photos", "hunched toward monitor", "pallor and frozen breath", "pale face" ], "visual_inferences": [ "She is wearing casual teenage clothing suitable for a modern daytime setting." ], "creative_decisions": [ "Keeping her upper body consistently hunched forward emphasizes her intense focus and psychological weight from the disturbing discoveries." ], "confidence": "high" }, { "element_id": "scattered_crime_scene_photos", "element_type": "persistent_prop", "character_name": "", "description": "A scattered pile of disturbing crime scene photographs is spread out on the desk. These photos sit low in the foreground, appearing slightly out of focus and displaying muted red tones.", "applies_to_shots": [ 5, 8 ], "element_scope": "close", "source_facts": [ "혜수 책상위에는 살인사건 확대 사진들이 어지럽게 널려있다", "disturbing photographs blurred across the lower foreground", "muted red flecks from the photographs", "crime-scene photo edges sit low in the foreground" ], "visual_inferences": [ "The photos have standard white borders and contain gruesome visual details that are obscured by the shallow depth of field." ], "creative_decisions": [ "Blurring the photos in the foreground ensures they remain a constant, morbid presence without stealing focus from the character's reactions." ], "confidence": "high" }, { "element_id": "police_office_blue_gray_lighting", "element_type": "environment_state", "character_name": "", "description": "The police station office is bathed in cold, low-key lighting. The ambient illumination casts deep blue-gray shadows and icy cyan glows from nearby monitors across the desk area.", "applies_to_shots": [ 5, 8, 10 ], "element_scope": "full", "source_facts": [ "Flat daytime office light is cooled into corpse-like blue-gray", "Low-key police-office fluorescents mix with icy monitor light", "cyan monitor glow falls off into damp blue-gray shadow" ], "visual_inferences": [ "The light source is primarily artificial overhead fluorescents and computer screens, with minimal warm natural sunlight." ], "creative_decisions": [ "Using dominant blue-gray and cyan tones establishes a chilling, clinical atmosphere that mirrors the grim nature of the crime scene photos and the GPS location discovery." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 21, "shot_index": 5, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [ { "scene_index": 21, "shot_index": 10, "description": "접힌 종이쪽지를 주머니 안에 밀어 넣은 채 멈춘 수리영의 손.", "keep_elements": [ { "label": "desk surface with stationery and pen", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "cold, low-key lighting with deep blue-gray shadows and cyan glow", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "503853bc18cebc64" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "수리영의 단독 행동 결심", "representative_moment": "수리영이 좌표가 적힌 쪽지를 주머니에 숨기며 손을 멈춘 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close-up at waist height, shallow depth of field, icy cyan and blue-gray tones", "t2i_prompt": "Photorealistic cinematic still. At waist height beside the desk, the camera angles steeply downward to a close view of C01 in a blue-gray hooded jacket, a Korean woman in her early 20s. Her torso is angled at three-quarter from the right, occupying the lower-right half of the frame. Her hand is frozen at the jacket pocket, fingertips resting against the fabric after pushing a folded note inside. Only the blank folded outer face of the paper note is visible at the pocket opening, concealing the writing. The space is bathed in cold, low-key lighting, where an icy cyan monitor glow falls off into deep blue-gray shadow around her waist, with a small warm desk-lamp glint catching the paper edge. [L12B01: A scratched vertical desk side panel cuts obliquely along the left edge of the frame, with a pen and loose stationery blurred on the desktop behind the hand.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "메모쪽지를 접어 주머니에 넣고", "At waist height beside the desk, the camera stays within arm’s reach and angles down to a close view of 수리영’s hand frozen after pushing the folded note into her pocket" ], "visual_inferences": [ "fixed_element police_office_blue_gray_lighting dictates the cold cyan and blue-gray ambient shadows", "close framing at waist height explicitly places head and lower legs out of frame, focusing entirely on the hand and pocket", "desktop elements like the side panel and pen serve as the immediate frame-edge anchor in a close-up" ], "creative_decisions": [ "restricting the depth of field to sharply define the fingers and the blank note against the jacket fabric, blurring out the desk to emphasize the secret action", "using a single warm glint on the paper edge to contrast against the overarching clinical, cold cyan lighting" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "ceiling", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "copier", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "filing cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "notice board", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "printer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "a463a3a07453455c", "owned_hash": "9f13f326b3d6be40", "camera_direction_hash": "0098795679b684f3", "owned_usage_hash": "ff3a4a9d25a43450", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 21, "dependency_reason": "zoom_in_detail from Shot 8 of the same scene, preserving lighting and spatial continuity.", "_shot_index": 10, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 10 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At waist height beside the desk, the camera stays within arm’s reach and angles down to a close view of 수리영’s hand frozen after pushing the folded note into her pocket. The hand and pocket sit in the lower-right half of the frame, with the desktop edge and a pen blurred along the left, preserving the secrecy of the action before the camera withdraws.", "lighting_mood": "The cyan monitor glow falls off into damp blue-gray shadow around her waist, with a small warm desk-lamp glint catching the paper edge.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C01", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The shot isolates 수리영’s hand and pocket action rather than her face or full appearance, so a base identity reference is safer than requiring an outlook match." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L12B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "police_office_blue_gray_lighting", "element_type": "environment_state", "character_name": "", "description": "The police station office is bathed in cold, low-key lighting. The ambient illumination casts deep blue-gray shadows and icy cyan glows from nearby monitors across the desk area.", "applies_to_shots": [ 5, 8, 10 ], "element_scope": "full", "source_facts": [ "Flat daytime office light is cooled into corpse-like blue-gray", "Low-key police-office fluorescents mix with icy monitor light", "cyan monitor glow falls off into damp blue-gray shadow" ], "visual_inferences": [ "The light source is primarily artificial overhead fluorescents and computer screens, with minimal warm natural sunlight." ], "creative_decisions": [ "Using dominant blue-gray and cyan tones establishes a chilling, clinical atmosphere that mirrors the grim nature of the crime scene photos and the GPS location discovery." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 21, "shot_index": 8, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character", "id": "C01", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "520c492fd46186c7" }, { "scene_index": 22, "heading": "S#22. 경찰서 밖 / D", "beat_title": "자전거에 올라탄 수리영", "representative_moment": "경찰서 앞 젖은 아스팔트 위, 자전거 안장에 걸터앉아 출발 직전 긴장된 표정으로 앞을 주시하는 수리영(C01).", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide, static eye-level full-body composition, 35mm wide angle, overcast cool blue-gray tones", "t2i_prompt": "Photorealistic cinematic still. Shot from an eye-level standing height near the street-side curb, holding a wide full-body composition facing diagonally back across the wet pavement toward the police station entrance and bicycle rack. C01O01 in a casual hooded jacket, a Korean woman in her early 20s with long natural black hair and youthful oval face, positioned in the right third in left profile. She is perched on the saddle of a stationary parked bicycle, one foot planted firmly on the ground, gaze fixed off-screen in a tense pause before movement. The bicycle frame and wheels are clearly visible beneath her. Behind her, the building facade and recessed glass door stretch across the background. Overcast daylight flattens the scene into cool blue-gray tones. The slick pavement reflects faint vertical streaks of the station facade, the window, and the bicycle wheels. Small gold reflections from the sign panel barely warm the wet asphalt. A bollard and curb trace the edge of the driveway.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#22. 경찰서 밖 / D", "주차된 자전거를 타고 이내 달리는 수리영", "경찰서 건물 앞, 자전거 안장에 엉덩이를 댄 채 한 발로 땅을 짚고 서 있는 수리영의 전신.", "facing diagonally back across number 1 (front apron) toward number 6 (bicycle parking rack), number 8 (parked bicycle), and number 4 (main entrance doors)" ], "visual_inferences": [ "tense pause before movement", "overcast daylight with cool blue-gray tones", "gold reflections from station signage barely warming the wet asphalt" ], "creative_decisions": [ "positioned character in the right third to emphasize the empty driveway and building behind her", "used an eye-level 35mm composition to anchor both the character and the background architecture in a single frame" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bicycle rack", "usage_kind": "anchor", "source_phrase": "bicycle rack" }, { "owned_token": "bollard", "usage_kind": "anchor", "source_phrase": "A bollard" }, { "owned_token": "curb", "usage_kind": "anchor", "source_phrase": "street-side curb" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "recessed glass door" }, { "owned_token": "entrance", "usage_kind": "anchor", "source_phrase": "police station entrance" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "building facade" }, { "owned_token": "pavement", "usage_kind": "anchor", "source_phrase": "wet pavement" }, { "owned_token": "sidewalk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign panel", "usage_kind": "anchor", "source_phrase": "sign panel" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "the window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e27346e452746149", "owned_hash": "21a299687d1203fb", "camera_direction_hash": "0a868d6b8eb21ab0", "owned_usage_hash": "eefc1d51fb070419", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bicycle rack", "violating_phrase": "bicycle rack", "reason": "The bicycle rack is mentioned only as a background anchor to direct the camera's spatial orientation, matching the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "bollard", "violating_phrase": "A bollard", "reason": "The bollard is used as a static environmental marker tracing the driveway's edge, aligning with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "curb", "violating_phrase": "street-side curb", "reason": "The curb is referenced for camera height positioning and layout boundary description, with no instructions to redraw.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "recessed glass door", "reason": "The door is described as a static background element stretching behind the character, matching the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "entrance", "violating_phrase": "police station entrance", "reason": "The entrance is used as a directional landmark for the camera facing angle, serving as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "facade", "violating_phrase": "building facade", "reason": "The facade is mentioned as a backdrop stretching behind the character and in the reflections, consistent with anchor usage.", "verdict": "anchor_reference" }, { "owned_object": "pavement", "violating_phrase": "wet pavement", "reason": "The pavement is mentioned to establish the wet environmental context and reflections without requesting any modification.", "verdict": "anchor_reference" }, { "owned_object": "sign panel", "violating_phrase": "sign panel", "reason": "The sign panel is referenced solely for its faint gold reflections on the asphalt, acting as a lighting anchor.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "the window", "reason": "The window is mentioned only within the context of its reflection on the slick pavement, matching the anchor classification.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 1, "visible_entities": [ "C01", "L13" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 22, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From across the driveway, the camera holds a wide, static eye-level full-body composition facing the police station entrance. 수리영 is side-on near the right third, perched on the bicycle saddle with one foot planted on the wet ground, while the building frontage and empty driveway stretch behind her in a tense pause before movement.", "lighting_mood": "Overcast daylight is flattened into cool blue-gray tones, with small gold reflections from station signage barely warming the wet asphalt.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L13" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L13B01", "reference_usage": "exact_background", "owned_objects": [ "bicycle rack", "bollard", "curb", "door", "entrance", "facade", "pavement", "sidewalk", "sign panel", "window" ], "camera_reference": { "camera_position": "near number 3 (street-side exit and curb zone), facing diagonally back across number 1 (front apron) toward number 6 (bicycle parking rack), number 8 (parked bicycle), and number 4 (main entrance doors)", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "include the bicycle parking area and street-side exit path prominently; keep the station entrance readable in the background as a fixed location anchor" }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O01", "policy": "required" }, { "kind": "background", "id": "L13B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "e0a419b360fcf033" }, { "scene_index": 23, "heading": "S#23. 어느 PC방 / 실내", "beat_title": "좌표 확인", "representative_moment": "모니터 화면 속 붉은 핀이 꽂힌 해안가 지도가 클로즈업된 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "extremely tight insert close-up, cold blue-gray monitor spill, pixel grid texture", "t2i_prompt": "Photorealistic cinematic still. An extremely tight insert close-up of a glowing monitor screen displaying a digital coastal map, the lens positioned almost perpendicular to the surface. Focus is locked on a sharp red location pin planted near the center of the map. The visible screen face shows a fine RGB pixel grid structure and slight screen moiré, highlighting the electronic medium. The display emits a cold digital blue-gray spill, sharply contrasted by the vivid red of the pin. Only a narrow matte black rim of the monitor bezel is visible along one edge, while the surrounding physical room falls into deep, near-black ambient shadow.", "outfit_assignments": [], "source_facts": [ "shot description: close-up of coastal map image on the monitor screen pierced by a red pin", "scene text: checking coordinates and locations on the monitor" ], "visual_inferences": [ "fixed_element pc_bang_ambient_lighting: cold digital blue-gray glow, near-black shadow", "through_device perception: fine RGB pixel grid structure and slight screen moiré" ], "creative_decisions": [ "extremely tight insert framing to focus entirely on the digital map and red pin", "matte black bezel sliver included along one edge to physically anchor the image as a monitor screen" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "54033504a0e8536f", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "4a7bbd0960f28996", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 2, "visible_entities": [], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 23, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The lens is almost perpendicular to the monitor surface in an extremely tight screen close-up, cropping out nearly all of the PC room except a thin sliver of dark bezel. The red location pin and coastal map occupy the central frame, with pixel texture and slight screen moiré making the image feel mediated and urgent.", "lighting_mood": "The map’s cold digital blue-gray glow is punctured by the sharp red pin, with the surrounding room falling into near-black PC-bang shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "pc_bang_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "A dimly lit computer room cubicle. The space is enveloped in deep, near-black shadows, illuminated primarily by the cold, blue-gray spill of light from a glowing monitor. Subtle traces of red light reflect faintly off surfaces within the dark physical environment.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "near-black PC-bang shadow", "cold digital blue-gray glow", "PC booth darkness", "Cool monitor spill", "muted red UI reflection" ], "visual_inferences": [ "The overall space relies on high contrast, with the monitor serving as the primary light source cutting through the pitch-black surroundings." ], "creative_decisions": [ "Using a consistent cool blue-gray and subtle red color palette to unify the digital screen action and the physical backpack action." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [], "forbidden_refs": [], "readiness_policy": "not_applicable", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "bdb51ea92859902a" }, { "scene_index": 23, "heading": "S#23. 어느 PC방 / 실내", "beat_title": "종이 지도 확보와 짐싸기", "representative_moment": "수리영의 손이 접힌 종이 지도 여러 장을 열린 백팩 안으로 밀어 넣는다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low angle close-up, cool blue-gray monitor spill with deep shadows and muted red UI accents", "t2i_prompt": "Photorealistic cinematic still. Shot from low beside the open P12 backpack at about chair height, the camera angles down into the dark bag interior. C01 in a dark sleeve, a young Korean woman positioned at a three-quarter angle from the left looking off-screen, her hand entering from the upper left, paused mid-action pushing several folded P14 paper maps inside. The P12 backpack's unzipped, partly collapsed mouth occupies the lower half of the frame, facing upward and slightly toward the camera. The creased printed faces of the P14 maps are partly visible on the top folds, angled down into the backpack so only fragments of lines and coast shapes show. Cool blue-gray monitor spill and weak fluorescent overheads cast shadows inside the bag, while a muted red UI reflection flickers on the paper edges. [: The dimly lit underside of a desk leg and near-black PC booth shadows compress the frame edges], serving as scale references around the close action.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Shot 4 description: 수리영의 손이 접힌 종이 지도 여러 장을 열린 백팩 안으로 밀어 넣은 채 멈춘 찰나.", "Camera positioning: Low beside the open backpack, about chair height, the camera angles down into the bag.", "Lighting: Cool monitor spill and weak fluorescent overheads, muted red UI reflection." ], "visual_inferences": [ "fixed_element pc_bang_ambient_lighting: deep, near-black shadows and high contrast around the local area.", "Close framing rule dictates that the rest of C01's body is outside the frame; only the hand in a dark sleeve enters." ], "creative_decisions": [ "Replaced forbidden L01 ID with L02 to define the frame-edge background surface, fulfilling the close framing requirement to describe immediate surroundings without phantom background references.", "Emphasized the map's creased folds catching the red UI reflection to highlight it as the key narrative object while following the single perspective rule." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "prop" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "c44c905e8b0dff81", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "440df046f47876fa", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 4, "visible_entities": [ "C01", "P12", "P14" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 23, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Low beside the open backpack, about chair height, the camera angles down into the bag as 수리영’s hand finishes pushing several folded paper maps inside. The bag mouth takes the lower half of the frame, her hand enters from the upper left, and the desk leg and PC booth darkness remain as scale references around the action.", "lighting_mood": "Cool monitor spill and weak fluorescent overheads create blue-gray shadows inside the backpack, with a muted red UI reflection flickering on the folded map edges.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "P12", "P14" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C01", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "Only 수리영’s hand and packing action are visible, so full outlook-specific identity matching would not meaningfully apply." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": null, "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "pc_bang_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "A dimly lit computer room cubicle. The space is enveloped in deep, near-black shadows, illuminated primarily by the cold, blue-gray spill of light from a glowing monitor. Subtle traces of red light reflect faintly off surfaces within the dark physical environment.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "near-black PC-bang shadow", "cold digital blue-gray glow", "PC booth darkness", "Cool monitor spill", "muted red UI reflection" ], "visual_inferences": [ "The overall space relies on high contrast, with the monitor serving as the primary light source cutting through the pitch-black surroundings." ], "creative_decisions": [ "Using a consistent cool blue-gray and subtle red color palette to unify the digital screen action and the physical backpack action." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character", "id": "C01", "policy": "required" }, { "kind": "prop", "id": "P14", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [ { "contract_id": "rc_001", "scope": { "scene_index": 23, "shot_indices": [ 4 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P14", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] } ], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "a4f15a45107ca1f8" }, { "scene_index": 24, "heading": "S#24. 해안가 / 소나기", "beat_title": "빗속의 절박한 질주", "representative_moment": "쏟아지는 폭우 속에서 자전거 핸들을 꽉 쥐고 앞을 향해 필사적으로 달리는 C01의 젖은 얼굴.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Tracking backward just below eye level, tight frontal close shot, bruised blue-gray storm lighting", "t2i_prompt": "Photorealistic cinematic still. C01O02 in a navy waterproof windbreaker, a young Korean woman with youthful oval face, soft facial features, and long natural black hair flowing backward opposite to the forward motion, tracked from slightly below eye level, facing the camera and leaning forward. Her rain-soaked face fills the upper-center frame, eyes gazing into the distance with intense focus. Tightly clenched hands grip the wet bicycle handlebar cutting across the lower third of the frame, the upper stem of the bicycle frame partially visible below. Cold, bruised blue-gray storm light catches hard silver rain highlights and faint red glints from distant warning lights on her wet skin. Heavy wind-driven rain streaks slash through the foreground between the lens and the rider. [: rain-slick coastal road behind her reflecting the gray sky as elongated smears, collapsed into gray motion blur.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" } ], "source_facts": [ "S#24 action: '천둥 번개와 함께 잔뜩 낀 먹구름이 세찬 비를 뿌린다. 자전거를 타고 빗속을 달리는 수리영.'", "shot description: '쏟아지는 빗속에서 자전거 핸들을 꽉 쥐고 앞을 노려보는 수리영의 젖은 얼굴.'" ], "visual_inferences": [ "motion freeze: hair flowing backward opposite to motion, leaning forward over handlebars", "wet textures on skin, hair, and clothing due to the storm" ], "creative_decisions": [ "tight frontal close framing to emphasize desperation and speed", "cold bruised blue-gray palette with red glints to elevate the stormy atmosphere", "motion-blurred background to imply rapid forward movement without showing the full environment" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "b6503309f719d34b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "e17d7050f0ac4e22", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 2, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 24, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Tracking backward just ahead of the handlebars, the camera stays slightly below 수리영’s eye line in a tight frontal close shot that includes both her rain-soaked face and clenched hands. Her face fills the upper center while the handlebar cuts across the lower third, and rain streaks slash through the foreground as the road behind her collapses into gray motion blur.", "lighting_mood": "Storm light is cold and bruised blue-gray, with hard silver rain highlights and occasional red glints from distant warning lights cutting through the frame.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": null, "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O02", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "a3e4435d8a948d5f" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "벽에 그려진 붉은 원", "representative_moment": "수리영의 시선으로 열린 방문 틈을 통해 방 안쪽 벽에 선명하게 그려진 붉은 원(P08)을 발견하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "subjective POV eye-level framing through a doorway gap, sharp focus on the background mark, muted blue-gray dusk tones", "t2i_prompt": "Photorealistic cinematic still. A half-open bedroom doorframe stands in the middle-left foreground, revealing a stark P08 red circle mark painted on the bedroom wall in the middle-center background. The eye-level camera peers diagonally through the narrow door gap, focusing sharply on the P08 mark held at eye height, leaving the foreground door edge slightly soft. The surrounding room, including a partially visible bed and floor below, is bathed in muted blue-gray ambient dusk light. The damp, shadowed atmosphere flattens the interior plane, contrasting the heavy cold dusk tones against the vivid red of the ritual sign.", "outfit_assignments": [], "source_facts": [ "열린 자신의 방을 힐끗 보는데 벽에 또 붉은 원이 그려져있다.", "From just outside the half-open bedroom door at eye level, the camera peers diagonally through the narrow gap as if adopting 수리영’s discovered line of sight.", "Dusk seeps in as muted blue-gray ambient light, while the red circle reads as a sharp ritual accent against the damp, shadowed room." ], "visual_inferences": [ "fixed_element red_circle_wall_mark: A distinctly drawn red circle mark is painted on the bedroom wall. The surrounding environment is bathed in a muted blue-gray ambient light typical of low dusk.", "Since this is a subjective POV shot from 수리영's perspective, no characters are physically visible in the frame." ], "creative_decisions": [ "Used depth of field to keep the foreground doorframe slightly out of focus while holding the P08 mark on the back wall sharp, enhancing the feeling of spying through a gap.", "Emphasized the color contrast between the cool blue-gray ambient dusk light and the vivid red of the mark to amplify the ominous, occult horror tone." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background", "prop" ], "owned_object_usage": [ { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "partially visible bed" }, { "owned_token": "bedding", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "through the narrow door gap" }, { "owned_token": "doorframe", "usage_kind": "anchor", "source_phrase": "A half-open bedroom doorframe stands in the middle-left foreground" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "and floor below" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "painted on the bedroom wall in the middle-center background" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "6014ac0d92c65a28", "owned_hash": "162ce02cb64b6290", "camera_direction_hash": "4d64b221dd8cac5e", "owned_usage_hash": "b57794ca8c5632a4", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "doorframe", "violating_phrase": "A half-open bedroom doorframe stands in the middle-left foreground", "reason": "The doorframe is mentioned only to establish the foreground perspective and frame the shot, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "painted on the bedroom wall in the middle-center background", "reason": "The wall is referenced as the background surface that hosts the red circle mark, serving purely as a spatial anchor for the mark's location.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "through the narrow door gap", "reason": "The door is referenced only to describe the camera's line of sight through the gap, serving as an environmental anchor.", "verdict": "anchor_reference" }, { "owned_object": "bed", "violating_phrase": "partially visible bed", "reason": "The bed is described as partially visible and bathed in ambient dusk light, acting as a spatial anchor with no redraw instruction.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "floor below", "reason": "The floor is mentioned to specify the distribution of the ambient dusk light in the room, which is a clear anchoring usage.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "atmosphere_reference from S12_Shot16 to carry over the dim, cool gray ambient lighting and eerie mood.", "_shot_index": 4, "visible_entities": [ "L04", "P08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From just outside the half-open bedroom door at eye level, the camera peers diagonally through the narrow gap as if adopting 수리영’s discovered line of sight. The doorframe hangs large in the foreground on both sides, while the red circle mark on the far bedroom wall is held sharp in the middle distance before the camera slips closer.", "lighting_mood": "Dusk seeps in as muted blue-gray ambient light, while the red circle reads as a sharp ritual accent against the damp, shadowed room.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "half-open bedroom doorframe", "screen_zone": "middle_left", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "red circle mark on bedroom wall", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L04", "P08" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B02", "reference_usage": "exact_background", "owned_objects": [ "bed", "bedding", "curtain", "curtain track", "door", "doorframe", "floor", "shelf", "wall", "wardrobe", "window" ], "camera_reference": { "camera_position": "in number 2 (shared living dining zone) just outside number 10 (young bedroom door), facing through the doorway into number 5 (young adult bedroom) toward number 22 (young bed) and the wall near number 24 (curtain).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "40mm normal lens from doorway", "framing_notes": "Frame number 10 as a doorway foreground and number 40 as the key overlay on the bedroom wall; keep numbers 22 and 24 visible for orientation." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "red_circle_wall_mark", "element_type": "environment_state", "character_name": "", "description": "A distinctly drawn red circle mark is painted on the bedroom wall. The surrounding environment is bathed in a muted blue-gray ambient light typical of low dusk.", "applies_to_shots": [ 4, 8 ], "element_scope": "full", "source_facts": [ "벽에 또 붉은 원이 그려져있다", "the red circle mark on the far bedroom wall is held sharp", "the marked wall behind her dissolves into a red blur", "Dusk seeps in as muted blue-gray ambient light", "Low dusk light and handheld shadow wrap her face in blue-gray" ], "visual_inferences": [ "The red circle is large and prominent enough to be clearly visible both from outside the doorway and as a background element behind the character.", "The mark appears painted or drawn directly onto the interior surface." ], "creative_decisions": [ "The stark color contrast between the cold blue-gray dusk lighting and the deep red of the wall mark creates an ominous, unsettling atmosphere across the sequence." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 16, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [ { "scene_index": 25, "shot_index": 8, "description": "사진을 든 채 충격으로 멍해진 수리영의 측면 얼굴.", "keep_elements": [ { "label": "the bedroom wall with the red circle mark", "kind": "environment", "subject_kind": "non_human_visual_element" }, { "label": "the muted blue-gray ambient lighting in the room", "kind": "environment", "subject_kind": "non_human_visual_element" } ] } ], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B02", "policy": "required" }, { "kind": "prop", "id": "P08", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [ { "contract_id": "rc_001", "scope": { "scene_index": 25, "shot_indices": [ 4 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P08", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] } ], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "fbd81b16abdcdac4" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "붉은 원 한가운데 꽂힌 사진", "representative_moment": "충격으로 멍해진 수리영(C01)이 피 묻은 옛날 사진을 손에 쥐고 있는 측면 얼굴.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Shoulder-height tight profile from a slightly lower eye line, shallow depth of field, muted blue-gray dusk lighting.", "t2i_prompt": "Photorealistic cinematic still. C01O02 in a navy waterproof windbreaker, a Korean woman in her early 20s with long natural black hair and soft facial features, appears as the same figure in right profile, occupying the right half of the frame. Viewed from shoulder height from a slightly lower eye line, her stunned face is looking down. The old photograph she clutches low intrudes at the bottom edge, its printed face angled partly toward the camera but cropped and soft, showing only faded shapes and dark stains catching a dull gold edge of light. The same room is bathed in a muted blue-gray ambient light typical of low dusk, with low dusk light and handheld shadows wrapping her face in blue-gray. The close frame-edge surface behind her features [: the same bedroom wall], where a distinctly drawn red circle mark painted on the wall dissolves into a red blur, out of focus and partially blocked by her head. A messy bedroom furniture corner recedes behind her shoulder at a diagonal, barely visible in the shallow depth of field.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" } ], "source_facts": [ "the marked wall behind her dissolves into a red blur", "Dusk seeps in as muted blue-gray ambient light", "Low dusk light and handheld shadow wrap her face in blue-gray" ], "visual_inferences": [ "fixed_element body_pose: right profile, looking down at the photograph", "The red circle is large and prominent enough to be partially blocked by her head", "Shallow focus blurring the background elements like the furniture corner and the wall mark" ], "creative_decisions": [ "Tight close framing on the right side to emphasize her stunned expression while preserving the evidence in the lower edge", "Dull gold light catching the old photo edge to draw attention against the cool blue-gray ambient tone" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "threshold", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "9a6877d9dd62b714", "owned_hash": "fb29af5ef1667bdd", "camera_direction_hash": "51583b3e1aeb4aa1", "owned_usage_hash": "33b79c756ae788fd", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 25, "dependency_reason": "같은 공간(옥탑방 안)에서 발생한 직전 샷(S25_Shot4)의 배경, 조명, 환경 디테일(붉은 원)을 유지하며 카메라만 피사체로 확대·이동(zoom_in_detail).", "_shot_index": 8, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At shoulder height beside 수리영, the handheld camera stays tight on her side profile from a slightly lower eye line, letting the photograph in her hand intrude at the bottom edge. Her stunned face occupies the right half of the frame while the marked wall behind her dissolves into a red blur, making the shock register before the evidence does.", "lighting_mood": "Low dusk light and handheld shadow wrap her face in blue-gray, with the red wall mark bleeding softly behind her and a dull gold edge catching the old photo.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L04B03", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "red_circle_wall_mark", "element_type": "environment_state", "character_name": "", "description": "A distinctly drawn red circle mark is painted on the bedroom wall. The surrounding environment is bathed in a muted blue-gray ambient light typical of low dusk.", "applies_to_shots": [ 4, 8 ], "element_scope": "full", "source_facts": [ "벽에 또 붉은 원이 그려져있다", "the red circle mark on the far bedroom wall is held sharp", "the marked wall behind her dissolves into a red blur", "Dusk seeps in as muted blue-gray ambient light", "Low dusk light and handheld shadow wrap her face in blue-gray" ], "visual_inferences": [ "The red circle is large and prominent enough to be clearly visible both from outside the doorway and as a background element behind the character.", "The mark appears painted or drawn directly onto the interior surface." ], "creative_decisions": [ "The stark color contrast between the cold blue-gray dusk lighting and the deep red of the wall mark creates an ominous, unsettling atmosphere across the sequence." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 25, "shot_index": 4, "ref_usage": "zoom_in_detail", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O02", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "a6bd9d05fa992ada" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "안개 낀 포구와 수리영의 기다림", "representative_moment": "안개가 짙게 깔린 포구의 어두운 갯바위 위, 무릎을 안고 바다를 응시하는 수리영(C01)의 쓸쓸한 뒷모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide rear-side silhouette from knee height, cold blue-gray dusk lighting", "t2i_prompt": "Photorealistic cinematic still. Shot from knee height, deviating from the eye-level background reference framing to keep the figure small against the vast fog. C01 in a dark navy waterproof windbreaker, a Korean woman in her early 20s, with back to camera, face fully swallowed in shadow with no discernible features, seated on the dark shoreline rock in the lower-left third. Her knees are hugged close to her chest, gazing into the distance across the harbor water toward the open channel. The environment retains a cold, low-key blue-gray tonal palette during the post-rain dusk. Thick, heavy fog engulfs the harbor, significantly reducing visibility. The water reflects a smeared gray sky and the dark underside of the mist, with no clear boats depicted. Faint cold highlights catch on the wet rock, while the concrete quay and walkway remain visible along the side of the frame, fading into the dense sea fog.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" } ], "source_facts": [ "비는 그치고 안개가 자욱하다. 텅 빈 포구.", "갯바위에 앉아있는 수리영.", "하염없이 안개가 몰려오는 바다를 바라보는데" ], "visual_inferences": [ "fixed_element harbor_dusk_fog: thick fog reducing visibility, wet surfaces reflecting dim ambient light", "camera lowered to knee height emphasizing small silhouette against vast water, modifying the eye-level background reference" ], "creative_decisions": [ "explicitly note camera height deviation in prompt for compositor handling", "low-key blue-gray color grading to match DP's cold dusk instruction", "emphasize the reflection on the water to anchor the atmospheric depth" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "channel", "usage_kind": "anchor", "source_phrase": "toward the open channel" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ladder", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "path", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "quay", "usage_kind": "anchor", "source_phrase": "the concrete quay" }, { "owned_token": "ring", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rock", "usage_kind": "anchor", "source_phrase": "seated on the dark shoreline rock" }, { "owned_token": "walkway", "usage_kind": "anchor", "source_phrase": "and walkway remain visible" }, { "owned_token": "water", "usage_kind": "anchor", "source_phrase": "across the harbor water" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "b888bc7a93334cff", "owned_hash": "caf968a9d29ff12b", "camera_direction_hash": "a08309055c0a18b4", "owned_usage_hash": "083ff1e1441490fe", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "channel", "violating_phrase": "toward the open channel", "reason": "The prompt mentions the channel only to establish the direction of the character's gaze, acting as a spatial anchor. This aligns with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "quay", "violating_phrase": "the concrete quay", "reason": "The concrete quay is mentioned as remaining visible within the frame, which serves as a background spatial anchor without requiring it to be redrawn.", "verdict": "anchor_reference" }, { "owned_object": "rock", "violating_phrase": "seated on the dark shoreline rock", "reason": "The rock is referred to as the surface the character sits on, with faint highlights on it, serving as a spatial and lighting anchor rather than a new object to be drawn.", "verdict": "anchor_reference" }, { "owned_object": "walkway", "violating_phrase": "and walkway remain visible", "reason": "The walkway is described as remaining visible fading into the fog, establishing its role as a background environmental anchor.", "verdict": "anchor_reference" }, { "owned_object": "water", "violating_phrase": "across the harbor water", "reason": "The harbor water is used as a spatial reference for gazing and its reflection of the sky is described, which serves as an environmental anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 2, "visible_entities": [ "C01", "L16" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From knee height on the empty harbor edge, the camera has drifted in to a wide rear-side silhouette of 수리영 seated on the dark rock, keeping her small against the fog-covered sea. Her knees are gathered close to her chest in the lower-left third, while the water and mist occupy most of the frame, swallowing facial detail and making the harbor feel abandoned.", "lighting_mood": "Post-rain dusk turns the harbor into low-key blue-gray silhouettes, with only faint cold highlights on the wet rock and no comforting warm source.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L16" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C01", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "수리영 is intentionally minimized as a dark silhouette with facial detail suppressed, so full outlook matching is less important than preserving her base identity and posture." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L16B02", "reference_usage": "exact_background", "owned_objects": [ "bollard", "channel", "curb", "ladder", "path", "quay", "ring", "rock", "walkway", "water" ], "camera_reference": { "camera_position": "near number 1 (tidal rock outcrop), facing east across number 3 (harbor water basin) toward number 6 (open water approach channel), with number 2 (concrete quay walkway) and number 4 (mooring edge) along the side of frame", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "empty harbor state; include the open water and rocks prominently, with no boat or mooring line depicted" }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "harbor_dusk_fog", "element_type": "environment_state", "character_name": "", "description": "The harbor and the surrounding sea are engulfed in thick, heavy fog during dusk, significantly reducing visibility. The environment retains a cold, blue-gray tonal palette, with wet surfaces like rocks and wooden structures subtly reflecting the dim ambient light after a recent heavy rain.", "applies_to_shots": [ 2, 8, 12 ], "element_scope": "full", "source_facts": [ "비는 그치고 안개가 자욱하다. 텅 빈 포구.", "Post-rain dusk turns the harbor into low-key blue-gray silhouettes", "Cold fog-diffused dusk keeps their faces in blue-gray half-shadow" ], "visual_inferences": [ "Surfaces are wet and slightly reflective from the recent downpour", "The background lacks distinct horizon lines or distant objects due to the dense mist" ], "creative_decisions": [ "Dominant blue-gray color grading is applied consistently to emphasize the isolation and cold atmosphere of the abandoned harbor" ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character", "id": "C01", "policy": "required" }, { "kind": "background", "id": "L16B02", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "717fbf8fddcbe555" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "수리영의 절박한 간청과 인우의 수락", "representative_moment": "안개 낀 포구에서 C01이 배에 탄 인우(보통명사)를 올려다보며 태워달라고 간청하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium upper-body view from slightly below eye level, cold blue-gray fog-diffused dusk lighting", "t2i_prompt": "Photorealistic cinematic still. C01O02 in a navy windbreaker, a Korean young woman, stands in the lower-center foreground, looking toward the man on the boat across the boat gunwale between them in the middle-right midground. Seen from slightly below her eye line at a three-quarter angle from the left, she is framed in a medium upper-body view, her clasped hands raised near the lower frame edge. Her soft facial features carry a pleading expression, while her long natural black hair and skin are visibly damp, strands clinging slightly to her neck with a faint sheen from the heavy mist. A Korean young man in fisher workwear occupies the upper frame, separated from her by the weathered boat gunwale crossing the air between them. Captured over his shoulder, he is bracing on the gunwale as a partial figure, looking down with his eyes fixed on her, establishing the height difference between the dock and the boat. The cold fog-diffused dusk casts their faces in blue-gray half-shadow, accented by small gold glints on a wet mooring rope and a muted red-brown rust line along the boat edge. The wet concrete quay walkway below grounds her lower position, reflecting the faint gray sky and blurred shapes of her hands. Behind them, thick heavy fog engulfs the harbor water, swallowing the far shoreline and dimming the silhouette of a fixed mooring bollard nearby.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" } ], "source_facts": [ "배에서 내린 인우(24세. 남)는 닷줄을 고정시키고 있다.", "At dock level beside the boat, the camera holds a tight upper-body view of 수리영 from slightly below her eye line, angled upward", "Her clasped hands sit at the lower frame edge, her pleading face occupies the center", "The dock boards reflect faint gray sky and blurred shapes of her hands near the lower frame." ], "visual_inferences": [ "fixed_element harbor_dusk_fog: thick fog engulfs the harbor water and swallows the far shoreline", "fixed_element suri_young_damp_appearance: her long natural black hair and skin are visibly damp with a faint sheen", "The height difference emphasizes her pleading state from the dock walkway up to the gunwale" ], "creative_decisions": [ "Applied fsc_001 and fsc_002 directly into the opening sentence to lock the spatial contract between the pleading girl and the man", "Used generic demographic descriptors for the unregistered male character to comply with ID policy while establishing the fg/bg interaction across the gunwale", "Highlighted the muted red-brown rust and small gold glints to contrast against the cold blue-gray dusk palette" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bollard", "usage_kind": "anchor", "source_phrase": "dimming the silhouette of a fixed mooring bollard nearby" }, { "owned_token": "channel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "edge strip", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ladder", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mooring ring", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "path", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "quay", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rock", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "seawall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "walkway", "usage_kind": "anchor", "source_phrase": "The wet concrete quay walkway below grounds her lower position" }, { "owned_token": "water", "usage_kind": "anchor", "source_phrase": "thick heavy fog engulfs the harbor water" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "c010965d813ca447", "owned_hash": "705af16646150144", "camera_direction_hash": "e6031fb2d661c71e", "owned_usage_hash": "8b7b3cc3ef6fb59f", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bollard", "violating_phrase": "dimming the silhouette of a fixed mooring bollard nearby", "reason": "The prompt mentions the bollard only as a background element obscured by fog, which is a passive anchor reference, consistent with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "quay", "violating_phrase": "The wet concrete quay walkway below", "reason": "The prompt references the quay as part of the walkway to ground the character's position and describe reflections, which is an anchor reference even though the producer declared it absent.", "verdict": "anchor_reference" }, { "owned_object": "walkway", "violating_phrase": "The wet concrete quay walkway below grounds her lower position", "reason": "The prompt uses the walkway as a spatial anchor to establish height and show reflections, aligning with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "water", "violating_phrase": "thick heavy fog engulfs the harbor water", "reason": "The harbor water is described as a background environmental element covered in fog, serving as an anchor reference as declared by the producer.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 26, "dependency_reason": "previous_shot_refs 의 atmosphere_reference (같은 씬의 앞선 샷에서 안개 낀 항구의 톤을 가져옴)", "_shot_index": 8, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At dock level beside the boat, the camera holds a tight upper-body view of 수리영 from slightly below her eye line, angled upward so her raised gaze leads to 인우 above the boat edge. Her clasped hands sit at the lower frame edge, her pleading face occupies the center, and 인우 remains higher and off to one side as a partial figure near the gunwale, preserving the dock-to-boat power difference.", "lighting_mood": "Cold fog-diffused dusk keeps their faces in blue-gray half-shadow, with small gold glints on wet rope and a muted red-brown rust line along the boat edge.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "shared_space_relation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "boat gunwale between them", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "looks_toward", "gesture_target_label": "man on the boat", "label": "수리영 pleading on dock", "screen_zone": "lower_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L16B01", "reference_usage": "exact_background", "owned_objects": [ "bollard", "channel", "curb", "edge strip", "ladder", "mooring ring", "path", "quay", "rock", "seawall", "walkway", "water" ], "camera_reference": { "camera_position": "near number 2 (concrete quay walkway) beside number 7 (fixed mooring bollard), facing diagonally toward number 11 (arrived small fishing boat) and number 12 (loose mooring line) over number 3 (harbor water basin)", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm wide angle", "framing_notes": "show the mooring action area clearly; keep number 4 and number 7 readable as the fixed landing points for the boat" }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "harbor_dusk_fog", "element_type": "environment_state", "character_name": "", "description": "The harbor and the surrounding sea are engulfed in thick, heavy fog during dusk, significantly reducing visibility. The environment retains a cold, blue-gray tonal palette, with wet surfaces like rocks and wooden structures subtly reflecting the dim ambient light after a recent heavy rain.", "applies_to_shots": [ 2, 8, 12 ], "element_scope": "full", "source_facts": [ "비는 그치고 안개가 자욱하다. 텅 빈 포구.", "Post-rain dusk turns the harbor into low-key blue-gray silhouettes", "Cold fog-diffused dusk keeps their faces in blue-gray half-shadow" ], "visual_inferences": [ "Surfaces are wet and slightly reflective from the recent downpour", "The background lacks distinct horizon lines or distant objects due to the dense mist" ], "creative_decisions": [ "Dominant blue-gray color grading is applied consistently to emphasize the isolation and cold atmosphere of the abandoned harbor" ], "confidence": "high" }, { "element_id": "suri_young_damp_appearance", "element_type": "character_state", "character_name": "수리영", "description": "The young woman's hair and skin are visibly damp, carrying lingering moisture from the recent rainstorm and the surrounding heavy mist. Her facial features are shaded by the cold blue-gray dusk lighting, emphasizing her chilled physical state in the harsh weather.", "applies_to_shots": [ 8, 12 ], "element_scope": "close", "source_facts": [ "소나기가 한차례 퍼 붓는가 싶더니", "damp hair texture and mist beads catching the edge light", "Muted dusk blue-gray shadows wrap her face" ], "visual_inferences": [ "Strands of wet hair cling slightly to her face and neck", "Her skin has a faint sheen due to the moisture from the sea fog" ], "creative_decisions": [ "Using the damp texture to visually connect her physical state to the harsh weather conditions she endured while waiting at the harbor" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 26, "shot_index": 2, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O02", "policy": "required" }, { "kind": "background", "id": "L16B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "4ee0b902e28c0edf" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "바다를 향한 수리영의 옅은 미소", "representative_moment": "수리영(C01)이 안개 낀 바다 너머를 응시하며 옅은 미소를 짓는 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Tight eye-level close-up, muted blue-gray dusk lighting with a faint cold gold rim", "t2i_prompt": "Photorealistic cinematic still. C01O02 in a navy waterproof windbreaker, a Korean woman in her early 20s, occupies the center-right of the frame in a tight eye-level close-up. Captured at a three-quarter angle from the left with her chin slightly lifted, she smiles faintly while gazing into the distance. Her damp hair and skin carry lingering moisture, with wet strands clinging to her face. Muted dusk blue-gray shadows wrap her features, accented by a faint cold gold rim light from the dying sky and wet highlights catching mist beads on her skin. Behind her profile, the frame edge reveals only [: a blurred blue-gray band of dense fog].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" } ], "source_facts": [ "수리영, 떠나기전, 바다를 바라보며 음성메시지를 남긴다.", "비는 그치고 안개가 자욱하다.", "안개 낀 바다 너머를 바라보며 옅게 미소 짓는 수리영의 얼굴 클로즈업." ], "visual_inferences": [ "fixed_element suri_young_damp_appearance: Damp hair and skin carrying lingering moisture from recent rain.", "fixed_element harbor_dusk_fog: Muted dusk blue-gray shadows wrap her face with a faint cold gold rim from the dying sky.", "Background is intentionally skipped due to close framing, leaving only a blurred band of fog at the frame edge." ], "creative_decisions": [ "Positioned her face in the center-right to leave visual space for her off-screen gaze toward the left.", "Emphasized wet highlights and mist beads to visually connect her physical state to the harsh weather conditions." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "channel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "edge strip", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ladder", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mooring ring", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "path", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "quay", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rock", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "seawall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "walkway", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "43f49bdbbf65c5cc", "owned_hash": "705af16646150144", "camera_direction_hash": "5dec281d654c50a5", "owned_usage_hash": "36e0587ce23f45f4", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No dependent scene specific to this precise moment", "_shot_index": 12, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 12 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the flow’s settled endpoint, the camera holds very close at face height about 40 cm from 수리영, just off her front in a tight close-up; her face occupies the center-right of frame while the fog-softened sea direction remains only as a blurred blue-gray band beyond her gaze. Eye-level, still, and intimate, the composition lets her faint smile emerge without cutting away, with damp hair texture and mist beads catching the edge light.", "lighting_mood": "Muted dusk blue-gray shadows wrap her face, with a faint cold gold rim from the dying sky and wet highlights on skin and hair.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L16B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "harbor_dusk_fog", "element_type": "environment_state", "character_name": "", "description": "The harbor and the surrounding sea are engulfed in thick, heavy fog during dusk, significantly reducing visibility. The environment retains a cold, blue-gray tonal palette, with wet surfaces like rocks and wooden structures subtly reflecting the dim ambient light after a recent heavy rain.", "applies_to_shots": [ 2, 8, 12 ], "element_scope": "full", "source_facts": [ "비는 그치고 안개가 자욱하다. 텅 빈 포구.", "Post-rain dusk turns the harbor into low-key blue-gray silhouettes", "Cold fog-diffused dusk keeps their faces in blue-gray half-shadow" ], "visual_inferences": [ "Surfaces are wet and slightly reflective from the recent downpour", "The background lacks distinct horizon lines or distant objects due to the dense mist" ], "creative_decisions": [ "Dominant blue-gray color grading is applied consistently to emphasize the isolation and cold atmosphere of the abandoned harbor" ], "confidence": "high" }, { "element_id": "suri_young_damp_appearance", "element_type": "character_state", "character_name": "수리영", "description": "The young woman's hair and skin are visibly damp, carrying lingering moisture from the recent rainstorm and the surrounding heavy mist. Her facial features are shaded by the cold blue-gray dusk lighting, emphasizing her chilled physical state in the harsh weather.", "applies_to_shots": [ 8, 12 ], "element_scope": "close", "source_facts": [ "소나기가 한차례 퍼 붓는가 싶더니", "damp hair texture and mist beads catching the edge light", "Muted dusk blue-gray shadows wrap her face" ], "visual_inferences": [ "Strands of wet hair cling slightly to her face and neck", "Her skin has a faint sheen due to the moisture from the sea fog" ], "creative_decisions": [ "Using the damp texture to visually connect her physical state to the harsh weather conditions she endured while waiting at the harbor" ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O02", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "0b1695a56b627f79" }, { "scene_index": 27, "heading": "S#27. 옥탑방 안 / 해질 무렵", "beat_title": "불안한 통화와 황량한 흔적", "representative_moment": "난장판이 된 거실 한가운데 서서 C02가 P05를 귀에 댄 채 불안한 눈빛으로 허공을 응시한다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-height observer angle, low-key dirty blue-gray dusk lighting", "t2i_prompt": "Photorealistic cinematic still. Positioned at chest height about six feet in front of the subject, slightly below the eye line. C02O03 in a dark gray field jacket, a Korean woman in her early 40s with neat dark hair and a composed adult face, stands at three-quarter angle from the right, occupying the center of the frame with her tight upper body visible. P05 is pressed against her ear in the upper-right of her silhouette, its sleek dark casing facing the camera, her widened eyes looking off-screen with sharp alarm. Low-key ashy dusk lighting fills the space in muted, dirty blue-gray shadows, casting a faint cold glint on P05 and her eyes. Behind her, the interior of the modest Korean rooftop living room is in a state of severe disarray, crowding both edges with scattered debris and overturned domestic objects. Dim, worn wall planes with moisture stains angle away to form a cramped diagonal corner in the background, where a dark window and a torn curtain remain anchored in the cold, desolate atmosphere.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O03" } ], "source_facts": [ "S#27 action: 수리영의 녹음메시지를 확인하는 혜수, 불안해하며 통화", "shot description: 난장판이 된 거실 한가운데, 핸드폰을 귀에 댄 채 눈이 커진 혜수의 상반신." ], "visual_inferences": [ "fixed_element rooftop_room_wreckage: severe disarray, heavily littered with scattered debris, ashy dusk lighting with dirty blue-gray shadows", "P05 held to the ear at three-quarter angle, reflecting faint cold glints in the dark room" ], "creative_decisions": [ "Medium framing at chest height to emphasize her upper body and the tense phone call while allowing the trashed room to crowd the frame edges.", "Muted blue-gray tones to enforce the desolate, cold mood established by the atmospheric reference shot." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "a dark window and a torn curtain remain anchored" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "threshold", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "a dark window and a torn curtain remain anchored" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e00df2074dec21ab", "owned_hash": "fb29af5ef1667bdd", "camera_direction_hash": "712214259274d272", "owned_usage_hash": "ad49c289b4e66e03", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "curtain", "violating_phrase": "a dark window and a torn curtain remain anchored", "reason": "The curtain is described as remaining anchored in the background, serving as a spatial reference rather than being redrawn, which aligns with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "a dark window and a torn curtain remain anchored", "reason": "The window is mentioned as remaining anchored in the background corner to establish the environment, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 25, "dependency_reason": "atmosphere_reference: S25_Shot8 (해질 무렵의 푸르스름하고 황량한 분위기와 색감 유지)", "_shot_index": 1, "visible_entities": [ "C02", "L04", "P05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 27, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At chest height about six feet in front of 혜수, slightly below her eye line, the camera frames her as a tight upper-body subject in the center while the trashed rooftop living room crowds both edges with overturned objects and torn domestic clutter. The phone is pressed to her ear in the upper-right of her silhouette, her widened eyes held sharp as if the room itself is closing around her.", "lighting_mood": "Low-key dusk light seeps through the room in dirty blue-gray tones, with a faint cold glint on the phone and 혜수's widened eyes.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02", "L04", "P05" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O03" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B03", "reference_usage": "exact_background", "owned_objects": [ "TV", "bed", "chair", "counter", "curtain", "curtain track", "door", "faucet", "floor lamp", "shelf", "sink", "stand", "table", "threshold", "wardrobe", "window" ], "camera_reference": { "camera_position": "near number 7 (metal entrance door), facing inward across number 1 (entrance threshold zone) and number 2 (shared living dining zone) toward number 8 (living window) and number 10 (young bedroom door).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm wide angle for full cramped-room coverage", "framing_notes": "Include number 7 in the foreground and number 8 across the room; overlay use includes numbers 41, 42, 43, and 44 for the ransacked and breached-room state." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "rooftop_room_wreckage", "element_type": "environment_state", "character_name": "", "description": "The interior of the modest Korean rooftop living room is in a state of severe disarray. The floor and surrounding areas are heavily littered with scattered debris, overturned domestic objects, and torn clutter, creating a chaotic and ruined atmosphere. The room is enveloped in low-key, ashy dusk lighting, characterized by muted, dirty blue-gray shadows that cast a cold, desolate tone over the wreckage.", "applies_to_shots": [ 1, 5 ], "element_scope": "full", "source_facts": [ "trashed rooftop living room crowds both edges with overturned objects and torn domestic clutter", "Low-key dusk light seeps through the room in dirty blue-gray tones", "emptied, wrecked space... scattered debris", "Ashy dusk light and muted blue-gray shadows" ], "visual_inferences": [ "The domestic items scattered around suggest an everyday living space that has been abruptly and violently disturbed." ], "creative_decisions": [ "Consolidated the lighting descriptions to ensure the exact 'dirty blue-gray' color palette and low-key contrast remain perfectly consistent across different framing sizes." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 25, "shot_index": 8, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C02O03", "policy": "required" }, { "kind": "background", "id": "L04B03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "cf144b1179094626" }, { "scene_index": 27, "heading": "S#27. 옥탑방 안 / 해질 무렵", "beat_title": "황량하게 흔들리는 현관문", "representative_moment": "어스름한 빛 속, 반쯤 열린 현관 철문이 빈 옥탑방 거실 너머로 삐걱이는 정지 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide diagonal view from the rear corner at standing eye level, low-key ashy dusk lighting", "t2i_prompt": "Photorealistic cinematic still. A wide diagonal view from the rear corner of the living room at standing eye level, looking across an emptied, wrecked space toward the existing door in the upper-middle distance. No characters are present. The barren floor is heavily littered with scattered debris, overturned domestic objects, and torn clutter, creating a hollow negative space in the foreground and midground. The view is oriented toward the existing door and the threshold leading to the outside stairwell. The adjacent doorframe is positioned diagonally. Enveloped in low-key, ashy dusk lighting, the room is flattened by muted, dirty blue-gray shadows, with a thin cold highlight tracing the edge of the existing door and the inner jamb.", "outfit_assignments": [], "source_facts": [ "S#27 action: '덜커덩 덜커덩. 바람에 현관문 소리. 황량하다.'", "shot description: 어스름한 빛 속, 바람에 반쯤 열린 현관 철문이 삐걱이며 멈춰 있는 황량한 거실 정지 구도." ], "visual_inferences": [ "fixed_element rooftop_room_wreckage: interior of the modest Korean rooftop living room is in a state of severe disarray", "The floor is heavily littered with scattered debris, overturned domestic objects, creating a ruined atmosphere", "No figures present to emphasize the hollow negative space and abandonment after the conversation" ], "creative_decisions": [ "Wide diagonal view from the rear corner to emphasize the empty distance to the half-open door", "Ashy dusk light and muted blue-gray shadows with a thin cold highlight on the door edge to amplify the desolate mood" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain track", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "looking across an emptied, wrecked space toward the existing door" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "threshold", "usage_kind": "anchor", "source_phrase": "the threshold leading to the outside stairwell" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e8f9651c1725eedf", "owned_hash": "fb29af5ef1667bdd", "camera_direction_hash": "0042dab348514ef8", "owned_usage_hash": "8994b3c726a0e644", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "looking across an emptied, wrecked space toward the existing door", "reason": "The prompt refers to the 'existing door' as a spatial anchor and lighting reference without requesting any modification or new creation, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "threshold", "violating_phrase": "the threshold leading to the outside stairwell", "reason": "The prompt mentions the 'threshold' purely as a directional cue for camera orientation, which matches the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 27, "dependency_reason": "exact_background from S27_Shot1 (same room, different moment)", "_shot_index": 5, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 27, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the rear corner of the living room at standing eye level, the camera holds a wide diagonal view across the emptied, wrecked space toward the half-open metal entrance door in the upper-middle distance. No character enters; the door occupies less than a third of the frame, while the barren floor, scattered debris, and dim walls create a hollow negative space that makes the creak feel accusatory.", "lighting_mood": "Ashy dusk light and muted blue-gray shadows flatten the room, with a thin cold highlight trembling along the metal door edge.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L04" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L04B03", "reference_usage": "exact_background", "owned_objects": [ "TV", "bed", "chair", "counter", "curtain", "curtain track", "door", "faucet", "floor lamp", "shelf", "sink", "stand", "table", "threshold", "wardrobe", "window" ], "camera_reference": { "camera_position": "near number 7 (metal entrance door), facing inward across number 1 (entrance threshold zone) and number 2 (shared living dining zone) toward number 8 (living window) and number 10 (young bedroom door).", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm wide angle for full cramped-room coverage", "framing_notes": "Include number 7 in the foreground and number 8 across the room; overlay use includes numbers 41, 42, 43, and 44 for the ransacked and breached-room state." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "rooftop_room_wreckage", "element_type": "environment_state", "character_name": "", "description": "The interior of the modest Korean rooftop living room is in a state of severe disarray. The floor and surrounding areas are heavily littered with scattered debris, overturned domestic objects, and torn clutter, creating a chaotic and ruined atmosphere. The room is enveloped in low-key, ashy dusk lighting, characterized by muted, dirty blue-gray shadows that cast a cold, desolate tone over the wreckage.", "applies_to_shots": [ 1, 5 ], "element_scope": "full", "source_facts": [ "trashed rooftop living room crowds both edges with overturned objects and torn domestic clutter", "Low-key dusk light seeps through the room in dirty blue-gray tones", "emptied, wrecked space... scattered debris", "Ashy dusk light and muted blue-gray shadows" ], "visual_inferences": [ "The domestic items scattered around suggest an everyday living space that has been abruptly and violently disturbed." ], "creative_decisions": [ "Consolidated the lighting descriptions to ensure the exact 'dirty blue-gray' color palette and low-key contrast remain perfectly consistent across different framing sizes." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 27, "shot_index": 1, "ref_usage": "exact_background", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L04B03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "0c345771157d3403" }, { "scene_index": 28, "heading": "S#28. 바다 + 갑판 / 해질 무렵", "beat_title": "사진을 확인하고 충격받은 인우", "representative_moment": "C03이 건네받은 사진을 보고 눈이 커지며 충격에 굳어지는 클로즈업 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "eye-level tight close-up with soft background blur, cold sea-blue dusk contrasted with gold accents", "t2i_prompt": "Photorealistic cinematic still. C03O04 in worn green workwear, an East Asian man in his mid 20s with a handsome face, short dark hair, and a clean facial outline, occupies the middle-center midground. Facing the camera, his shoulders are locked forward with his gaze directed down toward an unseen photograph implied just below the lower frame edge. His widened eyes sit nearly centered in the frame, and his rigid mouth is caught in the lower third. [: The immediate space dissolves directly behind his shoulders into a soft out-of-focus dusk-dark sea and blurred damp deck planks angled away]. Cold sea-blue dusk ambient light undercuts a thin gold backwash from the setting sun, leaving his skin pale and tense with damp shadows pooling in his eye sockets.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O04" } ], "source_facts": [ "C03 looks at the photograph and reacts in shock", "Cold sea-blue dusk undercuts a thin gold backwash from sunset", "sea and deck dissolve into soft dusk blur" ], "visual_inferences": [ "tight focus on C03's facial expression to highlight his sudden recognition", "the photograph itself is placed outside the bottom frame edge to keep focus entirely on C03's widened eyes and rigid mouth", "C03's gaze directed downward to the off-frame object" ], "creative_decisions": [ "framing the face to place widened eyes near center and mouth in lower third", "skipped close framing approach used for background to isolate the character from the environment", "using the contrast of pale damp skin and bruised gold light to heighten the tense mood" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "bench", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cleat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fish box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hull", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sea", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheelhouse", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "c59cedfc223c598c", "owned_hash": "34029bf563f7a9b9", "camera_direction_hash": "83d9c6f07417f22e", "owned_usage_hash": "213be4aa2f0f447f", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 5, "visible_entities": [ "C03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 28, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "On the deck directly in front of 인우, the camera holds at eye level under one meter away in a tight close-up, squared to his face as the sea and deck dissolve into soft dusk blur behind him. His widened eyes sit nearly centered and his rigid mouth is caught in the lower third, while the photograph remains only implied just below frame, making his recognition feel trapped inside his expression.", "lighting_mood": "Cold sea-blue dusk undercuts a thin gold backwash from sunset, leaving 인우's face pale and tense with damp shadow in the eye sockets.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "인우", "screen_zone": "middle_center", "target_id": "C03", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C03" ], "allowed_outlook_pairs": [ { "character_id": "C03", "outlook_id": "O04" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L17B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "fishing_boat_dusk_environment", "element_type": "environment_state", "character_name": "", "description": "The environment is a small fishing boat on the sea during dusk. The lighting is dominated by a cold sea-blue and blue-gray ambient tone, contrasted with a thin, bruised gold backwash from the setting sun. The surrounding structure includes the boat's deck and the entrance to a wheelhouse fitted with glass.", "applies_to_shots": [ 5, 9 ], "element_scope": "full", "source_facts": [ "Cold sea-blue dusk undercuts a thin gold backwash from sunset", "dusk light flickers in blue-gray and bruised gold through wet glass", "sea and deck dissolve into soft dusk blur", "wheelhouse doorway and helm compress behind them" ], "visual_inferences": [ "The ambient dusk light wraps around both the open deck and the enclosed wheelhouse consistently, casting similar color temperatures across the entire boat." ], "creative_decisions": [ "Utilizing the stark contrast between the cold blue sea tones and the fading golden sunset to heighten the dramatic tension of the confrontation on the boat." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C03O04", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "51126d6da2432d29" }, { "scene_index": 28, "heading": "바다 + 갑판 / 해질 무렵", "beat_title": "조타실 입구에서의 거친 제지", "representative_moment": "C01이 조타실로 향하는 C03의 어깨를 강하게 낚아채며 제지하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Handheld medium side composition, flicking dusk light in blue-gray and bruised gold through wet glass.", "t2i_prompt": "Photorealistic cinematic still. C01O02 in a navy waterproof windbreaker, a Korean woman in her early 20s with a youthful oval face, long natural black hair, and soft facial features, stands in the middle-left foreground, reaching for the shoulder of C03O04 in worn green fisher workwear, a Korean man in his mid 20s with a handsome face, short dark hair, and clean facial outline, positioned in the middle-right midground within the shared wheelhouse doorway. C01O02 is seen in right profile, her arm thrust forward across the frame, hand firmly gripping his shoulder to halt his mid-stride escape, her eyes fixed on C03O04, her long natural black hair swaying backward from the sudden forward momentum. C03O04 is angled in three-quarter left, twisting his torso away from her toward the dim interior controls, looking off-screen in panic. A hard shadow cuts across the open door. The cramped wheelhouse compresses the space behind them, while the wet window reflects flickering blue-gray dusk and a thin bruised gold backwash.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" }, { "character_id": "C03", "outlook_id": "O04" } ], "source_facts": [ "Suyeong follows In-woo to the wheelhouse to stop him from turning the boat.", "Cold sea-blue dusk undercuts a thin gold backwash from sunset.", "C01: Korean woman, early 20s, youthful oval face, long natural black hair, soft facial features.", "C03: Korean man, mid 20s, handsome face, short dark hair, clean facial outline." ], "visual_inferences": [ "fixed_element fishing_boat_dusk_environment: ambient dusk light wrapping the open deck and wheelhouse.", "Physical contact abruptly halts In-woo's escape momentum, shown by Suyeong's hair flowing opposite to her lunge.", "The narrow doorway forces the two characters into a tight foreground-midground relationship." ], "creative_decisions": [ "Medium side framing places C01 in the left foreground to emphasize her reaching gesture toward C03.", "Using the narrow doorway as a shared spatial anchor to heighten tension.", "Hard shadows cast across the doorway visually slice the frame, enhancing the abrupt nature of the physical intervention." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bench", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cleat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "A hard shadow cuts across the open door" }, { "owned_token": "fish box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hull", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sea", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheelhouse", "usage_kind": "anchor", "source_phrase": "within the shared wheelhouse doorway" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "while the wet window reflects flickering blue-gray dusk" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "9179d9d425278f22", "owned_hash": "34029bf563f7a9b9", "camera_direction_hash": "5e7a6abac74fdec9", "owned_usage_hash": "f893995e644b03b1", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "within the shared wheelhouse doorway", "reason": "The door is mentioned as a spatial anchor for the characters and shadow placement, with no instructions to redraw or recreate it.", "verdict": "anchor_reference" }, { "owned_object": "wheelhouse", "violating_phrase": "The cramped wheelhouse compresses the space behind them", "reason": "The wheelhouse is referenced to establish the composition and background space, acting as a spatial anchor rather than being newly drawn.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "the wet window reflects flickering blue-gray dusk", "reason": "The window is used as a surface for describing ambient reflections, serving as an anchor reference for environmental lighting.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 9, "visible_entities": [ "C01", "C03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 28, "shot_index": 9 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At the wheelhouse entrance on 수리영's side, shoulder height and about 1.2 meters from her flank, the handheld camera tracks parallel and catches the end of the intervention in a medium side composition. 수리영 occupies the left-mid foreground in profile, arm thrust across frame to 인우 on the right as her hand lands on his shoulder; the wheelhouse doorway and helm compress behind them, making the contact feel abrupt and physical.", "lighting_mood": "Handheld dusk light flickers in blue-gray and bruised gold through wet glass, with a hard shadow cutting across the doorway as 수리영 lunges.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "shared_space_relation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "reaches_for", "gesture_target_label": "인우's shoulder", "label": "수리영", "screen_zone": "middle_left", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "인우", "screen_zone": "middle_right", "target_id": "C03", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "C03" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" }, { "character_id": "C03", "outlook_id": "O04" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L17B01", "reference_usage": "exact_background", "owned_objects": [ "bench", "cleat", "deck", "door", "fish box", "hatch", "hull", "rail", "sea", "wheelhouse", "window" ], "camera_reference": { "camera_position": "on the open working deck near number 1, facing diagonally forward toward number 3 (wheelhouse exterior) and number 8 (wheelhouse doorway), with the rails numbers 4 and 5 framing the sea sightline number 16", "camera_height": "eye-level standing ~1.6m", "lens_hint": "35mm wide angle", "framing_notes": "include the tense deck confrontation elements numbers 17 and 18 when needed, and show the sudden movement cue number 19 leading toward the wheelhouse doorway number 8; camera metadata only, no camera icon on the plan" }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "fishing_boat_dusk_environment", "element_type": "environment_state", "character_name": "", "description": "The environment is a small fishing boat on the sea during dusk. The lighting is dominated by a cold sea-blue and blue-gray ambient tone, contrasted with a thin, bruised gold backwash from the setting sun. The surrounding structure includes the boat's deck and the entrance to a wheelhouse fitted with glass.", "applies_to_shots": [ 5, 9 ], "element_scope": "full", "source_facts": [ "Cold sea-blue dusk undercuts a thin gold backwash from sunset", "dusk light flickers in blue-gray and bruised gold through wet glass", "sea and deck dissolve into soft dusk blur", "wheelhouse doorway and helm compress behind them" ], "visual_inferences": [ "The ambient dusk light wraps around both the open deck and the enclosed wheelhouse consistently, casting similar color temperatures across the entire boat." ], "creative_decisions": [ "Utilizing the stark contrast between the cold blue sea tones and the fading golden sunset to heighten the dramatic tension of the confrontation on the boat." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 28, "shot_index": 5, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C01O02", "policy": "required" }, { "kind": "character_outlook", "id": "C03O04", "policy": "required" }, { "kind": "background", "id": "L17B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "026f7c8b8b2e097b" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "사진 속 아이의 얼굴을 가리키는 손길", "representative_moment": "수리영의 손가락이 사진 속 아이의 얼굴을 꾹 누른 채 멈춘 찰나.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Insert shot with a shallow downward diagonal angle from chest-to-waist height, heavily blurring the cramped background to isolate the fingertip and photograph.", "t2i_prompt": "Photorealistic cinematic still. C01's fingertip in the middle-left foreground reaches for the child's face in the photograph in the lower-center foreground. The single finger of C01, an early 20s Korean woman, presses firmly against the printed cheek of the small child on the hand-sized P09 old photograph. The printed image side faces the camera at a shallow diagonal, resting flat on a scratched, damp console. A narrow cold-blue wash from the wheelhouse window falls across the worn paper, while a faint dirty gold rim light catches the skin of the finger and the printed features. The background is heavily blurred out of focus, keeping the cramped, damp environment soft to emphasize the tight insert framing.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" } ], "source_facts": [ "S#29 action: '수리영의 손가락이 사진 속 아이의 얼굴을 꾹 누른 채 멈춘 찰나.'", "Shot description: insert-like held detail, photo sits in the lower-center of frame with the finger entering from the upper-left.", "DP instruction: A narrow cold-blue wash from the wheelhouse window falls across the photo, faint dirty gold edge catches the fingertip." ], "visual_inferences": [ "fixed_element wheelhouse_sunset_environment: cramped space with damp surfaces, cold blue-gray shadows, dirty gold rim lighting.", "The photo rests on the helm console below the wheelhouse window, anchoring it physically." ], "creative_decisions": [ "Rendered as an extreme close-up (insert scale) to focus strictly on the contact between the fingertip and the printed face.", "Clothing details for C01 are omitted because this is a tight insert on a single finger.", "Background is blurred to highlight the emotional tension anchored on the small printed photograph." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "prop", "background" ], "owned_object_usage": [ { "owned_token": "bench", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "console", "usage_kind": "anchor", "source_phrase": "resting flat on a scratched, damp console" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "panel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "throttle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "from the wheelhouse window" }, { "owned_token": "windshield", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "24f4f984a17c5808", "owned_hash": "e1d214202ccf0083", "camera_direction_hash": "53c5b0fecb3d3bc6", "owned_usage_hash": "f8df7b870ac95590", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "console", "violating_phrase": "resting flat on a scratched, damp console", "reason": "The prompt mentions the console only to describe the resting location of the photograph, acting as a spatial anchor rather than demanding a redraw of the background console.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "from the wheelhouse window", "reason": "The window is mentioned only as a light source for a cold-blue wash, acting as an environmental anchor rather than being redrawn.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "N/A", "_shot_index": 4, "visible_entities": [ "C01", "P09" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 29, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At chest-to-waist height close to the helm surface, the lens angles downward on a shallow diagonal and finishes its dolly-in as 수리영's fingertip presses onto the child's face in the photograph. This is an insert-like held detail: the photo sits in the lower-center of frame with the finger entering from the upper-left, while the blurred wheelhouse surface and damp edges keep the object grounded without letting it overtake the whole frame.", "lighting_mood": "A narrow cold-blue wash from the wheelhouse window falls across the photo, while a faint dirty gold edge catches the fingertip and the printed child's cheek.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "photograph with child's face", "screen_zone": "lower_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "reaches_for", "gesture_target_label": "child's face in the photograph", "label": "수리영's fingertip", "screen_zone": "middle_left", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "P09" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C01", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "Only 수리영's fingertip is visible in this insert, so a full outlook reference would be misleading for the image prompt." } ], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L18B01", "reference_usage": "exact_background", "owned_objects": [ "bench", "console", "door", "hatch", "panel", "shelf", "throttle", "wheel", "window", "windshield" ], "camera_reference": { "camera_position": "near number 2 (aft wheelhouse entrance), facing forward toward number 6 (helm console), number 7 (instrument panel), and number 3 (front windshield), with the side walls in peripheral view for overlay cues number 12 and number 13", "camera_height": "eye-level standing ~1.6m", "lens_hint": "28mm wide angle for tight interior", "framing_notes": "Cramped confrontation framing; include the helm area and forward windows prominently, keep the handheld photo cue number 11 available in the foreground or hand zone, and allow wall surfaces where number 12 and number 13 can be depicted by the background overlay." }, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract); do not enumerate forbidden phrasings here", "camera_reference specifies the framing under which the background was painted — match camera_position, camera_height, and framing", "if deviating from the reference framing (push-in, angle shift, ECU), state the deviation explicitly so the compositor can account for the scale difference" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "wheelhouse_sunset_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of the wheelhouse is a cramped space with damp surfaces. The environment maintains a persistent sunset lighting state, casting cold blue-gray shadows across the interior that are contrasted by faint, dirty gold rim lighting catching the edges of the surfaces and structures.", "applies_to_shots": [ 4, 9 ], "element_scope": "full", "source_facts": [ "S#29. 조타실 / 해질 무렵", "blurred wheelhouse surface and damp edges", "narrow cold-blue wash from the wheelhouse window", "faint dirty gold edge", "cramped wheelhouse wall", "Blue-gray interior shadow", "sickly gold rim" ], "visual_inferences": [ "The damp surfaces and cramped walls define the persistent physical state of the wheelhouse interior.", "The consistent contrast between the cold blue wash and dirty gold rim lighting reflects the distinct late sunset hour." ], "creative_decisions": [ "Combined the physical structural features (damp, cramped) and the lighting scheme (blue-gray and gold) into a single overarching environmental state to ensure consistent visual atmosphere across different shot scales." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character", "id": "C01", "policy": "required" }, { "kind": "background", "id": "L18B01", "policy": "required" }, { "kind": "prop", "id": "P09", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [ { "contract_id": "rc_001", "scope": { "scene_index": 29, "shot_indices": [ 4 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P09", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] } ], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "b50ca211442a5221" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "붉은 원의 환각과 수리영의 붕괴", "representative_moment": "의식을 잃은 C01이 C03의 어깨에 쓰러진 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Tight side close-up at shoulder height, cold blue-gray shadows contrasted by dirty gold rim light.", "t2i_prompt": "Photorealistic cinematic still. C01O02 in a navy windbreaker, a Korean woman in her early 20s, is positioned in the lower-left foreground, her body angled toward C03O04 in green workwear, a Korean man in his mid 20s, who occupies the middle-right foreground. Captured in a tight side close-up at shoulder height, C01O02 appears unconscious, slack and limp in right profile, her youthful oval face with closed eyes pressed firmly against C03O04's rigid shoulder. C03O04, in left profile, stands with his clean facial outline rendered in softer focus, his short dark hair catching the light, eyes fixed on C01O02. The cramped damp space maintains a persistent sunset state, casting cold blue-gray interior shadows across their side profiles, contrasted by a sickly, dirty gold rim light catching C03O04's shoulder and C01O02's long natural black hair. [: A cramped wheelhouse wall immediately behind the figures, its damp surface showing repeated circular markings and one restrained red circle partly obscured by the bodies, bordered by a salt-streaked fogged window.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O02" }, { "character_id": "C03", "outlook_id": "O04" } ], "source_facts": [ "S#29. 조타실 / 해질 무렵", "수리영을 노려보는 인우의 어깨에 그대로 쓰러진다", "인우 얼굴 너머에 보이는 조타실안 수많은 원... 또다시 발견된 붉은 원" ], "visual_inferences": [ "fixed_element wheelhouse_sunset_environment: cramped space with damp surfaces, cold blue-gray shadows, dirty gold rim lighting", "C01's face is sharp primary focus while C03 is softly focused to adhere to close framing constraints" ], "creative_decisions": [ "Applied shallow depth of field to C03 to keep C01 as the primary visible subject in close framing", "Replaced L01 with L02 to describe the background frame-edge surface as requested", "Anchored the wavering circular marks directly onto the damp L02 surface behind them to maintain spatial coherence" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "bench", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "panel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "throttle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "windshield", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "cf3aadc953adc6b0", "owned_hash": "e1d214202ccf0083", "camera_direction_hash": "335fc46aab915d2f", "owned_usage_hash": "d0d02a1bc02eca57", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 9, "visible_entities": [ "C01", "C03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 29, "shot_index": 9 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the flow’s tight side position at shoulder height, about two feet away and parallel to 인우's shoulder line, the handheld camera settles into a suspended side close-up as 수리영's forehead touches his shoulder. 수리영's tilted upper body cuts diagonally from lower-left into 인우's rigid torso on the right, with the cramped wheelhouse wall and circular marks wavering behind them like an occult pattern glimpsed too late.", "lighting_mood": "Blue-gray interior shadow swallows the side profiles, with a sickly gold rim on 인우's shoulder and a restrained red accent from the circular mark behind him.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "shared_space_relation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영", "screen_zone": "lower_left", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "인우", "screen_zone": "middle_right", "target_id": "C03", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "C03" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O02" }, { "character_id": "C03", "outlook_id": "O04" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "L18B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "wheelhouse_sunset_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of the wheelhouse is a cramped space with damp surfaces. The environment maintains a persistent sunset lighting state, casting cold blue-gray shadows across the interior that are contrasted by faint, dirty gold rim lighting catching the edges of the surfaces and structures.", "applies_to_shots": [ 4, 9 ], "element_scope": "full", "source_facts": [ "S#29. 조타실 / 해질 무렵", "blurred wheelhouse surface and damp edges", "narrow cold-blue wash from the wheelhouse window", "faint dirty gold edge", "cramped wheelhouse wall", "Blue-gray interior shadow", "sickly gold rim" ], "visual_inferences": [ "The damp surfaces and cramped walls define the persistent physical state of the wheelhouse interior.", "The consistent contrast between the cold blue wash and dirty gold rim lighting reflects the distinct late sunset hour." ], "creative_decisions": [ "Combined the physical structural features (damp, cramped) and the lighting scheme (blue-gray and gold) into a single overarching environmental state to ensure consistent visual atmosphere across different shot scales." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "C03O04", "policy": "required" } ], "forbidden_refs": [ { "kind": "background", "reason": "close framing skips chain_bg reference" } ], "readiness_policy": "skipped_by_policy", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": true, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "a9411cedbd2ef545" }, { "scene_index": 30, "heading": "S#30. 바다 / 해질 무렵", "beat_title": "안개 속으로 사라지는 어선", "representative_moment": "어선의 앞머리가 짙은 안개 속으로 반쯤 파묻힌 찰나의 몽환적인 실루엣.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low water-level tracking shot, golden backlight silhouette", "t2i_prompt": "Photorealistic cinematic still. The fishing boat bow in the middle-left midground moves toward the thick fog bank in the middle-center background. Captured from an observer low at water level, slightly ahead and off the front quarter, the frame focuses on the bow nosing into the dense white-gray wall. The forward half of the boat is swallowed by the thick fog, while the exposed rear hull edge forms a dark blue-black silhouette against the muted gold and cold gray mist. In the lower frame, the gently choppy, gold-shadowed sea features metallic highlights, its rippled surface reflecting the dying sunset and the boat's dark underside in broken streaks. The low setting sun diffuses through the fog, providing a golden backlight that exposes for silhouette rather than hull detail.", "outfit_assignments": [], "source_facts": [ "S30 action: '서서히 지는 해가 바다위에 황금빛 그림자를 드리운다. 세상과 경계를 지우는 것 같은 짙은 안개 띠. 서서히 안개 안으로 들어가는 인우의 배.'", "shot description: boat bow half-buried in fog, golden sunset" ], "visual_inferences": [ "low tracking camera angle at water level", "blue-black silhouette of the boat against backlit mist", "reflection of the dark underside on the rippled surface" ], "creative_decisions": [ "low water-level framing to emphasize the massive atmospheric barrier of the fog", "golden backlight to create a striking silhouette of the boat entering the unknown" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "anchor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "boat", "usage_kind": "anchor", "source_phrase": "fishing boat bow" }, { "owned_token": "bow", "usage_kind": "anchor", "source_phrase": "fishing boat bow in the middle-left midground" }, { "owned_token": "cleat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "net", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "railing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rope", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sea", "usage_kind": "anchor", "source_phrase": "gold-shadowed sea" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1b6e8abb7a85c958", "owned_hash": "0bce659e13ac9ab9", "camera_direction_hash": "0d23d8c801f8a79f", "owned_usage_hash": "800bb3575d89af65", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "boat", "violating_phrase": "The fishing boat bow", "reason": "The prompt references the boat to describe its movement into the fog and its silhouette, matching the producer's declaration as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "bow", "violating_phrase": "fishing boat bow", "reason": "The prompt references the bow's position and interaction with the fog as a compositional anchor point.", "verdict": "anchor_reference" }, { "owned_object": "sea", "violating_phrase": "gold-shadowed sea", "reason": "The prompt mentions the sea to describe its surface reflection and lighting, serving as an anchor reference.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 28, "dependency_reason": "atmosphere_reference for fog and sea tone", "_shot_index": 3, "visible_entities": [ "L17" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 30, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Low at water level, slightly ahead and off the front quarter of the fishing boat about five meters from the bow, the tracking camera looks back as the boat noses into the fog. The frame is tightened so the bow sits in the middle-left, half erased by a dense white-gray wall, while the exposed hull edge remains a dark silhouette against the gold backlit mist.", "lighting_mood": "Exposure favors the golden backlight through fog, rendering the boat as a blue-black silhouette against muted gold and cold gray mist.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "thick fog bank", "label": "fishing boat bow", "screen_zone": "middle_left", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "thick fog bank", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L17" ], "allowed_outlook_pairs": [], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L17B02", "reference_usage": "exact_background", "owned_objects": [ "anchor", "boat", "bow", "cleat", "deck", "lamp", "net", "railing", "rope", "sea" ], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "owned_objects lists all environment objects already drawn in the background reference — do not create new ones", "if focusing on an owned object, declare its reference-kind via the reference_phrase_kinds sidecar (producer contract)" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [ { "scene_index": 28, "shot_index": 9, "ref_usage": "atmosphere_reference", "kind": "location" } ], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "background", "id": "L17B02", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "9538ca100589c67e" } ], "consumed_bg_catalog_hash": "4d63094cd4bf1bb5", "consumed_shot_binding_hash": "c0627d8357285f2f" }, "step_id": "scene_detail", "run_id": "75a4844b-09a7-419c-9a03-d96c3ab2c5e2", "resolved_model": "gemini-pro", "updated_at": "2026-06-23T16:14:37.168462+00:00", "project_config_snapshot": {} }