{ "status": "completed", "completed_count": 58, "applicable_count": 58, "failed_count": 0, "schema_version": 13, "config_hash": "217c7304484a7c7e", "data": { "scenes": [ { "scene_index": 1, "heading": "S#1. 숲속 / 해 질 무렵", "beat_title": "도망치는 남자의 절박한 달리기", "representative_moment": "피 흘리는 C06이 신발 한 짝이 벗겨진 채 우거진 수풀을 헤치며 필사적으로 달리는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide tracking shot from an observer behind, muted late sunset ambient light", "t2i_prompt": "Photorealistic cinematic still. C06O06 in a grey windproof jacket, an Asian man, appears in the middle-center midground, moving toward the thick undergrowth ahead in the upper-center background. Captured from behind at a three-quarter angle from the right, the figure is frozen mid-stride, looking off-screen in desperation. One bare foot is suspended airborne over the forest floor, with the jacket flapping backward in the direction opposite to the forward motion. A prominent, bleeding bite mark is visible on the outer thigh, crimson blood visibly staining the leg. Muted ambient light of late sunset casts deep ominous shadows across the dense woodland, keeping the fleeing figure legible against the subdued wild vegetation.", "outfit_assignments": [ { "character_id": "C06", "outlook_id": "O06" } ], "source_facts": [ "우거진 수풀을 헤치며 도망치는. 벗겨진 한쪽 신발.", "허벅지에 커다란 이빨 자국. 피가 흐르고.", "mid-stride foot airborne", "man's blood as a limited crimson accent" ], "visual_inferences": [ "fixed_element korean_man_fleeing_state: continuous mid-stride running posture, desperate escape, dirt on bare foot", "fixed_element forest_sunset_environment: dense woodland, tangled undergrowth, ominous shadows" ], "creative_decisions": [ "Wide rear three-quarter tracking composition to preserve the forward pursuit line and full-body motion", "Subdued late sunset lighting to contrast the crimson blood on the thigh and emphasize isolation" ], "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": "736085d6b819407f", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "9851bcc9d9779420", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "작품의 첫 번째 씬이므로 참조할 이전 씬이 없음", "_shot_index": 1, "visible_entities": [ "C06", "L01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 1, "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 waist height several meters behind and slightly off the Korean man's side, track through the undergrowth in a full-body rear three-quarter view. Keep him in the middle of the frame with one bare foot suspended during the stride, while the dense vegetation ahead occupies the upper half and preserves the forward pursuit line.", "lighting_mood": "Muted ambient light appropriate to late sunset keeps the fleeing figure legible against the subdued woodland 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": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "undergrowth ahead", "screen_zone": "upper_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "undergrowth ahead", "label": "한국인 남자", "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", "L01" ], "allowed_outlook_pairs": [ { "character_id": "C06", "outlook_id": "O06" } ], "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": "korean_man_fleeing_state", "element_type": "character_state", "character_name": "Korean man", "description": "The Korean man is in a continuous mid-stride running posture, desperately trying to escape. One of his feet is entirely bare, having lost a shoe. There is a prominent, bleeding bite mark on his thigh, with crimson blood visibly staining his leg.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "우거진 수풀을 헤치며 도망치는. 벗겨진 한쪽 신발.", "허벅지에 커다란 이빨 자국. 피가 흐르고.", "mid-stride foot airborne", "man's blood as a limited crimson accent" ], "visual_inferences": [ "His clothing is slightly disheveled and stained due to the frantic escape through the wild undergrowth.", "The bare foot is covered in dirt from running on the forest floor." ], "creative_decisions": [ "Positioning the bloody bite mark on the outer thigh so it remains clearly visible from both rear three-quarter and side-on angles." ], "confidence": "high" }, { "element_id": "forest_sunset_environment", "element_type": "environment_state", "character_name": "", "description": "The environment is a dense woodland filled with thick, tangled undergrowth and wild vegetation. The lighting reflects a late sunset, casting subdued, low ambient light and deep, ominous shadows across the forest floor.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "숲속 / 해 질 무렵", "우거진 수풀", "Muted ambient light appropriate to late sunset keeps the fleeing figure legible", "Late-sunset ambient light is held low and restrained" ], "visual_inferences": [ "The surrounding foliage forms a visually oppressive and dark background, emphasizing the isolation of the location." ], "creative_decisions": [ "Keeping the ambient light strictly directional from the fading sun to enhance the contrast between the dark vegetation and the visible figures." ], "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": "C06O06", "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": "491b78d79cd37a3f" }, { "scene_index": 1, "heading": "S#1. 숲속 / 해 질 무렵", "beat_title": "절체절명의 공격", "representative_moment": "피를 흘리며 도망치던 C06이 쏜살같이 날아든 검은 인영(C11)에게 덮쳐지기 직전의 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "At chest height several meters beside the escape line, track laterally in a wide side-on composition", "t2i_prompt": "Photorealistic cinematic still. C06O06 in a grey windproof jacket, a Korean adult man, is caught mid-stride recoiling forward in the lower-right midground. An airborne black humanoid figure in tattered dark clothing moves toward the Korean man from the upper-left midground. Captured from several meters away at chest height in a wide side-on composition, the two full bodies are separated by a narrow pocket of visible space in a dense woodland filled with thick, tangled undergrowth. C06O06, seen in right profile and looking off-screen, is frozen in a continuous running posture, one bare dirt-covered foot lifted off the ground, displaying a prominent bleeding bite mark on his outer thigh with crimson blood visibly staining his leg. The airborne black figure, positioned in right profile with arms outstretched diagonally toward the man, appears as an ominous dark mass; its face is fully swallowed in shadow with no discernible features, and its body volume is preserved in silhouette form against the dim background. The subdued late-sunset ambient light casts deep shadows across the forest floor, highlighting the crimson blood as a limited accent amidst the dark vegetation.", "outfit_assignments": [ { "character_id": "C06", "outlook_id": "O06" } ], "source_facts": [ "우거진 수풀을 헤치며 도망치는. 벗겨진 한쪽 신발.", "허벅지에 커다란 이빨 자국. 피가 흐르고.", "어디선가 쏜살같이 달려온 검은 인영이 남자를 덮친다." ], "visual_inferences": [ "fixed_element: korean_man_fleeing_state (continuous mid-stride posture, bare foot covered in dirt, bleeding bite mark prominently displayed)", "fixed_element: forest_sunset_environment (dense woodland, tangled undergrowth, low late-sunset ambient light)" ], "creative_decisions": [ "Chest-height wide side-on framing from a distance to capture both full bodies separated by a narrow gap the instant before impact.", "Emphasizing the black figure's ominous nature by rendering it as an airborne silhouette with preserved volume but fully obscured facial features." ], "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": "1b359479d6e611af", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "c7ba19696978dc77", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 1, "dependency_reason": "동일 씬 내 연속되는 도주 및 피습 상황으로, 숲의 배경 요소와 시간대(해 질 무렵)의 조명 상태를 일관되게 유지하기 위함", "_shot_index": 4, "visible_entities": [ "C06", "C11", "L01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 1, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At chest height several meters beside the escape line, track laterally in a wide side-on composition and freeze the instant before impact. Place the airborne black figure in the upper-left reaching diagonally toward the Korean man at lower-right, with both full bodies separated by a narrow pocket of visible space.", "lighting_mood": "Late-sunset ambient light is held low and restrained, with the black figure reading as an ominous dark mass and the man's blood as a limited crimson accent.", "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": "none", "gesture_target_label": "", "label": "한국인 남자", "screen_zone": "lower_right", "target_id": "C06", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "한국인 남자", "label": "검은 인영", "screen_zone": "upper_left", "target_id": "C11", "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", "C11", "L01" ], "allowed_outlook_pairs": [ { "character_id": "C06", "outlook_id": "O06" }, { "character_id": "C11", "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": [ { "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": [ { "element_id": "korean_man_fleeing_state", "element_type": "character_state", "character_name": "Korean man", "description": "The Korean man is in a continuous mid-stride running posture, desperately trying to escape. One of his feet is entirely bare, having lost a shoe. There is a prominent, bleeding bite mark on his thigh, with crimson blood visibly staining his leg.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "우거진 수풀을 헤치며 도망치는. 벗겨진 한쪽 신발.", "허벅지에 커다란 이빨 자국. 피가 흐르고.", "mid-stride foot airborne", "man's blood as a limited crimson accent" ], "visual_inferences": [ "His clothing is slightly disheveled and stained due to the frantic escape through the wild undergrowth.", "The bare foot is covered in dirt from running on the forest floor." ], "creative_decisions": [ "Positioning the bloody bite mark on the outer thigh so it remains clearly visible from both rear three-quarter and side-on angles." ], "confidence": "high" }, { "element_id": "forest_sunset_environment", "element_type": "environment_state", "character_name": "", "description": "The environment is a dense woodland filled with thick, tangled undergrowth and wild vegetation. The lighting reflects a late sunset, casting subdued, low ambient light and deep, ominous shadows across the forest floor.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "숲속 / 해 질 무렵", "우거진 수풀", "Muted ambient light appropriate to late sunset keeps the fleeing figure legible", "Late-sunset ambient light is held low and restrained" ], "visual_inferences": [ "The surrounding foliage forms a visually oppressive and dark background, emphasizing the isolation of the location." ], "creative_decisions": [ "Keeping the ambient light strictly directional from the fading sun to enhance the contrast between the dark vegetation and the visible figures." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 1, "shot_index": 1, "ref_usage": "", "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": "C06O06", "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": "af35c8503f0118ab" }, { "scene_index": 2, "heading": "S#2. 도시 뒷골목 / N", "beat_title": "포식의 시작", "representative_moment": "가로등 불빛 아래, 붉은 눈동자의 C10이 전봇대에 밀어붙인 C09의 목을 물어뜯는 순간", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium two-shot from chest height at a perpendicular side angle, with dim, low-key streetlamp illumination casting heavy shadows.", "t2i_prompt": "Photorealistic cinematic still. Shot from chest height at a perpendicular side angle, a medium two-shot captures the upper bodies of two figures. On the right, C09O09 in a beige trench coat, a Korean adult woman in left profile, is physically compressed, her shoulder firmly braced and buckling against a shared sturdy, vertical concrete utility pole positioned immediately behind her. Sharing this rigid spatial anchor, C10O10 in a dark long coat, a Korean adult man in right profile, acts as the force initiator, pressing in aggressively from the left. His face is positioned tightly against her exposed neck, his teeth bared at the contact point, and his visible eye features a distinct, glowing red pupil. The urban back alley is bathed in dim, low-key streetlamp illumination, creating high-contrast shadows that isolate their entwined, overlapping profiles while the dark alleyway recedes behind them.", "outfit_assignments": [ { "character_id": "C09", "outlook_id": "O09" }, { "character_id": "C10", "outlook_id": "O10" } ], "source_facts": [ "Shot 1: 희미한 가로등 불빛 아래 전봇대에 밀착된 채 서 있는 한국인 남자와 한국인 여자의 옆모습", "S#2 원문: 남자의 붉은 눈동자, 목의 살점을 뜯는 이빨" ], "visual_inferences": [ "fixed_element man_red_eyes_biting: aggressive positioning against the neck, glowing red pupil", "fixed_element woman_shoulder_pinned: shoulder buckling against the pole, neck exposed", "fixed_element concrete_utility_pole: acts as a rigid compositional boundary", "fixed_element dim_streetlamp_lighting: high-contrast shadows isolate the subjects" ], "creative_decisions": [ "Chest-height side angle captures their overlapping profiles as specified in the DP directions.", "Muted color palette with deep shadows emphasizes the glowing red eye as a supernatural visual cue.", "The utility pole functions as a physical shared anchor to bind the foreground figures into the same interaction space." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "b68b5a9307f7474c", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "6197973d7377744c", "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": [ "C09", "C10", "L02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 2, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From chest height at a perpendicular side angle, hold a static medium two-shot of the pair compressed against the utility pole. The woman is braced on the right while the man presses in from the left, their overlapping profiles occupying the central band with the alley receding behind them.", "lighting_mood": "Dim streetlamp illumination isolates muted profiles against the subdued nighttime alley.", "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": [ "C09", "C10", "L02" ], "allowed_outlook_pairs": [ { "character_id": "C09", "outlook_id": "O09" }, { "character_id": "C10", "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": 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": "man_red_eyes_biting", "element_type": "character_state", "character_name": "한국인 남자", "description": "A Korean man positions his face aggressively against the neck of the woman from the left side. His visible eye has a distinct, glowing red pupil, serving as a striking focal point in the darkness during his predatory attack.", "applies_to_shots": [ 1, 3 ], "element_scope": "close", "source_facts": [ "남자의 붉은 눈동자", "목의 살점을 뜯는 이빨", "pressing in from the left", "man's open mouth, embedded teeth, and partial red eye" ], "visual_inferences": [ "The red eye glows slightly to stand out prominently against the dim ambient streetlamp lighting." ], "creative_decisions": [ "The man's profile and face are tightly composed to emphasize the red eye as a supernatural and dangerous visual cue." ], "confidence": "high" }, { "element_id": "woman_shoulder_pinned", "element_type": "character_state", "character_name": "한국인 여자", "description": "A Korean woman is positioned on the right side of the interaction, with her shoulder firmly braced and buckling against a vertical utility pole. Her neck remains exposed to the man on her left as her body is physically compressed against the hard surface.", "applies_to_shots": [ 1, 3 ], "element_scope": "close", "source_facts": [ "전봇대에 뒤엉켜있는 두 사람", "woman is braced on the right", "buckling against pole", "Her shoulder and neck occupy the right foreground" ], "visual_inferences": [ "Her shoulder is pressed tightly against the rough texture of the pole, showing signs of physical struggle and restriction." ], "creative_decisions": [ "Her posture visually emphasizes absolute vulnerability, showing her completely trapped between the attacker and the unyielding pole." ], "confidence": "high" }, { "element_id": "dim_streetlamp_lighting", "element_type": "environment_state", "character_name": "", "description": "The urban back alley is bathed in dim, low-key streetlamp illumination. The subdued lighting isolates the two subjects in the darkness, creating high-contrast shadows that obscure the receding alleyway behind them.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "도심 희미한 가로등 불빛", "Dim streetlamp illumination isolates muted profiles", "Dim streetlamp light sustains low-key contrast" ], "visual_inferences": [ "The light source comes from above and slightly off-screen, casting long shadows that highlight only the edges of the figures." ], "creative_decisions": [ "A muted color palette with heavy, dramatic shadows is employed to strictly direct the viewer's attention to the characters' interaction." ], "confidence": "high" }, { "element_id": "concrete_utility_pole", "element_type": "persistent_prop", "character_name": "", "description": "A sturdy, vertical concrete utility pole stands positioned immediately behind the woman. It acts as a permanent structural anchor in the scene, physically limiting the woman's movement on the right side.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "전봇대에 뒤엉켜있는 두 사람", "compressed against the utility pole", "utility pole remains only as a narrow background anchor" ], "visual_inferences": [ "The pole has a rough, standard concrete texture typical of older urban back alleys in Korea." ], "creative_decisions": [ "The pole functions as a rigid compositional boundary that claustrophobically boxes the woman into the frame." ], "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": "C09O09", "policy": "required" }, { "kind": "character_outlook", "id": "C10O10", "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": "38f57477f7f277dc" }, { "scene_index": 2, "heading": "S#2. 도시 뒷골목 / N", "beat_title": "포식자의 습격과 절명", "representative_moment": "어두운 전봇대 옆에서 C10이 붉은 눈을 번뜩이며 C09의 목을 물어뜯는 순간", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "tight over-shoulder impact composition at neck height, dim streetlamp low-key contrast", "t2i_prompt": "Photorealistic cinematic still. Shot at neck height within arm's reach for a tight over-shoulder impact composition. C09O09 in a beige trench coat, a Korean adult woman, occupies the right foreground. Seen from over her shoulder, her body is slack and severely injured, her shoulder firmly braced and buckling against a vertical concrete utility pole, gazing off-screen. Her neck remains fully exposed to the one driving the action: C10O10 in a black long coat, a Korean adult man, pressing in from the center-left at a three-quarter angle from the right. C10O10's face is aggressively positioned, his open mouth and embedded teeth locked at her neck, his gaze directed down at the shared contact point. His visible eye features a distinct, glowing red pupil. The concrete utility pole serves as a narrow structural background anchor immediately behind the woman. Dim streetlamp light sustains low-key contrast, isolating the figures in high-contrast shadow to obscure the omitted background, with the man's red eye forming a restrained crimson accent.", "outfit_assignments": [ { "character_id": "C09", "outlook_id": "O09" }, { "character_id": "C10", "outlook_id": "O10" } ], "source_facts": [ "전봇대에 뒤엉켜있는 두 사람", "남자의 붉은 눈동자, 목의 살점을 뜯는 이빨… 푹 힘없이 쓰러지는 여자", "도심 희미한 가로등 불빛", "Her shoulder and neck occupy the right foreground while the man's open mouth, embedded teeth, and partial red eye hold the center-left" ], "visual_inferences": [ "fixed_element man_red_eyes_biting: aggressive posture from the left, visible red pupil glowing against dim ambient light", "fixed_element woman_shoulder_pinned: physically compressed against the concrete pole, unable to escape", "utility pole acts as a rigid compositional boundary boxing the woman into the frame" ], "creative_decisions": [ "Utilized a tight over-shoulder framing to emphasize the violent physical contact at the neck and the glowing red eye", "Used deep, heavy shadows from low-key streetlamp lighting to naturally obscure the skipped background and isolate the characters" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "3c3c4183a41247ac", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "00d3c812b362499b", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 2, "dependency_reason": "같은 씬의 연속된 액션 중 클로즈업 컷이므로 이전 샷의 인물 위치와 환경 톤(어두운 가로등 조명)을 일관되게 유지", "_shot_index": 3, "visible_entities": [ "C09", "C10" ], "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 neck height within arm's reach and slightly behind the woman's shoulder, stop the tracking move on a tight over-shoulder impact composition. Her shoulder and neck occupy the right foreground while the man's open mouth, embedded teeth, and partial red eye hold the center-left; the utility pole remains only as a narrow background anchor.", "lighting_mood": "Dim streetlamp light sustains low-key contrast, with the man's red eye forming the restrained crimson accent.", "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": [ "C09", "C10" ], "allowed_outlook_pairs": [ { "character_id": "C09", "outlook_id": "O09" }, { "character_id": "C10", "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": 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_red_eyes_biting", "element_type": "character_state", "character_name": "한국인 남자", "description": "A Korean man positions his face aggressively against the neck of the woman from the left side. His visible eye has a distinct, glowing red pupil, serving as a striking focal point in the darkness during his predatory attack.", "applies_to_shots": [ 1, 3 ], "element_scope": "close", "source_facts": [ "남자의 붉은 눈동자", "목의 살점을 뜯는 이빨", "pressing in from the left", "man's open mouth, embedded teeth, and partial red eye" ], "visual_inferences": [ "The red eye glows slightly to stand out prominently against the dim ambient streetlamp lighting." ], "creative_decisions": [ "The man's profile and face are tightly composed to emphasize the red eye as a supernatural and dangerous visual cue." ], "confidence": "high" }, { "element_id": "woman_shoulder_pinned", "element_type": "character_state", "character_name": "한국인 여자", "description": "A Korean woman is positioned on the right side of the interaction, with her shoulder firmly braced and buckling against a vertical utility pole. Her neck remains exposed to the man on her left as her body is physically compressed against the hard surface.", "applies_to_shots": [ 1, 3 ], "element_scope": "close", "source_facts": [ "전봇대에 뒤엉켜있는 두 사람", "woman is braced on the right", "buckling against pole", "Her shoulder and neck occupy the right foreground" ], "visual_inferences": [ "Her shoulder is pressed tightly against the rough texture of the pole, showing signs of physical struggle and restriction." ], "creative_decisions": [ "Her posture visually emphasizes absolute vulnerability, showing her completely trapped between the attacker and the unyielding pole." ], "confidence": "high" }, { "element_id": "dim_streetlamp_lighting", "element_type": "environment_state", "character_name": "", "description": "The urban back alley is bathed in dim, low-key streetlamp illumination. The subdued lighting isolates the two subjects in the darkness, creating high-contrast shadows that obscure the receding alleyway behind them.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "도심 희미한 가로등 불빛", "Dim streetlamp illumination isolates muted profiles", "Dim streetlamp light sustains low-key contrast" ], "visual_inferences": [ "The light source comes from above and slightly off-screen, casting long shadows that highlight only the edges of the figures." ], "creative_decisions": [ "A muted color palette with heavy, dramatic shadows is employed to strictly direct the viewer's attention to the characters' interaction." ], "confidence": "high" }, { "element_id": "concrete_utility_pole", "element_type": "persistent_prop", "character_name": "", "description": "A sturdy, vertical concrete utility pole stands positioned immediately behind the woman. It acts as a permanent structural anchor in the scene, physically limiting the woman's movement on the right side.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "전봇대에 뒤엉켜있는 두 사람", "compressed against the utility pole", "utility pole remains only as a narrow background anchor" ], "visual_inferences": [ "The pole has a rough, standard concrete texture typical of older urban back alleys in Korea." ], "creative_decisions": [ "The pole functions as a rigid compositional boundary that claustrophobically boxes the woman into the frame." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 2, "shot_index": 1, "ref_usage": "", "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": "C10O10", "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": "c4e14ffc26c4f116" }, { "scene_index": 3, "heading": "S#3. 현장 / D", "beat_title": "현장 지휘하는 혜수 (뉴스 화면)", "representative_moment": "지역 뉴스 화면 속, 형사 무리 한가운데 서서 사건 현장을 가리키며 수사를 지휘하는 혜수의 모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Chest-height dolly-in, medium framing, through-device regional news footage with subdued brightness and slight broadcast flattening", "t2i_prompt": "Photorealistic cinematic still. A Korean middle-aged woman with neat black hair in a dark leather blouson stands at a three-quarter angle from the right in the middle-left midground, leaning into a pointing gesture toward the incident area in the middle-right background. The camera is positioned at chest height, framing her from the waist up. An Asian adult detective stands in the foreground occupying the same crime scene space, seen from over the shoulder with his back to the camera, his shoulder turned toward the woman and his eyes fixed on her. The woman's gaze is directed off-screen. Around them, the cropped shoulders and partial profiles of other Asian detectives form a human boundary. A police line marks the controlled boundary of the crime scene in the lush green forest behind the group. Subdued daytime source illumination with slight broadcast flattening and restrained contrast washes over the figures.", "outfit_assignments": [], "source_facts": [ "S#3 action: '형사무리 속에 사건을 지휘하는 혜수의 모습이 보인다.'", "shot description: Hyesu's upper body in the middle of the detective group, pointing to one side", "scene location: lush green forest crime scene" ], "visual_inferences": [ "The scene is viewed entirely through a regional news broadcast surface, necessitating demographic descriptors instead of direct entity IDs per reproduction surface rules", "Other detectives form a tight physical ring around Hyesu, emphasizing her command" ], "creative_decisions": [ "Applied 'middle-left midground' and 'pointing toward the incident area in the middle-right background' to satisfy the frame spatial contract", "Added 'occupying the same crime scene space' to satisfy the foreground/background shared anchor rule", "Employed slight broadcast flattening to the lighting to convey the through-device perception mode without rendering literal screen frames" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "970c5ac597095a14", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "142b9dfd0fa2b4ed", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 1, "dependency_reason": "동일한 숲속 사건 현장을 배경으로 하므로 공간적 연속성을 유지", "_shot_index": 3, "visible_entities": [ "C02", "L01" ], "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": "Within the regional-news image, dolly in at chest height from the front of the detective group until Hyesu's upper body anchors the center-left. Retain cropped shoulders and partial profiles of nearby detectives around her while her pointing hand cuts toward the right edge of the broadcast frame.", "lighting_mood": "Daytime source illumination is reproduced with subdued brightness and slight broadcast flattening while preserving restrained contrast.", "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": "points_to_anchor", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "incident area", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "points_to", "gesture_target_label": "incident area at screen right", "label": "혜수", "screen_zone": "middle_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": [ "C02", "L01" ], "allowed_outlook_pairs": [ { "character_id": "C02", "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": "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": 1, "ref_usage": "", "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": "933e6460aca43556" }, { "scene_index": 4, "heading": "S#4. 옥탑방 / D", "beat_title": "적막한 옥탑방 전경", "representative_moment": "맑은 낮, 높은 곳에서 허름한 빌라 옥탑방을 내려다보는 부감 샷.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "steep downward omniscient high angle, muted natural color, subdued brightness", "t2i_prompt": "Photorealistic cinematic still. The rooftop room sits firmly in the middle-center midground, while the villa rooftop spans the lower-center midground. Captured from a steep downward omniscient high angle, the older multi-unit building is seen from above, with the sprawling Incheon outskirts visible in the distance. Clear daytime ambient light casts muted natural colors and subdued brightness over the exterior. A water tank, a drying platform, and a clothesline anchor the quiet rooftop deck along the parapet. A small door and window break the wall of the rooftop room, and a narrow staircase leads downward. The camera focuses entirely on the weathered textures and isolated architecture of the rooftop.", "outfit_assignments": [], "source_facts": [ "scene_text: 인천 변두리 부감, 오래된 빌라 옥탑방 전경.", "shot description: 맑은 낮, 오래된 다세대 빌라 옥상 위에 자리 잡은 옥탑방을 높은 곳에서 내려다본 부감 구도" ], "visual_inferences": [ "isolated rooftop room surrounded by the sprawling cityscape", "weathered textures on the rooftop deck and parapet" ], "creative_decisions": [ "steep downward angle focuses on the architectural structure and its isolation", "muted daytime lighting emphasizes a grounded, bleak urban realism" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "anchor", "source_phrase": "and a clothesline" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "A small door" }, { "owned_token": "drying platform", "usage_kind": "anchor", "source_phrase": "a drying platform" }, { "owned_token": "parapet", "usage_kind": "anchor", "source_phrase": "along the parapet" }, { "owned_token": "rooftop deck", "usage_kind": "anchor", "source_phrase": "the quiet rooftop deck" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "The rooftop room sits firmly" }, { "owned_token": "staircase", "usage_kind": "anchor", "source_phrase": "a narrow staircase" }, { "owned_token": "water tank", "usage_kind": "anchor", "source_phrase": "A water tank" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "and window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "7599fdbce2510e9d", "owned_hash": "de779143439b7423", "camera_direction_hash": "401cb79299225717", "owned_usage_hash": "708e993bb2374570", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "clothesline", "violating_phrase": "and a clothesline anchor", "reason": "The prompt mentions the clothesline as a static spatial anchor on the rooftop deck, which aligns with the producer's 'anchor' classification.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "A small door", "reason": "The door is mentioned descriptively as breaking the wall of the existing rooftop room structure, matching the producer's 'anchor' classification.", "verdict": "anchor_reference" }, { "owned_object": "drying platform", "violating_phrase": "a drying platform", "reason": "The drying platform is referenced descriptively to establish spatial context on the rooftop deck, which is a valid anchor use.", "verdict": "anchor_reference" }, { "owned_object": "parapet", "violating_phrase": "along the parapet", "reason": "The parapet is mentioned solely to provide spatial reference for other background objects, which matches the 'anchor' classification.", "verdict": "anchor_reference" }, { "owned_object": "rooftop deck", "violating_phrase": "the quiet rooftop deck", "reason": "The rooftop deck serves as a spatial reference for the overall composition, which aligns with the producer's 'anchor' declaration.", "verdict": "anchor_reference" }, { "owned_object": "rooftop room", "violating_phrase": "The rooftop room sits firmly", "reason": "The rooftop room is referenced to position the camera and describe the composition of the scene, confirming its role as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "staircase", "violating_phrase": "a narrow staircase leads downward", "reason": "The staircase is mentioned as a structural anchor pointing downwards, requiring no redrawing or modification of the object.", "verdict": "anchor_reference" }, { "owned_object": "water tank", "violating_phrase": "A water tank", "reason": "The water tank is referenced as a static environmental element anchoring the deck, which matches the producer's 'anchor' classification.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "and window", "reason": "The window is described as a static structural detail of the rooftop room, validating the 'anchor' classification.", "verdict": "anchor_reference" } ] } } ], "scene_type": "voiceover", "dependent_scene_index": -1, "dependency_reason": "", "_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": "From high above the rooftop, continue the descending dolly-in while maintaining a steep downward angle on the older multi-unit villa. Settle with the rooftop room near the middle of the frame, the surrounding roof forming a clear lower spatial base and the Incheon outskirts remaining visible beyond the building.", "lighting_mood": "Clear daytime ambient light presents the rooftop layout with subdued brightness and muted natural color.", "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": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "rooftop room", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "villa rooftop", "screen_zone": "lower_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": [ "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", "parapet", "rooftop deck", "rooftop room", "staircase", "water tank", "window" ], "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": [], "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": "a0c5aab5a63cecb9" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "황급히 뉴스를 끄는 민숙", "representative_moment": "수리영의 기척에 놀라 황급히 상체를 기울여 TV 전원 버튼을 누르는 C04의 측면 모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Observer POV medium framing, slightly below shoulder height, subdued interior daylight mixed with fading screen glow", "t2i_prompt": "Photorealistic cinematic still. Captured from a medium observer viewpoint slightly below shoulder height, C04O04 in a burgundy t-shirt, a Korean early-40s woman, occupies the left and middle of the frame in right-facing profile. Her torso is pitched forward in a mid-lunge posture, eyes looking down, with one arm extended into the lower-right foreground, a fingertip firmly pressed against the power button of the TV. The TV occupies less than a third of the frame on the right, its front screen and dark casing angled partly toward the camera, displaying a fading news image. Behind the figure, a small window lets in a narrow beam of sunlight. The space reflects a subdued interior daylight environment; gentle natural sunlight from the window provides basic illumination, combining with the dimming glow of the TV to maintain a realistic and slightly dim atmosphere.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O04" } ], "source_facts": [ "S#5 action: '수리영이 방문을 열고 기지개를 켜며 나오자 후다닥 TV를 끄는 민숙.'", "shot description: 다급하게 상체를 기울이며 손가락 끝으로 TV 전원 버튼을 세게 누르고 있는 민숙의 측면 구도" ], "visual_inferences": [ "fixed_element rooftop_interior_lighting: subdued interior daylight environment, gentle natural sunlight provides basic illumination.", "Minsuk's urgent reaction is visualized as a mid-lunge posture with her finger pressed against the TV button." ], "creative_decisions": [ "Framed from slightly below shoulder height with Minsuk in right profile on the left/middle, keeping the TV on the right edge to anchor their physical connection in the same space.", "Lighting perfectly balances the narrow window daylight and the fading TV screen to create a grounded, dim atmosphere without over-stylization." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "redraw", "source_phrase": "power button of the TV. The TV occupies less than a third of the frame" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "small window lets in a narrow beam of sunlight" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "6321fe76ceda9126", "owned_hash": "bac52cdc87402e26", "camera_direction_hash": "dcd730e068ef6b1f", "owned_usage_hash": "8947953752819e34", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "TV", "violating_phrase": "fingertip firmly pressed against the power button of the TV. The TV occupies less than a third of the frame on the right, its front screen and dark casing angled partly toward the camera, displaying a fading news image", "reason": "The TV's state of displaying a fading news image and its power button interaction are structurally required by the shot intent of C04 pressing the TV power button, qualifying for allowed visual state change.", "verdict": "allowed_visual_state_change" }, { "owned_object": "window", "violating_phrase": "Behind the figure, a small window lets in a narrow beam of sunlight", "reason": "The window is mentioned as a spatial anchor and a source of natural sunlight without any attempt to redesign or redraw the physical object itself, aligning with the producer's echo.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 5, "visible_entities": [ "C04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From slightly below shoulder height at Minsuk's side, finish the dolly-in within arm's reach of the television controls. Frame her in right-facing profile across the left and middle of the image, torso pitched forward and one hand crossing into the lower-right to strike the power button while the television occupies less than a third of the frame.", "lighting_mood": "Narrow sunlight from the small window combines with the fading television image in a subdued interior 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": [ "C04" ], "allowed_outlook_pairs": [ { "character_id": "C04", "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": "background_ref_attached", "bg_id": "L04B04", "reference_usage": "exact_background", "owned_objects": [ "TV", "TV stand", "chair", "counter", "dining table", "door", "drying platform", "floor lamp", "sink", "stovetop", "window" ], "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": [ { "element_id": "rooftop_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "A subdued interior daylight environment inside the rooftop room. Gentle natural sunlight provides basic illumination, maintaining a realistic and slightly dim atmosphere without harsh stylized distortion.", "applies_to_shots": [ 5, 9 ], "element_scope": "full", "source_facts": [ "Narrow sunlight from the small window combines with the fading television image in a subdued interior palette.", "Interior daylight keeps the wrist natural and subdued" ], "visual_inferences": [ "The gentle natural lighting casts soft shadows, creating a visually unified tone that bridges both the wider room view and the isolated detail shot." ], "creative_decisions": [ "Ensure the daylight remains subdued and natural across different framing sizes to ground the mood of the scene in quiet reality." ], "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": "C04O04", "policy": "required" }, { "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": "8b421ee87656c8d2" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "손목의 붉은 표식", "representative_moment": "민숙의 손목에 선명하게 드러난 작은 원형 표식 타이트 샷", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "insert shot, sharp detail on wrist, soft background blur", "t2i_prompt": "Photorealistic cinematic still. The wrist and circular mark of C04 in a magenta t-shirt, a Korean woman in her 40s, are sharply focused in the middle-center foreground. The isolated insert shot frames the forearm nearly perpendicular to the lens, capturing the small, clear circular mark with a narrow margin of skin for scale. Her alive figure is partially visible in the soft background, wrist held near her torso, looking away with her gaze directed off-screen. Subdued interior daylight provides gentle natural illumination over the skin, maintaining a slightly dim and realistic atmosphere without harsh stylized distortion. The surrounding background room is completely out of focus, preserving the strict isolation of the detail.", "outfit_assignments": [ { "character_id": "C04", "outlook_id": "O04" } ], "source_facts": [ "민숙의 손목에 선명하게 드러난 작은 원형 표식의 타이트 샷", "A subdued interior daylight environment inside the rooftop room.", "Gentle natural sunlight provides basic illumination" ], "visual_inferences": [ "fixed_element rooftop_interior_lighting: realistic and slightly dim atmosphere on the wrist", "shallow depth of field blurs the background to isolate the mark as an insert shot" ], "creative_decisions": [ "Applied a macro/insert framing scale with shallow focus to emphasize the small circular mark while keeping the rest of the body and the room out of focus", "Integrated the DP's direction of the wrist held near the torso and looking away into the soft background layer" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "0176007cc494651b", "owned_hash": "bac52cdc87402e26", "camera_direction_hash": "859db78eb3262ab1", "owned_usage_hash": "d7faed32ebc64f25", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "Same room and continuous time from the previous shot, utilizing the established subdued interior lighting.", "_shot_index": 9, "visible_entities": [ "C04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 9 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At wrist height and nearly perpendicular to Minsuk's forearm, complete the dolly-in as an isolated detail shot. Hold the small circular mark sharply near the center while retaining a narrow margin of wrist and forearm for scale, with no unrelated figure entering the frame.", "lighting_mood": "Interior daylight keeps the wrist natural and subdued while allowing the unusually clear circular mark to read without stylized distortion.", "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": "민숙 wrist and circular mark", "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" ], "allowed_outlook_pairs": [ { "character_id": "C04", "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": [ { "subject_id": "C04", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The insert shows only Minsuk's wrist and circular mark, so a full outlook reference would not align with the visible anatomy." } ], "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": [ "TV", "TV stand", "chair", "counter", "dining table", "door", "drying platform", "floor lamp", "sink", "stovetop", "window" ], "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": [ { "element_id": "rooftop_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "A subdued interior daylight environment inside the rooftop room. Gentle natural sunlight provides basic illumination, maintaining a realistic and slightly dim atmosphere without harsh stylized distortion.", "applies_to_shots": [ 5, 9 ], "element_scope": "full", "source_facts": [ "Narrow sunlight from the small window combines with the fading television image in a subdued interior palette.", "Interior daylight keeps the wrist natural and subdued" ], "visual_inferences": [ "The gentle natural lighting casts soft shadows, creating a visually unified tone that bridges both the wider room view and the isolated detail shot." ], "creative_decisions": [ "Ensure the daylight remains subdued and natural across different framing sizes to ground the mood of the scene in quiet reality." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 5, "ref_usage": "", "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": "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": "a8a55769e72cc3a3" }, { "scene_index": 6, "heading": "S#6. 해안가 도로 / D", "beat_title": "해안도로를 달리는 활기찬 수리영", "representative_moment": "C01이 이어폰을 낀 채 미소를 지으며 자전거 페달을 밟아 해안도로를 달리고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Travel parallel at torso height beside the rider, medium-distance profile with natural daytime ambient light", "t2i_prompt": "Photorealistic cinematic still. C01O01, a Korean woman in her early 20s in a khaki sweatshirt, occupies the middle-center midground, moving toward the road ahead at screen right, with the coastal road ahead visible in the middle-right background. Captured from a parallel travel angle at torso height, she is seated on the bicycle saddle in mid-pedal, one knee bent in the down stroke. Her working leg, the bicycle frame, and the front wheel occupy the lower half of the frame. Seen in right profile, her oval youthful face with soft features and a natural skin tone bears a bright smile, her eyes gazing into the distance with earphones in. Her long black hair trails toward the left in the direction opposite to her forward motion, leaving open forward space on the right. The sea stretches beside the road under natural daytime ambient light, keeping the energetic ride bright but restrained in color and contrast.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#6 action: '수리영 자전거를 타고 해안가 도로를 달린다. 이어폰을 끼고... 밝고 활기차 보인다.'", "shot description: seated mid-pedal downstroke, smiling in profile, hair trailing back due to bicycle momentum" ], "visual_inferences": [ "bicycle frame and front wheel are visible in the lower frame to anchor the riding posture", "hair trailing leftward visually reinforces the rightward forward momentum" ], "creative_decisions": [ "parallel travel composition at torso height to keep her smiling face in the upper center and the working leg mechanism below", "bright natural lighting balances the cheerful mood while keeping the overall color palette restrained according to the project tone" ], "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": "a84c690549d494fc", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "c07d975f30896f1a", "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": [ "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": "Travel parallel at torso height beside Su-ri-young in a medium-distance profile, timing the frame to one pedal's downstroke. Keep her smiling face and earphones in the upper center, her working leg and bicycle in the lower half, and leave forward space on the right as her hair trails toward the left.", "lighting_mood": "Natural daytime ambient light keeps the energetic ride bright but restrained in color and contrast.", "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": "coastal road ahead", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "road ahead at screen right", "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", "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": "f55d53c202fa5eb6" }, { "scene_index": 7, "heading": "S#7. 유기견 보호원 / D", "beat_title": "순한 양이 된 맹견", "representative_moment": "사납게 짖던 로트와일러가 기세가 꺾인 채 바닥에 얌전히 엎드려 있는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low floor-height observer, close framing with low three-quarter view, restrained daytime ambient lighting", "t2i_prompt": "Photorealistic cinematic still. A Rottweiler dog showing short black fur, a broad heavy head, a short thick muzzle, and pinned-back triangular ears lies prone in the middle-center midground. Captured by a low observer at floor height in a low three-quarter view from the left, the dog rests with its forelegs extended across the lower frame, looking down quietly. [L06: Restrained natural daytime ambient light maintains subdued brightness and low contrast, casting soft, muted tones across the floor, while the near side of the enclosure boundary runs along one edge of the frame between the camera and the dog's confined space.]", "outfit_assignments": [], "source_facts": [ "S#7 action: 로트와일러 기세가 수그러들며 순한 양이 됨", "shot description: 로트와일러가 귀를 뒤로 젖힌 채 얌전히 바닥에 엎드린 정지 상태" ], "visual_inferences": [ "fixed_element shelter_ambient_lighting: restrained daytime ambient light with subdued brightness", "Camera is positioned outside the enclosure looking in, keeping the barrier edge visible at floor height", "Entity traits (short black fur, broad heavy head, thick muzzle) applied directly as generic descriptors based on subject policy" ], "creative_decisions": [ "Floor-height camera angle emphasizes the dog's lowered, submissive posture in close framing", "Enclosure edge anchored at the side of the frame provides spatial context without requiring the full absent background" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "85f2848f5af20660", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "c5c333671f2bcba7", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 3, "visible_entities": [ "C12", "L06" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 7, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Just outside the enclosure boundary near floor height, let the forward move resolve into a locked close hold on the lowered Rottweiler. Frame the dog in a low three-quarter view across the central and lower portions, with its forelegs extended and ears pinned back while the enclosure edge remains visible at one side for spatial context.", "lighting_mood": "Natural daytime ambient light appropriate to the shelter preserves subdued contrast and the dog's dark form without theatrical effects.", "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": "로트와일러", "screen_zone": "middle_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", "L06" ], "allowed_outlook_pairs": [ { "character_id": "C12", "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": "C12", "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": 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": "shelter_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the animal shelter is illuminated by restrained, natural daytime ambient light. The environment maintains subdued brightness and low contrast throughout the space, completely devoid of any harsh or theatrical lighting. The background features soft, muted tones that unify the overall atmosphere of the shelter.", "applies_to_shots": [ 3, 6 ], "element_scope": "full", "source_facts": [ "Natural daytime ambient light appropriate to the shelter preserves subdued contrast", "Restrained daytime ambient light with subdued brightness" ], "visual_inferences": [ "The light is likely diffused through windows or simple overhead fixtures, creating a calm and uniformly slightly dim atmosphere suitable for both humans and resting animals." ], "creative_decisions": [ "Keeping the background gently out of focus (soft context) across different subjects to maintain a cohesive spatial feeling without distracting from the main 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": [ { "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": "10af8234d9eafed1" }, { "scene_index": 7, "heading": "S#7. 유기견 보호원 / D", "beat_title": "민숙과 연락이 닿지 않아 의아해하는 수리영", "representative_moment": "C01이 미간을 찌푸린 채 귀에 P01을 대고 응답 없는 전화를 기다린다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level close facial frame, shallow depth of field, restrained daytime ambient lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a young Korean woman, viewed at three-quarter angle from the left by an eye-level observer directly in front. Her face dominates the frame, furrowed brow and slightly compressed mouth centered. She holds a fixed listening posture, eyes looking off-screen. P01, an ordinary handheld smartphone, is pressed tightly to her ear, its narrow side and rear edge facing the camera. [: softly blurred interior of the daytime animal shelter, featuring out-of-focus background surfaces at the frame edges]. Restrained natural daytime ambient light casts subdued brightness and low-key facial contrast, enhancing the soft muted tones of the environment.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "shot description: 수리영이 귀에 스마트폰을 댄 채 미간을 찌푸린 얼굴", "S#7 action: 민숙에게 연락을 취해보지만, 받지 않는다." ], "visual_inferences": [ "fixed_element shelter_ambient_lighting: restrained natural daytime ambient light, subdued brightness, low contrast", "P01 held steadily against the ear without obscuring the face, conveying the tension of an unanswered call" ], "creative_decisions": [ "Close facial composition to focus heavily on C01's furrowed brow and compressed mouth, isolating her from the shelter background using a shallow depth of field.", "Positioning P01 to expose its narrow side and rear edge to the camera, aligning with the eye-level frontal perspective while keeping the facial expression fully visible." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "9626485060c1189b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "75f85e841c4e7589", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 7, "dependency_reason": "Same location, utilizing the continuous ambient lighting and mood established in Shot 3 while pushing in for a close frame.", "_shot_index": 6, "visible_entities": [ "C01", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 7, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Finish the stage’s eye-level dolly-in directly in front of 수리영, holding a close facial frame with her furrowed brow and slightly compressed mouth occupying the center while the smartphone remains narrow against one side of her face. Keep the shelter context soft at the frame edges, letting the unanswered call register through her fixed listening posture rather than a change of angle.", "lighting_mood": "Restrained daytime ambient light with subdued brightness and low-key facial contrast.", "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", "P01" ], "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": "shelter_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the animal shelter is illuminated by restrained, natural daytime ambient light. The environment maintains subdued brightness and low contrast throughout the space, completely devoid of any harsh or theatrical lighting. The background features soft, muted tones that unify the overall atmosphere of the shelter.", "applies_to_shots": [ 3, 6 ], "element_scope": "full", "source_facts": [ "Natural daytime ambient light appropriate to the shelter preserves subdued contrast", "Restrained daytime ambient light with subdued brightness" ], "visual_inferences": [ "The light is likely diffused through windows or simple overhead fixtures, creating a calm and uniformly slightly dim atmosphere suitable for both humans and resting animals." ], "creative_decisions": [ "Keeping the background gently out of focus (soft context) across different subjects to maintain a cohesive spatial feeling without distracting from the main action." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 7, "shot_index": 3, "ref_usage": "", "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": "3a22e862a3a260db" }, { "scene_index": 8, "heading": "S#8. 번화가·대형 마트 / 해질 무렵", "beat_title": "엄마의 부재와 직원의 전언", "representative_moment": "마트 진열대 앞에서 수리영(C01)이 엄마가 조퇴했다는 여직원의 말을 듣고 살짝 입을 벌린 채 당황하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Close side-on two-shot at shoulder height, shallow depth of field, muted dusk lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a Korean young woman in her early 20s with long black hair, an oval young face, soft facial features and natural skin tone, holds the left-center of the frame at shoulder height. She is in right profile, leaning slightly toward the employee across the shared aisle space from each other, eyes fixed on the employee, her slightly open mouth visible. At the right edge, a Korean middle-aged woman in worker uniform stands in left profile, her shoulder and side of face partially visible in soft focus, weight shifted toward C01O01, eyes fixed on the young woman. [L08: stocked supermarket shelves receding obliquely in the soft background, illuminated by overhead store lights balanced against muted twilight spilling from the storefront, with subdued brightness and restrained contrast].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Scene time is explicitly set to dusk (해질 무렵).", "수리영 is the main subject looking at the female employee.", "The employee is wearing a store uniform and standing opposite." ], "visual_inferences": [ "fixed_element su_ri_young_appearance: casual attire appropriate for cycling applied to C01O01", "fixed_element dusk_time_lighting: muted twilight tones mixing with store lights in the background", "Employee rendered in soft focus at the right edge to strictly satisfy the close framing scale constraint" ], "creative_decisions": [ "Positioned the camera at shoulder height to focus on C01's facial expression in clear profile", "Used a shared aisle space as an explicit fg/bg shared anchor to connect the two interacting figures", "Maintained low-key contrast and subdued brightness to reflect the tense occult thriller tone" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "automatic door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "carton", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "checkout counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gondola shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "produce display", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "refrigerated case", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "service counter", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "0cabaf5882d8737c", "owned_hash": "f09214d4c2733b3b", "camera_direction_hash": "ab26fc5d6b5c7975", "owned_usage_hash": "e63172fb259451b4", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "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": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At shoulder height beside the pair, complete the dolly-in on a close side-on two-shot: 수리영 holds the left-center in clear profile with her slightly open mouth visible, while 한국인 여성 직원 remains opposite at the right edge. Preserve their conversational axis and use restrained depth separation so 수리영's confusion carries the focus without removing the employee from the exchange.", "lighting_mood": "Store illumination is balanced against muted dusk color beyond the market, with subdued brightness and restrained contrast.", "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": 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": "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": "su_ri_young_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman maintaining a consistent physical appearance across the sequence. She wears the same everyday casual attire suitable for riding a bicycle and running errands. Her hairstyle, facial features, and clothing remain completely unchanged as she transitions from the interior of the store to the exterior street.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "수리영 is the main subject in both the interior and exterior shots.", "The scene happens sequentially as she directly walks out of the mart." ], "visual_inferences": [ "She is wearing practical, casual clothes appropriate for cycling and neighborhood mobility." ], "creative_decisions": [ "Her attire is kept visually distinct from the store employees' uniforms to clearly mark her as a visitor." ], "confidence": "high" }, { "element_id": "dusk_time_lighting", "element_type": "environment_state", "character_name": "", "description": "The ambient lighting consistently reflects an early evening, dusk setting. Muted twilight tones are visible in the surrounding atmosphere, establishing a continuous time frame. This natural dimming light consistently contrasts with the bright artificial illumination from the store's interior and the street neon signs.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Scene time is explicitly set to dusk (해질 무렵).", "Shot 3 mentions muted dusk color beyond the market.", "Shot 5 mentions dusk street ambience mixing with neon signs." ], "visual_inferences": [ "The sky and exterior background hold deep twilight colors that spill into the visible exterior spaces of both shots." ], "creative_decisions": [ "Maintaining a consistent low-key color palette between the exterior backdrop of the indoor shot and the primary environment of the outdoor shot." ], "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": "af0b4bd4550ef4be" }, { "scene_index": 8, "heading": "S#8. 번화가·대형 마트 / 해질 무렵", "beat_title": "연락 두절과 귀가 결심", "representative_moment": "마트 밖으로 나온 C01이 고개를 숙이고 손에 든 P01 스마트폰 화면을 확인하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium upper-body frame at chest height, tilted slightly downward, muted dusk low-key lighting", "t2i_prompt": "Photorealistic cinematic still. A medium upper-body frame at chest height, tilted slightly downward. C01O01 in a khaki sweatshirt, a young Korean woman, is positioned in the middle-left at three-quarter angle from the right. She is slightly hunched over, looking down with eyes fixed on the P01 smartphone held in her hands in the lower-center frame at a natural hand-held scale. The front display of the P01 smartphone is angled upward toward her, remaining partly visible to the camera. The brightly illuminated sliding door and window of the storefront recede behind her, opening negative space. An explicitly lit sign glows against the muted twilight tones, as the dusk street ambience mixes with the artificial light in a low-key palette.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영 마트 밖을 나오면서 연락을 취해보지만 받지않는다.", "마트 밖으로 나온 수리영이 손에 든 스마트폰 액정을 내려다보는 상체", "해질 무렵" ], "visual_inferences": [ "fixed_element dusk_time_lighting: dusk street ambience mixing with store interior lights", "fixed_element su_ri_young_appearance: casual attire maintained from the previous shot" ], "creative_decisions": [ "chest-height medium framing tilting downward to anchor the downward gaze and phone in the lower-center frame", "receding storefront on one side to open negative space and emphasize isolation", "muted twilight palette contrasting with bright artificial background elements" ], "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": "awning", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bench", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cart rack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor mat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "poster", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shopping cart", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign", "usage_kind": "anchor", "source_phrase": "explicitly lit sign glows against the muted twilight tones" }, { "owned_token": "sliding door", "usage_kind": "anchor", "source_phrase": "brightly illuminated sliding door and window" }, { "owned_token": "trash can", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "brightly illuminated sliding door and window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "311bb268189a40db", "owned_hash": "174d81985ee583f6", "camera_direction_hash": "84eff83885c74c66", "owned_usage_hash": "b26ce2e6fa99c728", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "sign", "violating_phrase": "explicitly lit sign glows against the muted twilight tones", "reason": "The sign is referenced as a background lighting anchor to establish the ambient mood, with no instruction to redraw its structure, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "sliding_door", "violating_phrase": "brightly illuminated sliding door and window", "reason": "The sliding door is used as a spatial background anchor receding behind the character to frame the scene, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "brightly illuminated sliding door and window", "reason": "The window is mentioned alongside the sliding door as a compositional background anchor, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 8, "dependency_reason": "Shot 8_3 location reference (exact_background) applied to maintain the same exterior store environment and continuous dusk lighting.", "_shot_index": 5, "visible_entities": [ "C01", "L09", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 8, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Outside the store at chest height, retreat gently while maintaining a medium upper-body frame and tilting slightly downward along 수리영's lowered gaze. Place her in the middle-left and the smartphone in the lower-center at a natural hand-held scale, with the storefront receding behind her as the dolly-out opens a little negative space.", "lighting_mood": "Dusk street ambience mixes with the explicitly lit neon signs in a muted, low-key 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", "L09", "P01" ], "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": "L08B02", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "awning", "bench", "bollard", "cart rack", "floor mat", "poster", "shopping cart", "sign", "sliding door", "trash can", "window" ], "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": [ { "element_id": "su_ri_young_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman maintaining a consistent physical appearance across the sequence. She wears the same everyday casual attire suitable for riding a bicycle and running errands. Her hairstyle, facial features, and clothing remain completely unchanged as she transitions from the interior of the store to the exterior street.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "수리영 is the main subject in both the interior and exterior shots.", "The scene happens sequentially as she directly walks out of the mart." ], "visual_inferences": [ "She is wearing practical, casual clothes appropriate for cycling and neighborhood mobility." ], "creative_decisions": [ "Her attire is kept visually distinct from the store employees' uniforms to clearly mark her as a visitor." ], "confidence": "high" }, { "element_id": "dusk_time_lighting", "element_type": "environment_state", "character_name": "", "description": "The ambient lighting consistently reflects an early evening, dusk setting. Muted twilight tones are visible in the surrounding atmosphere, establishing a continuous time frame. This natural dimming light consistently contrasts with the bright artificial illumination from the store's interior and the street neon signs.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Scene time is explicitly set to dusk (해질 무렵).", "Shot 3 mentions muted dusk color beyond the market.", "Shot 5 mentions dusk street ambience mixing with neon signs." ], "visual_inferences": [ "The sky and exterior background hold deep twilight colors that spill into the visible exterior spaces of both shots." ], "creative_decisions": [ "Maintaining a consistent low-key color palette between the exterior backdrop of the indoor shot and the primary environment of the outdoor shot." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 8, "shot_index": 3, "ref_usage": "", "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": "31282a12375195d3" }, { "scene_index": 9, "heading": "S#9. 해안가 / N", "beat_title": "어두운 밤 해안도로를 질주하는 수리영", "representative_moment": "어두운 해안도로 위, 수리영(C01)이 자전거 안장에 앉아 상체를 앞을 향해 기울인 채 페달을 밟아 내리는 중간 자세의 실루엣 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Tracking wide side profile at bicycle-seat height, low-key night silhouette", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a young Asian woman riding the bicycle in the middle-center midground, moving toward the coastal road ahead in the middle-right background. Captured from an observer tracking perspective at bicycle-seat height beside and slightly ahead, holding a wide side profile. She is in right profile with her torso pitched forward, eyes looking off-screen to the right, frozen in the exact midpoint of the pedal downstroke. One knee is bent in the down stroke, her long black hair and clothing trailing back in the direction opposite to her motion. The bicycle's side profile, including its frame and wheels, is clearly visible across the frame. Low-key nighttime illumination renders the figure and the bicycle primarily as a subdued, rim-lit silhouette against the dark coastal road environment, preserving body volume with faint highlights catching the edges to establish the feeling of speed.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "scene text: 수리영 riding a bicycle hastily home along the coastal road at night", "shot description: silhouette of 수리영 on the bicycle saddle, torso pitched forward, mid-pedal downstroke, clothes trailing" ], "visual_inferences": [ "tracking camera at bicycle-seat height to capture the full pedal action and speed", "low-key rim lighting to preserve body volume within the night silhouette" ], "creative_decisions": [ "wide side profile framing to include the entire bicycle body as required by the large prop rule", "positioning the character in the middle-center moving toward the middle-right to satisfy the spatial contract" ], "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": "cc42bfb55f0c0201", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "679526b95516e320", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 6, "dependency_reason": "previous_shot_refs indicates Scene 6 Shot 1 as the location reference for the coastal road environment", "_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": "Track at bicycle-seat height beside and slightly ahead of 수리영, holding a wide side profile as she travels toward screen right along the coastal road. Capture the exact midpoint of the pedal downstroke, with her forward-pitched torso and trailing clothing forming one driving silhouette while enough road remains visible to establish speed and direction.", "lighting_mood": "Low-key nighttime illumination renders 수리영 and the bicycle primarily as a subdued silhouette against the coastal-road setting.", "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": "coastal road ahead", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "coastal road ahead at screen right", "label": "수리영 riding the 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": "", "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": "a917d73ab2c43c11" }, { "scene_index": 10, "heading": "S#10. 옥탑방 / N", "beat_title": "옥탑방 침입", "representative_moment": "발목 높이의 카메라가 살짝 열린 현관문 틈새로 낯선 남자의 신발 앞코가 어둠 속으로 들어가는 순간을 비춘다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Ankle-height insert shot with voyeuristic low-key contrast.", "t2i_prompt": "Photorealistic cinematic still. A narrow entrance-door gap is visible in the middle-center foreground. Captured from an ankle-height low observer in a voyeuristic perspective, the slightly opened iron door forms two close vertical frame edges around a narrow view of the pitch-dark interior. A Korean man's shoe toe, stepping forward with his back to the camera implied, is placed low in the frame just inside the threshold. The explicitly dark interior is held in deep low-key contrast against the readable, weathered edges of the iron door. The adjacent weathered exterior wall of the old rooftop room features a rough, aged texture dimly lit by ambient night light. The man's body and face are entirely withheld beyond the upper frame edge, focusing the suspense solely on the foot crossing into the darkness.", "outfit_assignments": [], "source_facts": [ "Scene text: The door opens slightly, the man goes inside.", "Shot description: 현관 철문이 아주 살짝 열린 틈새로 보이는 어두운 실내와 신발 앞코", "DP direction: At ankle height, complete the dolly-in until the slightly opened iron entrance door forms two close vertical frame edges around a narrow view of the dark interior." ], "visual_inferences": [ "fixed_element rooftop_exterior_wall: The weathered exterior wall of an old rooftop room at night. The surface is textured and dimly lit in deep low-key contrast, serving as the consistent architectural structure.", "An ankle-height low camera angle naturally places the shoe toe in the lower frame while implying the rest of the body is withheld beyond the upper frame edge, enforcing a suspenseful voyeuristic perspective." ], "creative_decisions": [ "Deep low-key lighting heavily shadows the interior to create a stark contrast with the readable, weathered edges of the slightly opened iron door.", "Limiting the visible elements to just the door gap and the shoe toe maximizes the sense of an unseen threat entering the enclosed space." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dried herb leaf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "flyer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "iron door", "usage_kind": "anchor", "source_phrase": "the slightly opened iron door" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "the old rooftop room" }, { "owned_token": "steel staircase", "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": "a3be1e691248161c", "owned_hash": "345a622acf84cac2", "camera_direction_hash": "8815dd49d3d0eea0", "owned_usage_hash": "8dffc77d45644df7", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "iron door", "violating_phrase": "the slightly opened iron door", "reason": "The iron door is used as a compositional frame and anchor for the perspective, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "rooftop room", "violating_phrase": "the old rooftop room", "reason": "The rooftop room is referenced to describe ambient night lighting on its exterior wall as a background anchor, which matches the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 4, "dependency_reason": "Same rooftop location (L03) as established in S4, utilizing exact_background to maintain consistent architecture.", "_shot_index": 6, "visible_entities": [ "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 10, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At ankle height, complete the dolly-in until the slightly opened iron entrance door forms two close vertical frame edges around a narrow view of the dark interior. Keep the shoe toe low and just inside the gap at natural scale, withholding the man's face and body so the threshold mechanism carries the suspense.", "lighting_mood": "The explicitly dark interior is held in deep low-key contrast against the readable edges of the slightly opened door.", "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": "narrow entrance-door gap", "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": [ "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": "L03B03", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "dried herb leaf", "drying platform", "flyer", "iron door", "parapet", "rooftop deck", "rooftop room", "steel staircase", "water tank", "window" ], "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": [ { "element_id": "rooftop_exterior_wall", "element_type": "environment_state", "character_name": "", "description": "The weathered exterior wall of an old rooftop room at night. The surface is textured and dimly lit in deep low-key contrast, serving as the consistent architectural structure that surrounds both the iron entrance door and the adjacent window.", "applies_to_shots": [ 6, 7 ], "element_scope": "full", "source_facts": [ "Setting is the exterior of an old multi-family villa rooftop and rooftop room at night.", "Shot 6 features the iron entrance door.", "Shot 7 features a controlled margin of the rooftop wall around the lit window." ], "visual_inferences": [ "The exterior wall shares the same structural material and aged appearance across the door and window areas." ], "creative_decisions": [ "Using rough, aged concrete or brick textures for the wall to emphasize the dilapidated atmosphere of the old rooftop." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 4, "shot_index": 1, "ref_usage": "", "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": "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": "8eea6df98f5fcc26" }, { "scene_index": 10, "heading": "S#10. 옥탑방 / N", "beat_title": "창문에 비친 끔찍한 실루엣", "representative_moment": "불 꺼진 옥탑방 외벽 가운데, 불 켜진 창문 너머로 거대한 남자가 민숙을 덮쳐 어깨를 뜯어먹는 기괴한 그림자가 비친다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "static straight-on wide composition from outside at window height, stark low-key contrast", "t2i_prompt": "Photorealistic cinematic still. The lit rooftop-room window occupies the middle-center background, centering the wide composition. The weathered exterior wall of the old rooftop room surrounds the window with a dim, textured margin in the night. Inside the illuminated window plane, two backlit figures appear in stark low-key contrast, their body volumes preserved against the interior glow while their faces remain entirely hidden in deep shadow. A large East Asian male figure in right profile exerts forceful downward pressure, his head bent sharply toward the other's shoulder. A smaller East Asian female figure in left profile is held in place beneath him, her shoulders recoiling as she faces off-screen. The violent contact is captured as a single frontal tableau, isolated by the restrained darkness of the exterior.", "outfit_assignments": [], "source_facts": [ "S#10 text: 창문에 비친 그림자. 민숙을 덮쳐 어깨를 물어뜯는 듯한 모습", "shot staging: From outside at window height, hold a static, straight-on wide composition with the lit window centered and a controlled margin of rooftop wall around it.", "character angles: large figure is in right profile with head bent toward shoulder, smaller figure is in left profile recoiling." ], "visual_inferences": [ "fixed_element rooftop_exterior_wall: The weathered exterior wall surrounds the window, providing architectural consistency.", "Silhouettes possess depth and volume (body volumes preserved), avoiding a flat cutout appearance while hiding facial details.", "Used violence guideline wording ('exerts forceful downward pressure', 'held in place') to accurately depict the power imbalance of the attack without explicit gore." ], "creative_decisions": [ "Framed the two figures as a unified tableau within the lit window, utilizing stark low-key contrast to isolate the terrifying action against the dark exterior." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dried herb leaf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "flyer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "iron door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "old rooftop room" }, { "owned_token": "steel staircase", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "lit rooftop-room window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d77f7b05a7aec147", "owned_hash": "345a622acf84cac2", "camera_direction_hash": "b6e5cc8c05e1ef02", "owned_usage_hash": "9e03b62f9c8480f7", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "rooftop room", "violating_phrase": "old rooftop room", "reason": "The rooftop room is referenced solely as a spatial frame to set the composition around the window, aligning with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "Inside the illuminated window plane, two backlit figures appear", "reason": "The illumination and the silhouettes shown inside the window are allowed visual state and content changes explicitly required by the shot_intent's depiction of the struggle.", "verdict": "allowed_visual_state_change" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 7, "visible_entities": [ "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 10, "shot_index": 7 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From outside at window height, hold a static, straight-on wide composition with the lit window centered and a controlled margin of rooftop wall around it. Inside the window plane, the 거대한 그림자 rises behind and above the 작은 그림자, its bent head meeting the smaller silhouette at shoulder level without revealing literal bodies or facial detail.", "lighting_mood": "The lit window isolates two deep silhouettes in stark low-key contrast, with restrained surrounding brightness.", "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": "background", "gesture_action": "none", "gesture_target_label": "", "label": "lit rooftop-room window", "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": [ "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": "L03B03", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "dried herb leaf", "drying platform", "flyer", "iron door", "parapet", "rooftop deck", "rooftop room", "steel staircase", "water tank", "window" ], "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": [ { "element_id": "rooftop_exterior_wall", "element_type": "environment_state", "character_name": "", "description": "The weathered exterior wall of an old rooftop room at night. The surface is textured and dimly lit in deep low-key contrast, serving as the consistent architectural structure that surrounds both the iron entrance door and the adjacent window.", "applies_to_shots": [ 6, 7 ], "element_scope": "full", "source_facts": [ "Setting is the exterior of an old multi-family villa rooftop and rooftop room at night.", "Shot 6 features the iron entrance door.", "Shot 7 features a controlled margin of the rooftop wall around the lit window." ], "visual_inferences": [ "The exterior wall shares the same structural material and aged appearance across the door and window areas." ], "creative_decisions": [ "Using rough, aged concrete or brick textures for the wall to emphasize the dilapidated atmosphere of the old rooftop." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 10, "shot_index": 6, "ref_usage": "", "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": "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": "40ebdf88bcde6a67" }, { "scene_index": 11, "heading": "S#11. 옥탑 방 밖 / N", "beat_title": "역겨운 냄새에 멈칫하는 수리영", "representative_moment": "계단 중간에 서서 역겨운 냄새에 코를 찡그린 수리영(C01)의 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close three-quarter portrait at face height, low-key subdued nighttime ambient lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a Korean young woman in her 20s with long black hair, an oval young face, soft facial features, and a natural skin tone, seen at three-quarter angle from the right. Her gaze is fixed off-screen, nose pinched and brow furrowed in a distressed reaction, her climb stalled. One hand enters the lower frame edge, braced tightly on the pipe railing of [L11: an added exterior steel stair with thin steel-plate stringers, support posts, and open treads falling away into the deep shadow behind her]. Low-key, subdued nighttime ambient lighting envelops the outdoor space, casting deep shadows that conceal the surrounding background details and maintaining a dark atmosphere without any unmotivated artificial light sources. The close framing focuses entirely on her face and the frame-edge railing.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#11 action: '갑자기 훅- 느껴지는 역겨운 냄새. 뭐지?'", "shot description: 철제 파이프 난간을 손으로 잡은 수리영이 코끝을 찡그린 채 계단 중간에 서 있는 얼굴 클로즈업" ], "visual_inferences": [ "fixed_element nighttime_ambient_lighting: Low-key, subdued nighttime ambient lighting envelops the outdoor space, casting deep shadows.", "The background stair structure features open treads and thin steel plates as explicitly stated in the creator corrections, avoiding any concrete appearance." ], "creative_decisions": [ "Used a tight close framing at face height to isolate her distressed expression, with her hand entering at the bottom edge to anchor her physical interaction with the stairs.", "The deep shadows in the out-of-focus background emphasize the claustrophobic and tense atmosphere created by the low-key lighting." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "column", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "railing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "staircase", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "65f56205fcfb27ab", "owned_hash": "375bc23e2e583d54", "camera_direction_hash": "0bdea75885052eba", "owned_usage_hash": "00cdd3413f0491ef", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 4, "visible_entities": [ "C01", "L11" ], "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": "At face height on the added exterior stair, finish the dolly-in from slightly beside 수리영 into a close three-quarter portrait, placing her pinched nose and furrowed expression near center while one hand enters at the lower edge around the pipe railing. The open stair structure falls away behind her, and the frame holds at the instant her hurried climb stalls in reaction to the smell.", "lighting_mood": "Low-key nighttime ambient light keeps the close reaction legible without introducing an unmotivated source or color.", "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": "skipped_close_framing", "bg_id": "L11B01", "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": "nighttime_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "Low-key, subdued nighttime ambient lighting envelops the outdoor space. The illumination is minimal, casting deep shadows and maintaining a dark atmosphere without any bright or unmotivated artificial light sources.", "applies_to_shots": [ 4, 5 ], "element_scope": "full", "source_facts": [ "옥탑 방 밖 / N", "Low-key nighttime ambient light keeps the close reaction legible without introducing an unmotivated source or color.", "Subdued nighttime ambient light preserves the dark doorway as the principal low-key threat." ], "visual_inferences": [ "The minimal lighting creates deep shadows that conceal the surrounding background details." ], "creative_decisions": [ "Maintaining a consistent low-key exposure across different outdoor areas (stairs and rooftop) to heighten suspense and focus on the character's reaction." ], "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": "616edb38b5dcf599" }, { "scene_index": 11, "heading": "S#11. 옥탑 방 밖 / N", "beat_title": "불길한 예감과 열린 철문", "representative_moment": "옥상에 도착한 수리영(C01)이 멈춰 서서 반쯤 열려 있는 어두운 현관 철문을 주시하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide over-the-shoulder shot with subdued nighttime ambient lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, an East Asian young woman with long black hair, occupies the lower-left foreground seen from over the shoulder with back to camera, paused after her final step and looking toward the half-open iron entrance door in the middle-center background. The entrance-facing side of the door is square to the camera, with the open leaf revealing part of the pitch-black interior of the rooftop room. Low-key, subdued nighttime ambient lighting envelops the outdoor space, casting deep shadows across the concrete landing and the clear space separating the figure from the entrance. The minimal illumination preserves the dark doorway as the principal low-key threat.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "옥상에 올라와 보니 반쯤 열린 현관 철문.", "Settle at eye level slightly behind and to one side of 수리영, aligning the lens exactly along her eyeline so the half-open iron entrance door is seen straight on in the background." ], "visual_inferences": [ "fixed_element nighttime_ambient_lighting: Low-key, subdued nighttime ambient lighting envelops the outdoor space, casting deep shadows.", "Her posture is tense and paused, registering the awful smell and the ominous open door." ], "creative_decisions": [ "Utilized an over-the-shoulder wide framing to anchor the character in the lower-left foreground, establishing a clear line of sight to the centered entrance door to heighten spatial tension." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "column", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "the half-open iron entrance door in the middle-center background" }, { "owned_token": "landing", "usage_kind": "anchor", "source_phrase": "across the concrete landing" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "railing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "the pitch-black interior of the rooftop room" }, { "owned_token": "staircase", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "water tank", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "4603a9ad41afd8fa", "owned_hash": "375bc23e2e583d54", "camera_direction_hash": "ea036d85df09db27", "owned_usage_hash": "823b4400923d67a1", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "the half-open iron entrance door in the middle-center background", "reason": "The prompt describes the door's half-open state and alignment to establish the spatial composition, which correctly aligns with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "landing", "violating_phrase": "across the concrete landing", "reason": "The landing is referenced solely as a ground surface where shadows are cast, confirming its role as a spatial anchor in agreement with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "rooftop room", "violating_phrase": "the pitch-black interior of the rooftop room", "reason": "The rooftop room is mentioned only to define the dark interior space in the background, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 11, "dependency_reason": "동일 씬 내 이전 샷(Shot 4)의 옥상 배경(L11B01)과 야간 조명 상태(nighttime_ambient_lighting)를 유지하며 앵글을 이동", "_shot_index": 5, "visible_entities": [ "C01", "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 11, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Settle at eye level slightly behind and to one side of 수리영, aligning the lens exactly along her eyeline so the half-open iron entrance door is seen straight on in the background. Her shoulder and paused body occupy the lower-left foreground while the doorway sits centered with clear rooftop space separating them.", "lighting_mood": "Subdued nighttime ambient light preserves the dark doorway as the principal low-key threat.", "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": "half-open iron entrance door", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "looks_toward", "gesture_target_label": "half-open iron entrance door", "label": "수리영", "screen_zone": "lower_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", "L03" ], "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": "L11B01", "reference_usage": "exact_background", "owned_objects": [ "column", "door", "landing", "parapet", "railing", "rooftop room", "staircase", "wall light", "water tank" ], "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": [ { "element_id": "nighttime_ambient_lighting", "element_type": "environment_state", "character_name": "", "description": "Low-key, subdued nighttime ambient lighting envelops the outdoor space. The illumination is minimal, casting deep shadows and maintaining a dark atmosphere without any bright or unmotivated artificial light sources.", "applies_to_shots": [ 4, 5 ], "element_scope": "full", "source_facts": [ "옥탑 방 밖 / N", "Low-key nighttime ambient light keeps the close reaction legible without introducing an unmotivated source or color.", "Subdued nighttime ambient light preserves the dark doorway as the principal low-key threat." ], "visual_inferences": [ "The minimal lighting creates deep shadows that conceal the surrounding background details." ], "creative_decisions": [ "Maintaining a consistent low-key exposure across different outdoor areas (stairs and rooftop) to heighten suspense and focus on the character's reaction." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 11, "shot_index": 4, "ref_usage": "", "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": "L11B01", "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": "73903b1a68346095" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "참혹하게 쓰러진 엄마 발견", "representative_moment": "어두운 방 안 커튼 뒤로, 붉은 피 분장을 한 채 바닥에 고개를 떨구고 쓰러진 C21을 발견하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium framing from seated head height, looking through parted foreground curtains into a dim, low-key lit room with deep crimson accents", "t2i_prompt": "Photorealistic cinematic still. At seated head height, the view is framed by the existing parted curtain edges in the foreground. C21O13 in burgundy homewear, a Korean woman in her 40s, at three-quarter angle from the left, is slumped seated on the floor, looking down, motionless and lifeless. Her head is dropped deeply forward, the body collapsed against the floor. Dark red stage paint covers her, with torn shoulder and collarbone special-effects makeup visible. One hand remains tightly closed around a crumpled photograph, resting against her leg. The interior space is engulfed in a low-key nighttime atmosphere with faint, subdued ambient lighting, heavy shadows muting natural colors. She is positioned on the floor near the pre-existing bloodstain.", "outfit_assignments": [ { "character_id": "C21", "outlook_id": "O13" } ], "source_facts": [ "S#12 action: 호흡을 가두고 커튼을 젖히자 앉은 채로 고개를 떨군 채 움직이지 않는 엄마- 민숙!!!!", "S#12 action: 분장 표면이 뜯긴 어깨와 쇄골, 마치 동물에게 물어뜯긴 듯한 거친 상처가 고스란히 드러나있다.", "S#12 action: 엄마 손안에 구겨진 사진 한 장이 보인다.", "shot description: 옷과 피부가 피투성이가 된 민숙이 바닥에 주저앉아 고개를 푹 숙인 정지 상태" ], "visual_inferences": [ "fixed_element: low_key_night_interior_lighting (faint ambient light, heavy shadows)", "The parted curtain acts as a foreground framing device to reveal the subject", "Immobilized subject continuity requires the hand tightly closing around the crumpled photograph", "The dark red stage paint substitutes for the scenario's blood to adhere to safety guidelines while maintaining visual impact" ], "creative_decisions": [ "Placing the camera at seated head height just behind the curtains to execute the 'observer' POV and constrained reveal requested by the DP", "Using a bloodstain pool around her to ground the figure on the floor and provide the deep crimson visual contrast specified in the lighting mood" ], "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": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bloodstain", "usage_kind": "anchor", "source_phrase": "she is positioned on the floor near the pre-existing bloodstain" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "circle mark", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "the view is framed by the existing parted curtain edges in the foreground" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "footprint", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "anchor", "source_phrase": "One hand remains tightly closed around a crumpled photograph" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "teacup", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "8abd79b4621296dc", "owned_hash": "daee96141155af1e", "camera_direction_hash": "d0785e9dbe712eb6", "owned_usage_hash": "3ef30e98037c1f82", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "curtain", "violating_phrase": "the view is framed by the existing parted curtain edges in the foreground", "reason": "The prompt uses the existing curtain as a foreground framing element, functioning purely as a spatial anchor, which matches the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "photograph", "violating_phrase": "One hand remains tightly closed around a crumpled photograph", "reason": "The character is physically holding the photograph, with the phrase describing the hand's tight grip (physical interaction), and the 'crumpled' state is a physical deformation allowed under prop contact rather than a redesign.", "verdict": "allowed_prop_contact" }, { "owned_object": "bloodstain", "violating_phrase": "She is positioned on the floor near the pre-existing bloodstain", "reason": "The prompt uses the pre-existing bloodstain as a spatial anchor for the character's positioning, without any modification or redraw requested.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "동일한 옥탑방 내부 배경 참조(S5_Shot9)와 환경 조명 분위기 연속성 유지", "_shot_index": 11, "visible_entities": [ "C21", "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 11 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Just behind the opened curtain at seated head height, advance only to a medium distance and then hold on 민숙 alone, centered low in the frame with her head dropped and her blood-covered body collapsed against the floor. Keep the torn shoulder and collarbone readable without pushing into an exploitative detail shot, while the parted curtain edges provide a restrained reveal frame.", "lighting_mood": "Low-key nighttime interior ambience holds the room in subdued darkness while deep crimson blood remains the restrained visual accent.", "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": [ "C21", "L04" ], "allowed_outlook_pairs": [ { "character_id": "C21", "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": "L04B05", "reference_usage": "exact_background", "owned_objects": [ "TV", "TV stand", "bed", "bloodstain", "chair", "circle mark", "counter", "curtain", "dining table", "door", "drying platform", "floor lamp", "footprint", "photograph", "sink", "stovetop", "teacup", "window" ], "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": [ { "element_id": "low_key_night_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior space is engulfed in a low-key nighttime atmosphere with faint, subdued ambient lighting. Heavy shadows dominate the room, obscuring background details and muting natural colors. This restrained dark environment serves to starkly contrast with any deep crimson elements present in the scene.", "applies_to_shots": [ 11, 18, 21 ], "element_scope": "full", "source_facts": [ "희미한 거실 스탠드 불빛", "Low-key nighttime interior ambience holds the room in subdued darkness", "Subdued low-key interior light", "Subdued low-key night-interior illumination" ], "visual_inferences": [ "The primary light source is weak and distant, leaving the immediate area in heavy, oppressive shadow." ], "creative_decisions": [ "Consistent use of subdued darkness across all shots to maintain visual continuity and a horror atmosphere.", "Using the dim environment specifically as a canvas to make deep red visual accents stand out." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 9, "ref_usage": "", "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": "C21O13", "policy": "required" }, { "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": "8eb17ccca98ed6d4" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "손에 쥐어진 구겨진 사진", "representative_moment": "C21의 손에 꽉 쥐어진 구겨진 사진 P02가 어둠 속에서 타이트하게 보인다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low-angle insert shot isolating the photograph and fingers, shallow depth of field, subdued low-key lighting", "t2i_prompt": "Photorealistic cinematic still. Viewed from a low observer position at hand height, P02, a small, crumpled old black-and-white photographic print, occupies the center of the frame. It is held tightly by the motionless, bloodied, clenched fingers of C21O13 in dark purplish homewear, an East Asian middle-aged woman. The printed face of P02 is angled partly toward the camera, with its unknown image content unresolved in the heavy shadow. The prop maintains a natural, small scale within the grip, framed by the narrow ring of the fingers. [L04B05: Below the hand, a small patch of the dimly lit floor is visible, engulfed in subdued low-key night-interior shadow, muting natural colors and highlighting the deep crimson traces on the skin].", "outfit_assignments": [ { "character_id": "C21", "outlook_id": "O13" } ], "source_facts": [ "Shot 18: 민숙의 손에 꽉 쥐어진 구겨진 사진의 타이트 샷", "엄마 손안에 구겨진 사진 한 장이 보인다.", "Subdued low-key night-interior illumination holds the room in subdued darkness", "Only the tightly closed hand and the crumpled photograph are required in frame" ], "visual_inferences": [ "The printed face of the photograph is turned toward the lens but kept ambiguous by the heavy shadow.", "The fingers remain completely motionless in a rigid, tight grip.", "The patch of floor below remains minimally visible to provide spatial context without distracting from the prop." ], "creative_decisions": [ "Insert scale framing isolating just the hand and the prop against a very restricted background.", "Restrained dark palette to make the deep crimson traces on the hand stand out strongly against the low-key shadows." ], "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": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bloodstain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "circle mark", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "footprint", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "redraw", "source_phrase": "P02, a small, crumpled old black-and-white photographic print" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "teacup", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "a82e4e355d646d90", "owned_hash": "daee96141155af1e", "camera_direction_hash": "0c54ecc3563f588d", "owned_usage_hash": "a87404a0e80b336a", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "photograph", "violating_phrase": "It is held tightly by the motionless, bloodied, clenched fingers of C21O13, maintaining a natural, small scale within the grip, framed by the narrow ring of the fingers", "reason": "The prompt details the physical interaction geometry of holding the photograph (the grip and scale within the fingers) which is allowed under prop contact, and its crumpled state is required by the shot intent.", "verdict": "allowed_prop_contact" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "이전 샷 및 씬 5에서 확립된 민숙 방의 배경과 조명 분위기(exact_background) 연속성 유지", "_shot_index": 18, "visible_entities": [ "C21", "P02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 18 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From low beside 민숙 at hand height, complete the push into an insert that isolates the crumpled photograph while retaining a narrow ring of her gripping fingers and a small patch of the nearby floor. Angle the photograph's printed face partly toward the lens but keep its unknown image content unresolved, avoiding any enlargement beyond the hand's natural scale.", "lighting_mood": "Subdued low-key interior light emphasizes the creases of the photograph and the deep crimson traces on 민숙's bloodied 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": [ "C21", "P02" ], "allowed_outlook_pairs": [ { "character_id": "C21", "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": "L04B05", "reference_usage": "exact_background", "owned_objects": [ "TV", "TV stand", "bed", "bloodstain", "chair", "circle mark", "counter", "curtain", "dining table", "door", "drying platform", "floor lamp", "footprint", "photograph", "sink", "stovetop", "teacup", "window" ], "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": [ { "element_id": "low_key_night_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior space is engulfed in a low-key nighttime atmosphere with faint, subdued ambient lighting. Heavy shadows dominate the room, obscuring background details and muting natural colors. This restrained dark environment serves to starkly contrast with any deep crimson elements present in the scene.", "applies_to_shots": [ 11, 18, 21 ], "element_scope": "full", "source_facts": [ "희미한 거실 스탠드 불빛", "Low-key nighttime interior ambience holds the room in subdued darkness", "Subdued low-key interior light", "Subdued low-key night-interior illumination" ], "visual_inferences": [ "The primary light source is weak and distant, leaving the immediate area in heavy, oppressive shadow." ], "creative_decisions": [ "Consistent use of subdued darkness across all shots to maintain visual continuity and a horror atmosphere.", "Using the dim environment specifically as a canvas to make deep red visual accents stand out." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 9, "ref_usage": "", "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": "C21O13", "policy": "required" }, { "kind": "background", "id": "L04B05", "policy": "required" }, { "kind": "prop", "id": "P02", "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": [ 18 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P02", "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": "31dbff46172e8767" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "기괴한 표식의 발현", "representative_moment": "어두운 방 안, 침대 위쪽 벽면에 스스로 나타나기 시작한 기괴한 붉은 원형 표식.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "static uncanny observational close-up with an upward tilt", "t2i_prompt": "Photorealistic cinematic still. A half-formed red circular mark above the bed appears in the upper-center background. Captured from beside the bed at shoulder height with an upward tilt, the close frame isolates the [L04: bare camera-facing wall surface]. A narrow strip of the bed's side and upper edge angles across the lower frame edge, providing spatial reference. Engulfed in a low-key nighttime atmosphere with faint ambient lighting, heavy shadows dominate the restricted view, starkly contrasting the deep crimson of the partially formed mark against the restrained, muted natural colors of the wall.", "outfit_assignments": [], "source_facts": [ "침대 위 벽에 붉은 원이 생겨난다", "마치 누군가 붓으로 벽에 원으로 그리는 것처럼" ], "visual_inferences": [ "fixed_element low_key_night_interior_lighting: The interior space is engulfed in a low-key nighttime atmosphere with heavy shadows obscuring background details", "camera position: upward tilt from shoulder height explicitly framing the wall and the top edge of the bed" ], "creative_decisions": [ "static, uncanny observational composition tightly focused on the wall to emphasize the supernatural phenomenon", "subdued low-key night-interior illumination utilized specifically to isolate the deep crimson color of the mark" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bloodstain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "circle mark", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "footprint", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "teacup", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "8fb87d3c1a294bda", "owned_hash": "daee96141155af1e", "camera_direction_hash": "9ec9c9a6005adafa", "owned_usage_hash": "60c88bde619c39e2", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "연속성 유지: S12 이전 샷(Shot 11 등)의 어두운 야간 실내 조명(low-key night interior) 분위기와 그림자 상태 유지", "_shot_index": 21, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 21 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From beside the bed at shoulder height, hold an upward tilt toward the wall above the bed, with the half-formed red circle positioned in the upper center and occupying less than a third of the frame. Keep a narrow strip of the bed along the lower edge as spatial reference, using a static, uncanny observational composition.", "lighting_mood": "Subdued low-key night-interior illumination isolates the deep crimson mark against restrained natural colors.", "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": "half-formed red circular mark above the bed", "screen_zone": "upper_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" ], "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": "L04B05", "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": "low_key_night_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior space is engulfed in a low-key nighttime atmosphere with faint, subdued ambient lighting. Heavy shadows dominate the room, obscuring background details and muting natural colors. This restrained dark environment serves to starkly contrast with any deep crimson elements present in the scene.", "applies_to_shots": [ 11, 18, 21 ], "element_scope": "full", "source_facts": [ "희미한 거실 스탠드 불빛", "Low-key nighttime interior ambience holds the room in subdued darkness", "Subdued low-key interior light", "Subdued low-key night-interior illumination" ], "visual_inferences": [ "The primary light source is weak and distant, leaving the immediate area in heavy, oppressive shadow." ], "creative_decisions": [ "Consistent use of subdued darkness across all shots to maintain visual continuity and a horror atmosphere.", "Using the dim environment specifically as a canvas to make deep red visual accents stand out." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 11, "ref_usage": "", "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": [ { "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": "f289896c6e859767" }, { "scene_index": 13, "heading": "S#13. 옥탑방 외곽 / N", "beat_title": "공포에 질린 수리영", "representative_moment": "옥탑방 계단을 뛰어내려온 뒤 초점을 잃고 멍하니 서 있는 수리영의 모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level medium framing on the frontal axis, softly dollying in to a three-quarter portrait under low-key ambient night lighting.", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a young Korean woman with long black hair, standing at three-quarter angle from the left, halted after a hurried descent, waist-up visible near the center of the frame at eye level. She appears visibly distressed and physically frozen in shock, her hair slightly disheveled, gazing into the distance with an unfocused expression. Behind her, the existing stair of the building is visible as a background element in the shadowed courtyard. The environment is immersed in subdued, low-key ambient street lighting that restricts bright colors and leaves much of the background heavily shadowed.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영 혼자 공포에 질려 옥탑방을 빠져나온 상태", "초점을 잃고 멍하니 서있는 수리영", "halted after hurried descent", "다세대 빌라 골목길 및 마당" ], "visual_inferences": [ "fixed_element su_ri_young_distressed_state: physically frozen in shock, hair slightly disheveled", "fixed_element night_alleyway_environment: subdued ambient street lighting, heavily shadowed" ], "creative_decisions": [ "Eye-level medium framing to focus on her frozen state while keeping the architectural escape route (stair) visible in the background.", "Integration of creator-corrected metal exterior stair details to establish the exact setting." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bicycle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "building", "usage_kind": "anchor", "source_phrase": "the existing stair of the building is visible as a background element in the shadowed courtyard" }, { "owned_token": "courtyard", "usage_kind": "anchor", "source_phrase": "the existing stair of the building is visible as a background element in the shadowed courtyard" }, { "owned_token": "gate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "police car", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "the existing stair of the building is visible as a background element in the shadowed courtyard" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e6e7448628075148", "owned_hash": "e12f67b0acc69e0e", "camera_direction_hash": "171e32350c9b451c", "owned_usage_hash": "41108c8f06e4f46b", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "building", "violating_phrase": "the existing stair of the building is visible as a background element in the shadowed courtyard", "reason": "The prompt refers to the \"existing... building\" as a background element without proposing any redraw or modification, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "courtyard", "violating_phrase": "the existing stair of the building is visible as a background element in the shadowed courtyard", "reason": "The prompt mentions the \"shadowed courtyard\" merely as lighting and spatial context for the background, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "the existing stair of the building is visible as a background element in the shadowed courtyard", "reason": "The prompt specifically refers to the \"existing stair\" as a secondary background element with no attempt to modify or recreate it, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 4, "visible_entities": [ "C01", "L10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 13, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At eye level on 수리영's frontal axis, complete the dolly-in from a full-figure view and settle at medium distance, allowing her slight bodily turn to create a three-quarter portrait rather than a direct frontal address. Hold her near center while the exterior staircase remains secondary behind her, letting the arrested stillness and unfocused gaze carry the frame.", "lighting_mood": "Low-key ambient illumination appropriate to the night exterior keeps brightness subdued and colors restrained.", "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", "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": [ "bicycle", "building", "courtyard", "gate", "police car", "stair", "wall" ], "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": [ { "element_id": "su_ri_young_distressed_state", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman standing outdoors at night, appearing visibly distressed and physically frozen in shock. Her clothing and hair remain slightly disheveled, reflecting her immediate physical state after a hurried and panicked descent from a higher floor.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "수리영 혼자 공포에 질려 옥탑방을 빠져나온 상태", "halted after hurried descent", "초점을 잃고 멍하니 서있는 수리영" ], "visual_inferences": [ "Her hair and clothing are slightly messy to show she just ran down the stairs in a panic." ], "creative_decisions": [ "Maintaining a rigid, frozen posture across the shots to emphasize the lingering shock." ], "confidence": "high" }, { "element_id": "night_alleyway_environment", "element_type": "environment_state", "character_name": "", "description": "An exterior nighttime setting in a residential alleyway and courtyard. The environment is immersed in subdued, low-key ambient street lighting that restricts bright colors and leaves much of the background heavily shadowed.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "다세대 빌라 골목길 및 마당", "Low-key ambient illumination appropriate to the night exterior", "Restrained night-exterior ambient light" ], "visual_inferences": [ "The light sources are mostly distant or weak streetlamps, creating soft edges and a gloomy mood." ], "creative_decisions": [ "Using a dark, restrained color palette for the environment to contrast with the characters and build tension." ], "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" }, { "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": "ad003134c82f04cd" }, { "scene_index": 13, "heading": "S#13. 옥탑방 외곽 / N", "beat_title": "충격적인 소식의 전달", "representative_moment": "고개를 든 채 혜수의 얼굴을 마주보고 있는 수리영의 측면 구도", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium two-shot from shoulder height with subdued low-key contrast", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki casual top, a young Korean woman with long black hair, stands physically frozen in shock in the middle-left midground in right profile, her clothing and hair slightly disheveled, chin lifted and eyes fixed on C02O02's face. Across the narrow gap on the same depth plane, C02O02 in a leather blouson jacket, a middle-aged Korean woman with neat black hair, stands in the middle-right midground in left profile, leaning forward with eyes fixed on C01O01's raised face. A static medium two-shot from shoulder height frames their shared profile axis, focusing on the visual tension between them. Behind them, the existing metal stair extends against the building exterior, its open treads, thin steel stringers, and pipe railings softly visible in side view. The environment is immersed in subdued, low-key ambient street lighting that restricts bright colors and leaves much of the residential courtyard heavily shadowed, contrasting the pale skin tones against the dark background.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "수리영 혼자 공포에 질려 옥탑방을 빠져나온 상태", "도착한 혜수에게 엄마의 죽음을 알리고 있는 순간", "고개를 든 채 혜수의 얼굴을 마주보고 있는 수리영의 측면 구도" ], "visual_inferences": [ "fixed_element: 수리영 is physically frozen in shock, clothing and hair slightly disheveled reflecting her immediate physical state after a panicked descent", "fixed_element: subdued, low-key ambient street lighting that restricts bright colors and leaves much of the background heavily shadowed" ], "creative_decisions": [ "Frame spatial contract enforced in the first two sentences to secure midground positions and crossing sightlines on the shared depth plane.", "Included the Creator Corrections regarding the 'metal stair' (open treads, thin steel stringers, pipe railings) to anchor the specific architectural detail without conflicting with the reference.", "Employed a static medium two-shot from shoulder height to isolate their shared profile axis and capture the tension in the narrow gap." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bicycle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "building", "usage_kind": "anchor", "source_phrase": "against the building exterior" }, { "owned_token": "courtyard", "usage_kind": "anchor", "source_phrase": "much of the residential courtyard heavily shadowed" }, { "owned_token": "gate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "police car", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "the existing metal stair extends" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "0b8aa13d376339fa", "owned_hash": "e12f67b0acc69e0e", "camera_direction_hash": "cf575c700fc29baf", "owned_usage_hash": "d4d0ffc3e68f8fbb", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "building", "violating_phrase": "against the building exterior", "reason": "The prompt mentions the building exterior solely to establish the spatial context behind the characters, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "courtyard", "violating_phrase": "much of the residential courtyard heavily shadowed", "reason": "The prompt references the courtyard only to describe ambient lighting and shadowing in the scene, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "the existing metal stair extends", "reason": "The prompt explicitly describes the stair as existing and details its soft visibility in the background without commanding a redraw, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 13, "dependency_reason": "Shot 4의 배경 공간과 연속된 시간대를 공유하며 일관된 옥탑방 외곽 환경을 유지해야 함", "_shot_index": 6, "visible_entities": [ "C01", "C02", "L10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 13, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From shoulder height on the pair's shared profile axis, hold a static medium two-shot that favors 수리영's raised profile on one side while preserving 혜수's face directly across from her. Keep both women on the same depth plane with a narrow gap between their faces, turning the empty space into the visual channel for 수리영's revelation.", "lighting_mood": "Restrained night-exterior ambient light holds both profiles in subdued low-key contrast.", "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": "looks_toward", "gesture_target_label": "혜수's face", "label": "수리영", "screen_zone": "middle_left", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "looks_toward", "gesture_target_label": "수리영's raised face", "label": "혜수", "screen_zone": "middle_right", "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", "L10" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C02", "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": "background_ref_attached", "bg_id": "L10B01", "reference_usage": "exact_background", "owned_objects": [ "bicycle", "building", "courtyard", "gate", "police car", "stair", "wall" ], "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": [ { "element_id": "su_ri_young_distressed_state", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman standing outdoors at night, appearing visibly distressed and physically frozen in shock. Her clothing and hair remain slightly disheveled, reflecting her immediate physical state after a hurried and panicked descent from a higher floor.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "수리영 혼자 공포에 질려 옥탑방을 빠져나온 상태", "halted after hurried descent", "초점을 잃고 멍하니 서있는 수리영" ], "visual_inferences": [ "Her hair and clothing are slightly messy to show she just ran down the stairs in a panic." ], "creative_decisions": [ "Maintaining a rigid, frozen posture across the shots to emphasize the lingering shock." ], "confidence": "high" }, { "element_id": "night_alleyway_environment", "element_type": "environment_state", "character_name": "", "description": "An exterior nighttime setting in a residential alleyway and courtyard. The environment is immersed in subdued, low-key ambient street lighting that restricts bright colors and leaves much of the background heavily shadowed.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "다세대 빌라 골목길 및 마당", "Low-key ambient illumination appropriate to the night exterior", "Restrained night-exterior ambient light" ], "visual_inferences": [ "The light sources are mostly distant or weak streetlamps, creating soft edges and a gloomy mood." ], "creative_decisions": [ "Using a dark, restrained color palette for the environment to contrast with the characters and build tension." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 13, "shot_index": 4, "ref_usage": "", "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": "C02O02", "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": "848bfd4066766d32" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "사라진 시신과 흔적", "representative_moment": "핏자국과 시신이 감쪽같이 사라지고 지나치게 깨끗하게 정돈된 옥탑방 전경.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "eye-level wide shot with subdued night-interior lighting", "t2i_prompt": "Photorealistic cinematic still. The bed beneath the now-unmarked wall occupies the middle-right midground, and the curtain area where the body had been found is located in the middle-left background. Framed from eye level just inside the open bedroom door, the vertical edges of the doorway border the foreground, revealing an excessively orderly bedroom area. The space is completely devoid of any bloodstains or signs of a struggle. The bed is neatly made and tightly smoothed out, situated under the bare wall that is entirely clean with no red markings. The window curtain hangs smoothly and undisturbed. The center of the room is conspicuously unoccupied, drawing focus to the absence of the body. Subdued ambient night-interior illumination fills the unnaturally orderly room, characterized by restrained color and low-key contrast.", "outfit_assignments": [], "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다." ], "visual_inferences": [ "fixed_element orderly_bedroom_environment: neatly made bed, bare wall, smooth window curtain", "doorway framing emphasizes the empty center of the room", "absence of the body and blood is the visual focal point" ], "creative_decisions": [ "wide doorway composition to anchor the observer's point of view", "subdued ambient night-interior lighting with low-key contrast to highlight the uncanny neatness of the room" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "The bed beneath the now-unmarked wall occupies the middle-right midground" }, { "owned_token": "bedroom door", "usage_kind": "anchor", "source_phrase": "just inside the open bedroom door" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "the curtain area where the body had been found is located in the middle-left background" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steel door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage unit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wardrobe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "The window curtain hangs smoothly and undisturbed" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "68e542e5a79ba9aa", "owned_hash": "5035fab4fe2ae00f", "camera_direction_hash": "d2553b5e172529f4", "owned_usage_hash": "5299aa712982b1bc", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "The bed is neatly made and tightly smoothed out", "reason": "The prompt describes the bed's visual state as 'neatly made and tightly smoothed out' to convey the excessively tidied-up condition required by the shot_intent ('지나치게 깨끗하게 정돈된'), qualifying as an allowed visual state change.", "verdict": "allowed_visual_state_change" }, { "owned_object": "bedroom door", "violating_phrase": "just inside the open bedroom door", "reason": "The bedroom door is used as a spatial framing anchor to establish the camera's viewpoint, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "curtain", "violating_phrase": "The window curtain hangs smoothly and undisturbed", "reason": "The curtain is mentioned to establish its undisturbed layout and position as a spatial anchor, which is consistent with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "The window curtain", "reason": "The window is mentioned in the phrase 'window curtain' purely to locate the spatial anchor without any redraw instruction.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "핏자국과 시신이 있던 씬 12와의 극명한 대비를 보여주기 위해 동일한 방 구조와 카메라 위치를 유지하며 깨끗한 상태를 강조함.", "_shot_index": 5, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "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 eye level just inside the doorway, hold a static wide view across the excessively orderly room, keeping the bed, bare wall above it, and curtain legible as separate spatial zones. Leave the center conspicuously unoccupied so the absence of blood, the body, and the red mark reads before the camera moves farther inside.", "lighting_mood": "Subdued ambient night-interior illumination presents the unnaturally orderly room with restrained color and low-key contrast.", "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": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "bed beneath the now-unmarked wall", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "curtain area where the body had been found", "screen_zone": "middle_left", "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" ], "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": [ "TV", "TV stand", "backpack", "bed", "bedroom door", "chair", "counter", "curtain", "dining table", "doll", "drying platform", "floor lamp", "sink", "steel door", "storage unit", "stovetop", "wardrobe", "window" ], "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": [ { "element_id": "orderly_bedroom_environment", "element_type": "environment_state", "character_name": "", "description": "The bedroom area is excessively orderly and completely devoid of any bloodstains or signs of a struggle. The bed is neatly made, and the wall immediately above it is entirely bare, showing no red markings. A window curtain hangs smoothly and undisturbed nearby.", "applies_to_shots": [ 5, 7 ], "element_scope": "full", "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다.", "keeping the bed, bare wall above it, and curtain legible (Shot 5)", "retain the bare wall and orderly bed as muted, recognizable evidence (Shot 7)" ], "visual_inferences": [ "The bedding is smoothed out tightly, emphasizing an unnatural level of neatness.", "The surface of the bare wall is uniformly clean, highlighting the absence of the previous red mark." ], "creative_decisions": [ "The excessive cleanliness is presented as eerie and unsettling, contrasting sharply with the expectation of a violent crime scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 21, "ref_usage": "", "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": "d107ee720706a4ff" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "사라진 시신과 혈흔", "representative_moment": "C01이 핏자국 하나 없이 텅 비고 지나치게 깨끗해진 방을 보며 충격에 빠진 채 서 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "eye-level tight facial framing with shallow focus on stunned eyes, low-key ambient night-interior illumination", "t2i_prompt": "Photorealistic cinematic still. Captured at eye level, C01O01 in a khaki sweatshirt, an early-20s Korean woman, appears at three-quarter angle from the right, shoulders braced after a sudden stop. Her stunned face is placed off-center at a tight facial distance, with shallow focus isolating her wide eyes gazing off-screen. [: Low-key ambient night illumination fills the slim corridor of empty space stretching along her eyeline, leaving a neatly made bed and an entirely bare, excessively orderly wall without red markings softly blurred but recognizable as muted background evidence.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영 방 안 지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "수리영이 멍한 눈으로 텅 빈 방 안을 응시하는 얼굴 클로즈업" ], "visual_inferences": [ "fixed_element orderly_bedroom_environment as muted background evidence", "Su-ri-young's shoulders braced to reflect the sudden halt and shock" ], "creative_decisions": [ "placed her face off-center to preserve a slim corridor of empty room along her eyeline, emphasizing the void", "used shallow depth of field to isolate her stunned eyes while keeping the excessively clean bed and bare wall softly recognizable" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bedroom door", "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": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steel door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage unit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "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": "d9498f977bc4992b", "owned_hash": "5035fab4fe2ae00f", "camera_direction_hash": "c375854b321e34d4", "owned_usage_hash": "b657c01586d72213", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "Shot 5의 깨끗해진 방(침대와 빈 벽)을 수리영의 시선 뒤편 흐릿한 배경 단서로 연속성 유지", "_shot_index": 7, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 7 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At 수리영's eye level inside the room, finish the dolly-in at a tight facial distance, placing her face off-center while preserving a slim corridor of empty room along her eyeline. Use shallow focus on her stunned eyes but retain the bare wall and orderly bed as muted, recognizable evidence behind the direction of her stare.", "lighting_mood": "Low-key ambient night-interior illumination keeps 수리영's pale shock subdued against the restrained room 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" ], "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": "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": "orderly_bedroom_environment", "element_type": "environment_state", "character_name": "", "description": "The bedroom area is excessively orderly and completely devoid of any bloodstains or signs of a struggle. The bed is neatly made, and the wall immediately above it is entirely bare, showing no red markings. A window curtain hangs smoothly and undisturbed nearby.", "applies_to_shots": [ 5, 7 ], "element_scope": "full", "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다.", "keeping the bed, bare wall above it, and curtain legible (Shot 5)", "retain the bare wall and orderly bed as muted, recognizable evidence (Shot 7)" ], "visual_inferences": [ "The bedding is smoothed out tightly, emphasizing an unnatural level of neatness.", "The surface of the bare wall is uniformly clean, highlighting the absence of the previous red mark." ], "creative_decisions": [ "The excessive cleanliness is presented as eerie and unsettling, contrasting sharply with the expectation of a violent crime scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 5, "ref_usage": "", "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": "449cb69d0090a50b" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "혜수의 서늘한 시선과 형사의 짜증", "representative_moment": "C02 glares sharply at the detective from across the room as he turns to leave.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium-close composition at eye level, subdued ambient night-interior light", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a dark leather jacket, a Korean middle-aged woman with neat black hair and a mature oval face with distinct features, stands in the middle-left midground, looking toward a Korean middle-aged man in a detective coat near the right edge of the frame. C02O02 is seen at a three-quarter angle from the right, her shoulders squared toward him, her sharp gaze cutting across the space. The man is captured from over the shoulder, his turned torso acting as a secondary counterweight in the foreground, turning toward the open bedroom door to exit, looking off-screen. Captured at eye level, the clean dining table anchors the lower frame, its surface angling beneath the two figures sharing the space. Subdued ambient night-interior light and restrained contrast sharpen C02O02's expression, keeping the surrounding background in deep shadow.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "shot description: 혜수가 날카로운 눈빛으로 김형사를 노려보는 모습", "scene text: (김형사 째리고) 김형사. 먼저 가 있어. (김형사 나가면)", "DP instructions: At 혜수's eye level beside the table, let the pan settle into a medium-close composition with 혜수 in sharp three-quarter view" ], "visual_inferences": [ "the detective is placed near the right edge acting as a foreground counterweight", "the dining table serves as a shared spatial anchor in the lower frame beneath both figures", "the open bedroom door indicates the detective's exit path" ], "creative_decisions": [ "subdued ambient night-interior light to heighten the tension in 혜수's gaze", "used a generic descriptor for the detective (Korean middle-aged man in a detective coat) as he is not in the allowed entity list", "kept the background in restrained contrast to emphasize the characters' interaction" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bedroom door", "usage_kind": "anchor", "source_phrase": "turning toward the open bedroom door to exit" }, { "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": "dining table", "usage_kind": "anchor", "source_phrase": "the clean dining table anchors the lower frame" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steel door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage unit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "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": "a29a36fe1dbe9586", "owned_hash": "5035fab4fe2ae00f", "camera_direction_hash": "8d4672166bfabecc", "owned_usage_hash": "26c31d30f8943401", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bedroom door", "violating_phrase": "turning toward the open bedroom door to exit", "reason": "The bedroom door is used as a spatial reference for the character's movement, which matches the producer's declaration of anchor.", "verdict": "anchor_reference" }, { "owned_object": "dining table", "violating_phrase": "the clean dining table anchors the lower frame", "reason": "The dining table is mentioned solely as a compositional anchor to frame the lower shot, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "Maintains the exact background established in S14_Shot5 of the clean, ordered room.", "_shot_index": 11, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 11 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At 혜수's eye level beside the table, let the pan settle into a medium-close composition with 혜수 in sharp three-quarter view and 김형사 held nearer the edge of his exit path. Keep his shoulder and turned torso as a secondary counterweight while the focus isolates the hard line of 혜수's gaze across the frame.", "lighting_mood": "Subdued ambient night-interior light and restrained contrast sharpen 혜수's expression without introducing a specific ungrounded 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": { "reason": "looks_to_anchor", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "looks_toward", "gesture_target_label": "김형사", "label": "혜수", "screen_zone": "middle_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": [ "C02" ], "allowed_outlook_pairs": [ { "character_id": "C02", "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": "background_ref_attached", "bg_id": "L04B06", "reference_usage": "exact_background", "owned_objects": [ "TV", "TV stand", "backpack", "bed", "bedroom door", "chair", "counter", "curtain", "dining table", "doll", "drying platform", "floor lamp", "sink", "steel door", "storage unit", "stovetop", "wardrobe", "window" ], "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": 14, "shot_index": 5, "ref_usage": "", "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": "C02O02", "policy": "required" }, { "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": "9f8846f28f723a2e" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "헛것과 현실의 괴리 속 혼란", "representative_moment": "혜수의 손이 음료수 캔을 쥔 채 수리영 앞으로 내밀어진 순간", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Hand-height close framing with a three-quarter perspective, emphasizing the spatial tension of the human exchange over the foreground prop under low-key lighting.", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a dark leather jacket, a Korean middle-aged woman, leans through her extended arm in the middle-left foreground, reaching for the space directly before a young Asian woman in her 20s. Observed from a hand-height perspective at a three-quarter side angle, C02O02 appears at a three-quarter angle from the right, her eyes fixed on the seated young woman. The young Asian woman occupies the shared bench, seated with rounded shoulders at a three-quarter angle from the left, looking down, her form rendered slightly soft in the shallow depth of field. A beverage can, gripped in C02O02's hand, sits near the lower center of the image, held horizontally with its side angled toward the camera, crossing the air between them. [: A subdued nighttime coastal atmosphere enveloped in low-key ambient illumination, with a soft dark sea visible in the distance, keeping the overall visual palette muted and natural].", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "혜수, 사온 음료수를 수리영에게 건네며", "바다를 바라보며 멍한 채로 앉아있는 수리영", "At hand height from a three-quarter side angle, complete the dolly-in as 혜수's extended hand places the beverage can directly before seated 수리영." ], "visual_inferences": [ "fixed_element nighttime_coastal_environment: subdued nighttime coastal atmosphere, low-key ambient illumination.", "Su-ri-young (unregistered base) is rendered as a young Asian woman, seated on a shared bench with rounded shoulders, looking down.", "The beverage can is held horizontally in Hye-soo's hand, kept proportionally small near the lower center." ], "creative_decisions": [ "Applied a shallow depth of field to render the receiving young woman slightly soft, maintaining the primary focus on Hye-soo's reaching gesture to obey close framing rules.", "Placed the beverage can crossing the air between them to anchor the foreground/background spatial relationship.", "Used L02 instead of L01 to comply with the requested ID removal while supplying the frame-edge background surface description." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "8f909ce3ed093bba", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "f802d64cfb205266", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct visual dependency established with prior specific scenes; relies on the fixed environment state.", "_shot_index": 2, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "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 hand height from a three-quarter side angle, complete the dolly-in as 혜수's extended hand places the beverage can directly before seated 수리영. Keep the can below one fifth of the image near the lower center, with 혜수's reaching torso on one side and 수리영's rounded seated posture on the other so the human exchange remains larger than the prop.", "lighting_mood": "Low-key ambient illumination appropriate to the nighttime coast keeps the palette muted and natural.", "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": "space directly before 수리영", "label": "혜수", "screen_zone": "middle_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": [ "C02" ], "allowed_outlook_pairs": [ { "character_id": "C02", "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": 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": "nighttime_coastal_environment", "element_type": "environment_state", "character_name": "", "description": "The scene is enveloped in a subdued nighttime coastal atmosphere. Low-key ambient illumination washes over the area, keeping the overall visual palette muted, dark, and natural.", "applies_to_shots": [ 2, 6 ], "element_scope": "full", "source_facts": [ "Low-key ambient illumination appropriate to the nighttime coast keeps the palette muted and natural.", "Subdued nighttime coastal ambience keeps both figures readable without breaking the distant, low-key observational mood." ], "visual_inferences": [], "creative_decisions": [ "Maintained a consistently dim and subdued lighting setup to emphasize the lonely nighttime coastal setting across different camera distances." ], "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": "C02O02", "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": "bb009452c91c06b3" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "감시자의 시선 속에서 헤어지는 두 사람", "representative_moment": "C02O02가 멀어지는 C01O01을 향해 몸을 기울여 외치는 뒷모습을 누군가 먼 발치에서 지켜보는 상황.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Distant eye-level voyeuristic wide shot with subdued nighttime coastal ambience", "t2i_prompt": "Photorealistic cinematic still. C01O01 in casual wear, a Korean early-20s woman with long black hair, moves toward the route away from the bus stop in the upper-center background, while C02O02 in a leather blouson, a Korean mid-40s woman with neat black hair, stands in the lower-center midground looking toward C01O01 running away. Captured from a distant voyeuristic eye-level position, the wide frame includes both full figures. C01O01 is seen with her back to the camera, frozen mid-stride with one foot lifted off the ground, her hair flowing back in the direction opposite to motion. C02O02 also has her back to the camera, her weight pitched forward, her eyes fixed on C01O01. The open paved ground of the coastal bus-stop vicinity serves as the shared spatial anchor they both occupy, with the dark sea visible stretching behind them. The scene is enveloped in a subdued nighttime coastal atmosphere. Low-key ambient illumination washes over the area, keeping the overall visual palette muted, dark, and natural, maintaining a long observational distance that suggests an unseen watcher.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "S#15 action: '수리영 달려간다', '혜수 등 뒤에 대고 소리친다'", "shot description: 멀리서 혜수의 전신과 한 발을 공중에 띄운 mid-stride 자세로 달리는 수리영의 뒷모습을 한 프레임에 담은 넓은 구도", "beat #2 situation: 누군가 두 사람의 일거수일투족을 은밀하게 지켜보는 감시 상황" ], "visual_inferences": [ "fixed_element nighttime_coastal_environment: subdued nighttime coastal atmosphere, low-key ambient illumination", "Voyeuristic DP angle framing both characters from behind to suggest an unseen watcher's perspective" ], "creative_decisions": [ "Positioned C01 in the upper-center background and C02 in the lower-center midground to emphasize the separating distance and spatial tension", "Used the paved ground of the bus-stop vicinity as a shared spatial anchor to comply with fg_bg_shared_anchor_rule in a wide framing", "Applied mid-stride freeze wording for C01 to preserve the motion direction without implying a static posture" ], "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": "34ae45046ddb84f9", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "cfe0741af2685737", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 15, "dependency_reason": "Shot 2의 해안가 버스정류장 주변 환경 및 야간 해안의 조명/분위기 연속성 유지", "_shot_index": 6, "visible_entities": [ "C01", "C02", "L12" ], "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": "From a distant eye-level position behind 혜수, finish the withdrawal in a wide frame that includes both full figures: 혜수 remains nearer the lower frame while 수리영 recedes toward the upper center. Freeze the running rhythm with 수리영's back to camera and one foot lifted, preserving a long observational distance that suggests an unseen watcher.", "lighting_mood": "Subdued nighttime coastal ambience keeps both figures readable without breaking the distant, low-key observational mood.", "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": "moves_toward", "gesture_target_label": "route away from the bus stop", "label": "수리영", "screen_zone": "upper_center", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "looks_toward", "gesture_target_label": "수리영 running away", "label": "혜수", "screen_zone": "lower_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": [ "C01", "C02", "L12" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C02", "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": "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": "nighttime_coastal_environment", "element_type": "environment_state", "character_name": "", "description": "The scene is enveloped in a subdued nighttime coastal atmosphere. Low-key ambient illumination washes over the area, keeping the overall visual palette muted, dark, and natural.", "applies_to_shots": [ 2, 6 ], "element_scope": "full", "source_facts": [ "Low-key ambient illumination appropriate to the nighttime coast keeps the palette muted and natural.", "Subdued nighttime coastal ambience keeps both figures readable without breaking the distant, low-key observational mood." ], "visual_inferences": [], "creative_decisions": [ "Maintained a consistently dim and subdued lighting setup to emphasize the lonely nighttime coastal setting across different camera distances." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 15, "shot_index": 2, "ref_usage": "", "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": "C02O02", "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": "92563f6fc17b62fc" }, { "scene_index": 16, "heading": "S#16. 경찰서 사무실 / N", "beat_title": "불길함을 감지하고 수사 행동에 나선 혜수", "representative_moment": "C02가 스마트폰 액정을 내밀어 김형사에게 보여주며 다급히 반응을 살피는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium push-in at seated eye level, subdued ambient night shift lighting", "t2i_prompt": "Photorealistic cinematic still. In the middle-left midground, C02O02 in a leather jacket, a Korean middle-aged woman, reaches for the man across the desk, while a Korean man in his 30s seated opposite C02O02 occupies the middle-right midground, with the desk separating the two officers located in the lower-center midground. Seen at three-quarter angle from the left, C02O02 leans forward over her outstretched arm, her eyes fixed on the man. Seen at three-quarter angle from the right, the man remains seated, leaning slightly from the desk and looking down at the object. Between them, C02O02's hand firmly holds P01, an ordinary smartphone. The active screen of P01 faces the man and remains obliquely visible to the camera, displaying a saved contact number. The lit screen forms the nearest focal point in the lower center. The scene is lit by subdued ambient office illumination appropriate to a night shift, keeping the surroundings restrained with cool shadows while the glowing phone screen casts local light. A filing cabinet is visible in the dim background, anchoring the space.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "혜수 (C02) thrusts her smartphone towards 김형사 (C05) showing Min-suk's saved number.", "They are in the police station office during a night shift.", "C02 is leaning forward, C05 is seated at his desk." ], "visual_inferences": [ "The desk acts as a physical barrier separating the two characters.", "The light from the smartphone screen provides a local glow on their faces and hands.", "The background office environment is dimly lit." ], "creative_decisions": [ "Placed the smartphone obliquely so the screen is visible to the camera while facing the man, anchoring the focal point in the lower center.", "Used medium framing to keep both characters and their spatial tension visible without blending close-ups." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002", "fsc_003" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "ceiling light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "anchor", "source_phrase": "desk separating the two officers located in the lower-center midground" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "filing cabinet", "usage_kind": "anchor", "source_phrase": "A filing cabinet is visible in the dim background" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "fe3bb058c96321c8", "owned_hash": "bf88b469d99b7201", "camera_direction_hash": "621861bfc85669f0", "owned_usage_hash": "7512b9b7930668bd", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "desk", "violating_phrase": "desk separating the two officers located in the lower-center midground", "reason": "The desk is used as a spatial reference to position the characters, which is an anchor usage that matches the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "filing cabinet", "violating_phrase": "A filing cabinet is visible in the dim background, anchoring the space", "reason": "The filing cabinet is described as being in the dim background to anchor the space, which is an anchor reference matching the producer's echo.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct visual dependency on a prior scene specified.", "_shot_index": 3, "visible_entities": [ "C02", "C05", "L13", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 16, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At seated eye level beside the desk, push into the space between 혜수 and 김형사 as the smartphone is thrust forward, making its lit screen the nearest point near the lower center without allowing it to exceed one fifth of the frame. Hold 혜수 leaning in on one side and 김형사 seated opposite on the other, with both reactions remaining visible across the offered screen.", "lighting_mood": "Subdued ambient office illumination appropriate to the night shift keeps the image restrained while the active phone screen remains locally legible.", "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": "desk separating the two officers", "screen_zone": "lower_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "reaches_for", "gesture_target_label": "김형사 across the desk", "label": "혜수", "screen_zone": "middle_left", "target_id": "C02", "target_kind": "character" }, { "constraint_id": "fsc_003", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "김형사 seated opposite 혜수", "screen_zone": "middle_right", "target_id": "C05", "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", "C05", "L13", "P01" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O02" }, { "character_id": "C05", "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": "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": "background_ref_attached", "bg_id": "L13B02", "reference_usage": "exact_background", "owned_objects": [ "ceiling light", "counter", "desk", "door", "filing cabinet", "monitor", "window", "workstation" ], "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": [], "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": "C02O02", "policy": "required" }, { "kind": "background", "id": "L13B02", "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": "53c4104ebaef789c" }, { "scene_index": 17, "heading": "S#17. 옥탑방 / 동틀무렵", "beat_title": "적막하고 고요한 새벽녘의 옥탑방 → 검은 고양이가 개입하며 스산하고 불길한 분위기가 조성된 상태", "representative_moment": "옥상 담벼락 위에서 한쪽 앞발을 허공으로 뻗은 검은 고양이의 측면", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Parallel tracking shot at cat's height, cool dawn color grading", "t2i_prompt": "Photorealistic cinematic still. A black cat in right profile stands on the rooftop wall being crossed in the lower-center midground. The feline's body is captured from the side, with one forepaw suspended in the open space to the right, preserving the direction of its movement. Its gaze is fixed off-screen to the right. The camera tracks parallel at the cat's height, keeping its full body clearly visible along the structure. Beyond the wall in the upper frame, the exterior of [L03: the same illuminated rooftop room] provides a restrained point of warm contrast. Subdued cool dawn ambient light fills the outdoor space, highlighting the silhouette against the dim sky.", "outfit_assignments": [], "source_facts": [ "S#17 action: '검은 고양이가 불이 켜져 있는 옥탑 방을 슬쩍 보더니 담을 넘는다'", "shot description: 옥상 담벼락 위에서 한쪽 앞발을 허공으로 뻗은 검은 고양이의 측면" ], "visual_inferences": [ "the cat's mid-stride pose with one forepaw suspended indicates the exact moment of crossing", "the illuminated rooftop room from the exact_background reference establishes the spatial continuity" ], "creative_decisions": [ "placed the cat in the lower-center midground to anchor the movement along the wall", "cool dawn ambient light contrasts with the warm glow of the rooftop room for an eerie tone" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "cat", "usage_kind": "redraw", "source_phrase": "A black cat in right profile" }, { "owned_token": "clothesline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "room", "usage_kind": "anchor", "source_phrase": "the same illuminated rooftop room" }, { "owned_token": "staircase", "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": "bf50ccd0a32965c7", "owned_hash": "8fc5804b3b2415e0", "camera_direction_hash": "dc5c154844d5d608", "owned_usage_hash": "1e73623eeada5a83", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "cat", "violating_phrase": "A black cat in right profile stands on the rooftop wall being crossed in the lower-center midground. The feline's body is captured from the side, with one forepaw suspended in the open space to the right", "reason": "The cat's pose and state change (having one forepaw suspended) is explicitly and structurally required by the shot intent ('한쪽 앞발을 허공으로 뻗은 검은 고양이'), making it a permissible state modification under exception (1).", "verdict": "allowed_visual_state_change" }, { "owned_object": "room", "violating_phrase": "the same illuminated rooftop room", "reason": "The room is mentioned strictly as a point of warm contrast in the background with no instructions to modify or redraw its structure, matching the anchor usage.", "verdict": "anchor_reference" }, { "owned_object": "parapet", "violating_phrase": "rooftop wall", "reason": "The rooftop wall (parapet) is referenced only as a spatial anchor for the cat's position, with no instructions to redesign or modify it.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 10, "dependency_reason": "previous_shot_refs 의 exact_background 로 S10_Shot7 (이전 장면의 옥탑방 배경) 연속성 참조", "_shot_index": 2, "visible_entities": [ "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 17, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At the black cat's body height, track parallel to the rooftop wall and hold a broadside profile as one forepaw hangs beyond the edge. Frame the cat at medium distance with its full body readable along the wall, leaving open space ahead of the suspended paw to preserve the direction of crossing.", "lighting_mood": "Dawn ambient light remains subdued while the already illuminated rooftop room provides a restrained point of contrast.", "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": "none", "gesture_target_label": "", "label": "rooftop wall being crossed", "screen_zone": "lower_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": [ "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": "L03B01", "reference_usage": "exact_background", "owned_objects": [ "cat", "clothesline", "deck", "door", "parapet", "platform", "room", "staircase", "water tank", "window" ], "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": 10, "shot_index": 7, "ref_usage": "", "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": "aaad0ea1c4786578" }, { "scene_index": 18, "heading": "S#18. 안방 안 / D", "beat_title": "기괴한 환영의 시작", "representative_moment": "A half-formed vivid red circular mark appears on the bathroom tile wall reflected within the mirror.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "At sink-side eye level, tight close-up entirely within the mirror focusing on the anomalous reflection", "t2i_prompt": "Photorealistic cinematic still. Close framing at sink-side eye level, capturing a tight crop entirely within a bathroom mirror surface. In the off-center reflected midground, a half-formed vivid red circular mark appears on the tiled wall, providing the only forceful color accent against the muted surroundings. A narrow margin of the mirror frame edges the view, with no figures present in the reflection. [L04: dimly lit bathroom tile wall with restrained surfaces] fills the reflected space. A consistent subdued ambient lighting with no harsh direct light sources maintains an eerie, low-key atmosphere, keeping the reflected background restrained behind the anomalous red mark.", "outfit_assignments": [], "source_facts": [ "S#18 action: '거울에 비친 욕실 벽에 갑자기 또 발생하는 붉은 원.'", "shot description: A red circular mark is half formed on its reflected image in the bathroom mirror." ], "visual_inferences": [ "fixed_element subdued_interior_lighting: A consistent subdued ambient illumination fills the interior spaces, maintaining a low-key and muted atmosphere with no harsh direct light sources.", "The mirror frame forms the natural edge of the composition, containing the supernatural manifestation." ], "creative_decisions": [ "Excluding the character entirely from the frame to isolate the supernatural red mark as the primary focal point within the mirror.", "Using a muted, cool color temperature for the ambient bathroom environment to make the vivid red anomaly stand out forcefully." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "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": "drain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shower", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "toilet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "3d051b467c9393be", "owned_hash": "8ff82884a2aa851e", "camera_direction_hash": "e2334d9ae80062fc", "owned_usage_hash": "5a4f2f0cd8812932", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "preserves the spatial layout and interior tone of the previously established bathroom environment", "_shot_index": 11, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 11 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At sink-side eye level, the camera holds a tight crop entirely within the mirror and completes the stage's reflected-space pan toward the tiled wall. The half-formed red circle sits off-center in the reflected midground, with only a narrow margin of mirror around it and no part of Suri-young visible.", "lighting_mood": "Subdued ambient bathroom illumination keeps the reflected wall restrained while the explicitly red mark provides the only forceful color accent.", "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": "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": "subdued_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "A consistent subdued ambient illumination fills the interior spaces, maintaining a low-key and muted atmosphere. There are no harsh direct light sources, keeping the overall contrast restrained. This dim environmental lighting unifies the different areas of the apartment.", "applies_to_shots": [ 11, 15, 16 ], "element_scope": "full", "source_facts": [ "Shot 11 describes subdued ambient bathroom illumination.", "Shot 15 describes subdued ambient illumination and muted color.", "Shot 16 describes subdued ambient illumination preserving low-key facial contrast." ], "visual_inferences": [ "The uniform dim lighting across the bathroom and the living area reflects the eerie and unsettling atmosphere of the apartment." ], "creative_decisions": [ "Using restrained ambient light to isolate and emphasize specific focal points of horror, such as the vivid red mark or the uncanny facial distortion, without introducing distracting light sources." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 5, "ref_usage": "", "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": [ { "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": "6ccaa1b98f19242c" }, { "scene_index": 18, "heading": "S#18. 아파트 내부 / D", "beat_title": "액자 속 기괴한 환시", "representative_moment": "식탁 위 P03 사진 속 중년 여성의 웃는 얼굴이 기괴하게 아래로 흘러내리는 찰나.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Macro focus on the photograph surface, capturing the surreal distortion of the printed image", "t2i_prompt": "Photorealistic cinematic still. The framed photograph, P03, dominates the view, resting against [: a dim, flat table surface]. Inside P03, the reproduced printed image of an Asian middle-aged woman is seen at three-quarter angle from the left, originally smiling and looking off-screen. Her facial features on the glossy paper bizarrely stretch and melt downward in an impossible, uncanny flow, with the heavy distortion pooling toward the lower edge. The distorted printed face occupies most of the frame, leaving only a narrow trace of the front-facing picture frame boundary visible around it for context. Consistent subdued ambient illumination and muted colors maintain a low-key atmosphere, highlighting the horrifying downward deformation of the printed portrait.", "outfit_assignments": [], "source_facts": [ "Suri-young looks at the photo of her and Minsuk placed on the dining table.", "Minsuk's brightly smiling face in the photo bizarrely crushes and melts downward." ], "visual_inferences": [ "fixed_element subdued_interior_lighting: A consistent subdued ambient illumination fills the interior spaces, maintaining a low-key and muted atmosphere.", "The distortion happens strictly on the 2D plane of the photograph, requiring a tight focus on the paper surface." ], "creative_decisions": [ "Used dominant framing on P03 to isolate the hallucinatory horror, minimizing the surrounding table background to a generic flat surface (skipped close framing).", "Applied 'an Asian middle-aged woman' as a generic descriptor instead of a character ID, conforming to the reproduction surface rule for the face inside the photo." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "prop" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "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": "drain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shower", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "toilet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1a8376f2181aa1c3", "owned_hash": "8ff82884a2aa851e", "camera_direction_hash": "cc532509f0f298f3", "owned_usage_hash": "5a4f2f0cd8812932", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "Same subdued ambient lighting style as previously established in S5_Shot9, carrying over the low-key environmental mood.", "_shot_index": 15, "visible_entities": [ "P03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 15 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The table-height track settles squarely onto the photograph, tightening until Minsuk's pictured face occupies most of the frame while a trace of the surrounding photograph remains for context. Her smiling facial image stretches downward in an impossible flow, with the distortion weighted toward the lower edge to motivate the cut to Suri-young.", "lighting_mood": "Subdued ambient illumination and muted color keep attention on the uncanny downward deformation of the photographed 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": [ "P03" ], "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": "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": "subdued_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "A consistent subdued ambient illumination fills the interior spaces, maintaining a low-key and muted atmosphere. There are no harsh direct light sources, keeping the overall contrast restrained. This dim environmental lighting unifies the different areas of the apartment.", "applies_to_shots": [ 11, 15, 16 ], "element_scope": "full", "source_facts": [ "Shot 11 describes subdued ambient bathroom illumination.", "Shot 15 describes subdued ambient illumination and muted color.", "Shot 16 describes subdued ambient illumination preserving low-key facial contrast." ], "visual_inferences": [ "The uniform dim lighting across the bathroom and the living area reflects the eerie and unsettling atmosphere of the apartment." ], "creative_decisions": [ "Using restrained ambient light to isolate and emphasize specific focal points of horror, such as the vivid red mark or the uncanny facial distortion, without introducing distracting light sources." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 9, "ref_usage": "", "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": "prop", "id": "P03", "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": 18, "shot_indices": [ 15 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P03", "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": "dce20fe293df3c97" }, { "scene_index": 18, "heading": "S#18. 안방 안 / D", "beat_title": "기괴한 환영에 경악하며 극심한 공포를 느끼는 수리영", "representative_moment": "C01이 양손으로 머리를 감싸 쥐고 경악하며 입을 크게 벌리고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "tight handheld close-up at eye level with subdued ambient illumination", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a Korean young woman isolated in panic, occupies the middle-center midground, facing the camera. A tight close-up frame holds her face between both hands, fingers tightly gripping her temples. Her open mouth and widened expression dominate the center. She is looking at the camera. A consistent subdued ambient illumination preserves low-key facial contrast across her features without introducing any harsh direct light sources. The near edge of a table falls below and behind her tight facial framing. [: subdued interior surfaces deeply out of focus in the dim ambient light, maintaining a low-key atmosphere with no harsh direct light sources]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#18 action: Suri-young holding her head, mouth wide open in shock.", "DP notes: At Suri-young's eye level directly across the table, tight handheld frame.", "DP notes: Subdued ambient illumination preserves low-key facial contrast." ], "visual_inferences": [ "fixed_element subdued_interior_lighting: consistent dim environmental lighting unifies the spaces without harsh light.", "The background is deeply out of focus due to the tight close-up framing.", "The table edge provides a spatial anchor below the face without distracting from the expression." ], "creative_decisions": [ "Framed the shot as a tight close-up on the face to maximize psychological distress while adhering to the single-camera rule.", "Excluded the L04 background ID as requested, using a generic L01 block for the deeply blurred background surface to fit the skipped_close_framing constraint." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "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": "drain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shower", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "toilet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "401a5e7d3efd2d8c", "owned_hash": "8ff82884a2aa851e", "camera_direction_hash": "2638037b991fcabd", "owned_usage_hash": "5a4f2f0cd8812932", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "Relies on prior shot 14 for established lighting mood, though specific background details are omitted via skipped_close_framing.", "_shot_index": 16, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 16 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At Suri-young's eye level directly across the table, a tight handheld frame holds her face between both hands without losing their grip around her temples. Her open mouth and widened expression occupy the center, while slight lateral instability preserves the shock of the preceding cut without changing the close distance.", "lighting_mood": "Subdued ambient illumination preserves low-key facial contrast without introducing a new visible light 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": { "reason": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "Suri-young isolated in panic", "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": 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": "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": "subdued_interior_lighting", "element_type": "environment_state", "character_name": "", "description": "A consistent subdued ambient illumination fills the interior spaces, maintaining a low-key and muted atmosphere. There are no harsh direct light sources, keeping the overall contrast restrained. This dim environmental lighting unifies the different areas of the apartment.", "applies_to_shots": [ 11, 15, 16 ], "element_scope": "full", "source_facts": [ "Shot 11 describes subdued ambient bathroom illumination.", "Shot 15 describes subdued ambient illumination and muted color.", "Shot 16 describes subdued ambient illumination preserving low-key facial contrast." ], "visual_inferences": [ "The uniform dim lighting across the bathroom and the living area reflects the eerie and unsettling atmosphere of the apartment." ], "creative_decisions": [ "Using restrained ambient light to isolate and emphasize specific focal points of horror, such as the vivid red mark or the uncanny facial distortion, without introducing distracting light sources." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 7, "ref_usage": "", "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": "44e8598bd41b0883" }, { "scene_index": 19, "heading": "S#19. 마트 / D", "beat_title": "CCTV 위치 확인", "representative_moment": "Suri-young and the male employee look up at the CCTV camera mounted on the mart's exterior wall.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide angle from slightly below shoulder height tilting up, muted early morning light", "t2i_prompt": "Photorealistic cinematic still. A wall-mounted CCTV camera is visible in the upper-center background, anchored to the exterior wall. C01O01 in early-morning casual attire, a young Korean woman, stands with her back to the camera in the lower frame, her neck craned upward with her eyes fixed on the CCTV camera above. Beside her on the same shared pavement, a Korean middle-aged man in a worker uniform also stands with his back to the camera, leaning back slightly to inspect the same CCTV camera high above. The viewpoint is positioned several steps behind and slightly below shoulder height, tilting up to keep the pair in the lower half as scale references while the exterior wall dominates the upper field. Subdued and muted early-morning ambient light bathes the environment in cool, restrained tones.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영과 남자직원이 마트 밖에서 벽면의 CCTV를 올려다보는 뒷모습", "이른 아침 (Early morning)" ], "visual_inferences": [ "fixed_element early_morning_ambient_light: soft and diffuse light without harsh deep shadows", "fixed_element suri_young_consistent_appearance: early-morning casual attire suitable for riding a bicycle" ], "creative_decisions": [ "Wide framing from below shoulder height to emphasize the elevation of the wall-mounted CCTV relative to the figures.", "Muted and restrained color tones to capture the tension and the specific hour before full daylight." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "background", "character" ], "owned_object_usage": [ { "owned_token": "AC unit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "CCTV camera", "usage_kind": "anchor", "source_phrase": "wall-mounted CCTV camera" }, { "owned_token": "awning", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "conduit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "downspout", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "junction box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "exterior wall" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "829aa03d054f47dc", "owned_hash": "7eb25826d1a39693", "camera_direction_hash": "cfea6816649c6c10", "owned_usage_hash": "8e12ef3a49c03dfd", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "CCTV camera", "violating_phrase": "A wall-mounted CCTV camera is visible in the upper-center background, anchored to the exterior wall", "reason": "The prompt references the CCTV camera solely as a visual anchor and spatial focal point for the characters' gaze, which aligns with the producer's 'anchor' declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "the exterior wall dominates the upper field", "reason": "The prompt mentions the wall as a compositional and spatial background anchor, matching the producer's 'anchor' declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 5, "visible_entities": [ "C01", "C08", "L09" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 19, "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 several steps behind and slightly below shoulder height, the camera frames both figures from the back and tilts upward along their raised sightlines to the wall-mounted CCTV. The pair remain in the lower half as scale references while the exterior wall and camera occupy the upper field, preparing the later arc around Suri-young.", "lighting_mood": "Natural ambient light appropriate to the stated early-morning exterior is kept subdued and muted.", "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": "wall-mounted CCTV", "screen_zone": "upper_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": [ "C01", "C08", "L09" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C08", "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": [ { "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": "L08B03", "reference_usage": "exact_background", "owned_objects": [ "AC unit", "CCTV camera", "awning", "bollard", "conduit", "downspout", "junction box", "wall", "window" ], "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": [ { "element_id": "early_morning_ambient_light", "element_type": "environment_state", "character_name": "", "description": "The exterior environment is illuminated by subdued and muted early-morning natural ambient light. The overall color palette remains restrained and slightly dim, reflecting the early hour before full daylight.", "applies_to_shots": [ 5, 6 ], "element_scope": "full", "source_facts": [ "이른 아침 (Early morning)", "Natural ambient light appropriate to the stated early-morning exterior is kept subdued and muted.", "Subdued early-morning ambient light maintains restrained color" ], "visual_inferences": [ "The lighting is soft and diffuse without harsh direct sunlight or deep shadows, creating a slightly cool and tense morning atmosphere." ], "creative_decisions": [ "Maintain the cool and muted color palette consistently across exterior shots to emphasize the early morning mood." ], "confidence": "high" }, { "element_id": "suri_young_consistent_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman wearing practical early-morning casual attire suitable for riding a bicycle. Her clothing, hairstyle, and overall external styling remain visually consistent throughout the sequence.", "applies_to_shots": [ 5, 6 ], "element_scope": "full", "source_facts": [ "수리영 (Suri-young)", "자전거를 주차하고 안으로 들어가는 수리영" ], "visual_inferences": [ "She wears comfortable and casual clothes suitable for a quick morning errand on a bicycle, with minimal morning styling." ], "creative_decisions": [ "Ensure her wardrobe and hair styling are kept consistent across the exterior shots to provide continuity, despite the change in camera angle and pose." ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 8, "shot_index": 5, "ref_usage": "", "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": "L08B03", "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": "7ee6e7811d343daf" }, { "scene_index": 19, "heading": "S#19. 마트 / D", "beat_title": "CCTV 위치 확인 및 다급한 통화", "representative_moment": "수리영이 마트 외벽을 등진 채 굳은 표정으로 카메라를 응시하며 스마트폰을 귀에 대고 통화하는 상체 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "eye-level upper-body close-up with shallow depth of field and subdued early-morning cool lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01, a young Korean woman in her 20s in a khaki sweatshirt, stands in the middle-center midground, facing the camera and looking directly at the camera. She holds a modern smartphone, P01, pressed to her ear at the edge of the frame, the phone's outer side facing the camera. This eye-level upper-body close-up focuses entirely on her sharp facial tension and fixed forward stare, clearly showing her long black hair, oval young face, soft features, and natural skin tone. The scene is illuminated by subdued and muted early-morning natural ambient light, creating a slightly cool and diffuse atmosphere without deep shadows. [: out-of-focus exterior wall of a commercial building, rendered as a soft background to isolate the subject, bathed in dim early-morning ambient light with a restrained color palette].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 혜수에게 전화 걸어 상황을 설명한다.", "이른 아침 (Early morning)", "Subdued early-morning ambient light maintains restrained color and clear facial tension." ], "visual_inferences": [ "The lighting is soft and diffuse without harsh direct sunlight or deep shadows, creating a slightly cool and tense morning atmosphere.", "She wears comfortable and casual clothes suitable for a quick morning errand on a bicycle." ], "creative_decisions": [ "Use an eye-level upper-body close-up framing with a shallow depth of field to intensely isolate Suri-young and separate her from the background.", "Maintain a cool and muted color palette consistently to emphasize the early morning mood and thriller suspense." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "AC unit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "CCTV camera", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "awning", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "conduit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "downspout", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "junction box", "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": "5631e0095c0332bb", "owned_hash": "7eb25826d1a39693", "camera_direction_hash": "cf8b570d070a1988", "owned_usage_hash": "110ae76a2d12c42c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 8, "dependency_reason": "동일한 이른 아침 외부 환경의 조명 톤(subdued early-morning ambient light) 및 수리영의 의상·스타일링(C01O01) 연속성 유지", "_shot_index": 6, "visible_entities": [ "C01", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 19, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The completed rear-to-front move resolves into a static eye-level upper-body close-up directly before Suri-young. She holds the smartphone against one ear at the side of frame while her fixed forward stare and tense face remain centered, separating her from the employee and the CCTV investigation behind her.", "lighting_mood": "Subdued early-morning ambient light maintains restrained color and clear facial tension.", "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": "Suri-young alone on the call", "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", "P01" ], "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": "L08B03", "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": "early_morning_ambient_light", "element_type": "environment_state", "character_name": "", "description": "The exterior environment is illuminated by subdued and muted early-morning natural ambient light. The overall color palette remains restrained and slightly dim, reflecting the early hour before full daylight.", "applies_to_shots": [ 5, 6 ], "element_scope": "full", "source_facts": [ "이른 아침 (Early morning)", "Natural ambient light appropriate to the stated early-morning exterior is kept subdued and muted.", "Subdued early-morning ambient light maintains restrained color" ], "visual_inferences": [ "The lighting is soft and diffuse without harsh direct sunlight or deep shadows, creating a slightly cool and tense morning atmosphere." ], "creative_decisions": [ "Maintain the cool and muted color palette consistently across exterior shots to emphasize the early morning mood." ], "confidence": "high" }, { "element_id": "suri_young_consistent_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman wearing practical early-morning casual attire suitable for riding a bicycle. Her clothing, hairstyle, and overall external styling remain visually consistent throughout the sequence.", "applies_to_shots": [ 5, 6 ], "element_scope": "full", "source_facts": [ "수리영 (Suri-young)", "자전거를 주차하고 안으로 들어가는 수리영" ], "visual_inferences": [ "She wears comfortable and casual clothes suitable for a quick morning errand on a bicycle, with minimal morning styling." ], "creative_decisions": [ "Ensure her wardrobe and hair styling are kept consistent across the exterior shots to provide continuity, despite the change in camera angle and pose." ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 8, "shot_index": 5, "ref_usage": "", "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": "a8f827d505bf18fb" }, { "scene_index": 20, "heading": "S#20. 마트 총무과 / 실내", "beat_title": "흐릿한 CCTV 화면", "representative_moment": "보안직원과 혜수가 책상 앞 모니터를 향해 몸을 기울이고 어두운 CCTV 화면을 주시하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "static medium side composition at seated eye level with subdued ambient office illumination", "t2i_prompt": "Photorealistic cinematic still. Beside the desk shared by both viewers in the lower-center foreground, two figures lean toward the computer monitor beyond the pair in the middle-right background. A Korean adult man in a mart security uniform in right profile, leaning forward over the desk, eyes fixed off-screen. Beside him on an adjacent depth plane, C02O02 in a dark leather blouson jacket, a Korean middle-aged woman at three-quarter angle from the right, bracing toward the monitor, eyes fixed off-screen. A static medium side composition at seated eye level captures their profiles and forward-set shoulders, creating a narrowing corridor toward the screen. The desk's side edge leads from the camera toward the monitor. The physical monitor housing frames a monochrome CCTV video feed, its display face angled away from this side view with a faint human-shaped form only partially readable. Subdued ambient office illumination supports low-key contrast.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "마트 총무과 실내에서 혜수와 보안직원이 컴퓨터로 CCTV 화면을 돌려보고", "화면 속 엄마가 한 남자와 대화하는 장면만 기이하게 어둡고 흐려 식별이 불가능하다", "security employee and Hye-su lean toward the monitor just beyond them" ], "visual_inferences": [ "fixed_element cctv_monitor: A desktop computer monitor placed on the office desk, actively displaying a monochrome CCTV video feed", "The physical monitor housing and its edges frame the glowing digital display" ], "creative_decisions": [ "static medium side composition at seated eye level places the figures on adjacent depth planes", "Subdued ambient office illumination supports low-key contrast", "The desk's side edge leads from the camera toward the monitor to anchor the lower frame" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "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": "Beside the desk shared by both viewers in the lower-center foreground" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "anchor", "source_phrase": "computer monitor beyond the pair in the middle-right background" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "7eaa20245830146c", "owned_hash": "012d1de71682b9e4", "camera_direction_hash": "60a133fd774cae8a", "owned_usage_hash": "558e60be55217e04", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "desk", "violating_phrase": "Beside the desk shared by both viewers in the lower-center foreground, two figures lean toward... leaning forward over the desk... The desk's side edge leads from the camera toward the monitor", "reason": "The desk is used purely as a spatial anchor and composition guide, matching the producer's echo of anchor.", "verdict": "anchor_reference" }, { "owned_object": "monitor", "violating_phrase": "The physical monitor housing frames a monochrome CCTV video feed, its display face angled away from this side view with a faint human-shaped form only partially readable", "reason": "The prompt describes the internal display content of the monitor screen, which is explicitly required by the shot_intent ('어두운 CCTV 화면을 주시하는 순간'), satisfying all conditions for allowed visual state change.", "verdict": "allowed_visual_state_change" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No previous sequence establishes an exact visual match required for this shot's unique foreground staging.", "_shot_index": 1, "visible_entities": [ "C02", "C13", "L14" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 20, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Beside the desk at seated eye level, a static medium side composition places the security employee and Hye-su on adjacent depth planes as both lean toward the monitor just beyond them. Their profiles and forward-set shoulders create a narrowing corridor toward the screen, while the desk anchors the lower portion of the frame for the coming move behind them.", "lighting_mood": "Subdued ambient office illumination supports low-key contrast without specifying an unestablished source or color.", "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": "background", "gesture_action": "none", "gesture_target_label": "", "label": "computer monitor beyond the pair", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "desk shared by both viewers", "screen_zone": "lower_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": [ "C02", "C13", "L14" ], "allowed_outlook_pairs": [ { "character_id": "C02", "outlook_id": "O02" }, { "character_id": "C13", "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": "C13", "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", "monitor", "window" ], "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": [ { "element_id": "cctv_monitor", "element_type": "persistent_prop", "character_name": "", "description": "A desktop computer monitor placed on the office desk, actively displaying a monochrome CCTV video feed. The on-screen footage features an unusually dark section containing a faint, barely distinguishable human silhouette. The physical monitor housing and its edges frame the glowing digital display.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "monitor just beyond them", "settles directly before the monitor", "minimal edge of the display housing", "monochrome feed", "unusually dark corner", "faint human-shaped form" ], "visual_inferences": [ "The monitor is a standard flat-screen office monitor with a standard bezel." ], "creative_decisions": [ "The contrast between the monitor's display housing and the internal monochrome footage is highlighted to draw attention to the obscure dark corner." ], "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": "C02O02", "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": "120aa3d49d6084fd" }, { "scene_index": 20, "heading": "S#20. 마트 총무과 / 실내", "beat_title": "흐릿한 단서", "representative_moment": "A close-up of a monochrome CCTV monitor screen, showing a faint, indistinguishable silhouette hidden in a dark corner.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "insert shot focusing directly on the monitor screen, restrained monochrome tones", "t2i_prompt": "Photorealistic cinematic still. A dark CCTV section with a faint human figure occupies the middle-center midground of the frame. The camera settles directly before a desktop computer monitor at screen-center height, leaving only a minimal edge of the monitor's display housing visible at the frame boundaries. The active front face is square to the camera, showing a monochrome video feed with restrained tonal separation. Within the footage, an unusually dark corner suppresses detail, holding the barely distinguishable human silhouette near the center, merging into the heavy shadows.", "outfit_assignments": [], "source_facts": [ "close-up of monochrome CCTV screen", "faint human shape in a dark corner", "minimal edge of the display housing", "monochrome feed" ], "visual_inferences": [ "fixed_element: barely distinguishable silhouette merging into the darkness", "office monitor with a standard bezel anchoring the digital image" ], "creative_decisions": [ "insert framing prioritizing the screen content to emphasize obscurity and tension", "restrained monochrome tonal separation to heighten the occult thriller atmosphere" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "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": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "anchor", "source_phrase": "desktop computer monitor's display housing" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "6ac0a1747117c9bb", "owned_hash": "012d1de71682b9e4", "camera_direction_hash": "91c394d0ba60dfec", "owned_usage_hash": "95112a6ff09b6cc8", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "monitor", "violating_phrase": "a desktop computer monitor at screen-center height, leaving only a minimal edge of the monitor's display housing visible at the frame boundaries. The active front face is square to the camera, showing a monochrome video feed", "reason": "The prompt describes the screen content of the existing monitor showing a CCTV feed, which is explicitly and structurally required by the shot intent ('A close-up of a monochrome CCTV monitor screen, showing a faint, indistinguishable silhouette hidden in a dark corner').", "verdict": "allowed_visual_state_change" } ] } } ], "scene_type": "normal", "dependent_scene_index": 20, "dependency_reason": "exact_background ref from prior shot inside the same market office", "_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": "The dolly-in settles directly before the monitor at screen-center height, leaving only a minimal edge of the display housing as the CCTV image reaches the frame boundaries. Within the monochrome feed, the unusually dark corner is held near center and the faint human-shaped form remains barely separable from it.", "lighting_mood": "The monochrome CCTV feed provides restrained tonal separation, with the specified dark section suppressing detail around the figure.", "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": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "dark CCTV section with faint human figure", "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": [], "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", "monitor", "window" ], "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": [ { "element_id": "cctv_monitor", "element_type": "persistent_prop", "character_name": "", "description": "A desktop computer monitor placed on the office desk, actively displaying a monochrome CCTV video feed. The on-screen footage features an unusually dark section containing a faint, barely distinguishable human silhouette. The physical monitor housing and its edges frame the glowing digital display.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "monitor just beyond them", "settles directly before the monitor", "minimal edge of the display housing", "monochrome feed", "unusually dark corner", "faint human-shaped form" ], "visual_inferences": [ "The monitor is a standard flat-screen office monitor with a standard bezel." ], "creative_decisions": [ "The contrast between the monitor's display housing and the internal monochrome footage is highlighted to draw attention to the obscure dark corner." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 20, "shot_index": 1, "ref_usage": "", "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": "771d6295895a7085" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "사건 사진과 실종 좌표를 확인하고 충격에 빠진 수리영", "representative_moment": "수리영이 두 눈을 크게 뜨고 입을 살짝 벌린 굳은 얼굴 클로즈업", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "tight eye-level framing on the face, preserving three-quarter angle with an out-of-focus desk edge", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a casual jacket, a young Korean woman with long black hair, a young oval face, gentle facial features, and natural skin tone, leaning forward at three-quarter angle from the left. Seen at eye level, her face and upper shoulders occupy most of the frame, frozen with widened eyes and slightly parted lips, her gaze fixed off-screen. Subdued, restrained ambient illumination casts a dim, low-contrast cool glow on her features, keeping the colors muted. [L13B01: an out-of-focus edge of an office desk with blurred case photographs barely visible behind and below her shoulder].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영이 두 눈을 크게 뜨고 입을 살짝 벌린 굳은 얼굴 클로즈업", "혜수 책상위에는 살인사건 확대 사진들이 어지럽게 널려있다", "수리영 망설이다가 커서를 누른다" ], "visual_inferences": [ "fixed_element suri_young_appearance: casual outerwear jacket or coat", "fixed_element office_ambience_lighting: subdued, restrained ambient illumination", "cool muted color tone reflecting the chilling nature of the murder case" ], "creative_decisions": [ "eye-level close-up positioning to strongly emphasize her frozen, shocked expression", "out-of-focus desk edge with blurred photos preserves the workstation context without distracting from her face" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "GPS map", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "case photograph", "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": "light fixture", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "fc8f6aca76de31a0", "owned_hash": "d7ad311eb57ce7ed", "camera_direction_hash": "9e70d44f2fe43ca7", "owned_usage_hash": "b5210f44bbc03e25", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 16, "dependency_reason": "atmosphere_reference: prior shot establishes the subdued, restrained ambient illumination of the police station office", "_shot_index": 7, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 7 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "On Suri-young's three-quarter frontal axis at eye level, the dolly-in tightens until her widened eyes and slightly parted lips occupy most of the frame. The forward movement stops at the instant her face freezes, keeping the nearby monitor outside the crop but preserving her eyeline toward it for the following arc.", "lighting_mood": "Subdued ambient office illumination holds restrained contrast on her frozen expression without adding an unestablished light 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" ], "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": "L13B01", "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": "suri_young_appearance", "element_type": "character_state", "character_name": "Suri-young", "description": "Suri-young is a young Korean woman wearing a casual outerwear jacket or coat with accessible side pockets. Her demographic features, hairstyle, and the fabric and color of her outerwear remain visually consistent across the sequence. The clothing is practical and matches the contemporary setting.", "applies_to_shots": [ 7, 12 ], "element_scope": "full", "source_facts": [ "수리영이 두 눈을 크게 뜨고 입을 살짝 벌린 굳은 얼굴 클로즈업", "수리영이 메모지가 들어간 겉옷 주머니에 손을 넣은 채" ], "visual_inferences": [ "The outerwear is appropriate for late winter or early spring in 2026 Korea, featuring a practical design that allows her to hide her hand inside the pocket." ], "creative_decisions": [ "Keep the clothing colors muted or neutral to align with the somber, tense mood of the police station and her shocking discovery." ], "confidence": "high" }, { "element_id": "office_ambience_lighting", "element_type": "environment_state", "character_name": "", "description": "The police station office interior is characterized by subdued, restrained ambient illumination. The lighting maintains a somewhat dim, low-contrast atmosphere. There are no overly bright or distracting unestablished directional light sources visible.", "applies_to_shots": [ 7, 12 ], "element_scope": "full", "source_facts": [ "Subdued ambient office illumination holds restrained contrast", "Subdued ambient office illumination keeps the departure grounded" ], "visual_inferences": [ "The office relies on standard but slightly dim overhead lighting, reinforcing a serious and gritty reality within the station." ], "creative_decisions": [ "Use a slightly cool color temperature for the ambient light to emphasize the chilling nature of the murder case and the missing person situation." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 16, "shot_index": 3, "ref_usage": "", "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": "620ec04c72d55fa4" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "좌표 확인", "representative_moment": "모니터 화면에 바다를 가리키는 지도와 붉은색 핀, 그리고 마지막 위치 좌표가 떠 있는 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "insert shot directly in front of the monitor, pushing in to the edges", "t2i_prompt": "Photorealistic cinematic still. The sea map, red pin, and coordinate numbers on the monitor occupy the middle-center midground. Positioned directly in front of the screen, the framing is an insert shot where the active display extends to nearly all four edges, bounded only by a slim physical bezel of the workstation. The GPS map interface presents a restrained, muted ocean topography, sharply contrasted by a vivid crimson red location pin. Crisp digital coordinate text and an offline status marker are clearly visible against the subdued layout. The screen emits a cool, low-key digital glow, ensuring the red pin acts as the decisive color accent in the stark modern interface.", "outfit_assignments": [], "source_facts": [ "Shot description: 모니터 화면에 바다 지도가 펼쳐져 있고 붉은색 핀과 좌표 숫자가 떠 있는 클로즈업", "S#21 action: 해상 지도위에 좌표. 37.484400 N / 126. 072486 E ... Status: offline" ], "visual_inferences": [ "Slim physical bezel visible at the edges to ground the UI in the physical workstation context.", "Cool, subdued digital glow from the monitor emphasizing the red location pin." ], "creative_decisions": [ "Insert framing that fills the screen with the UI, isolating the crucial plot information.", "High contrast between the muted sea map background and the vivid crimson pin to create a decisive visual accent." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "GPS map", "usage_kind": "redraw", "source_phrase": "The GPS map interface" }, { "owned_token": "case photograph", "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": "light fixture", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "anchor", "source_phrase": "on the monitor" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "anchor", "source_phrase": "of the workstation" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "3b1854c328b8342b", "owned_hash": "d7ad311eb57ce7ed", "camera_direction_hash": "50671c04299ae82a", "owned_usage_hash": "355263b96dd6ed85", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "GPS map", "violating_phrase": "The GPS map interface presents a restrained, muted ocean topography, sharply contrasted by a vivid crimson red location pin.", "reason": "The prompt describes the visual content displayed on the GPS map interface, which is structurally required by the shot_intent to show a sea map and a red pin, qualifying for allowed_visual_state_change.", "verdict": "allowed_visual_state_change" }, { "owned_object": "monitor", "violating_phrase": "on the monitor", "reason": "The monitor is referenced purely as the surface holding the coordinates and map content, with no intent to redraw its physical structure, matching the anchor verdict.", "verdict": "anchor_reference" }, { "owned_object": "workstation", "violating_phrase": "slim physical bezel of the workstation", "reason": "The workstation is mentioned only to define the framing boundaries of the shot, which matches the anchor usage.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 21, "dependency_reason": "Background continuity for the police station workstation established in the prior shot.", "_shot_index": 10, "visible_entities": [], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 10 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Directly in front of the monitor at screen-center height, the push-in ends with the active display nearly reaching all four frame edges while retaining a slim physical bezel. The sea map, red location pin, and coordinate line are isolated in a clear hierarchy, with the pin offset enough that the coordinate numbers remain unobstructed before the downward tilt begins.", "lighting_mood": "The subdued display treatment keeps the sea map restrained while the explicitly red location pin becomes the decisive color accent.", "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": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "sea map, red pin, and coordinate numbers on monitor", "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": [], "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": "L13B01", "reference_usage": "exact_background", "owned_objects": [ "GPS map", "case photograph", "counter", "desk", "door", "filing cabinet", "light fixture", "monitor", "window", "workstation" ], "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": 21, "shot_index": 7, "ref_usage": "", "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": "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": "281cebf2c92e8cb1" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "수리영의 단독 결정과 이탈", "representative_moment": "C01이 주머니에 메모지를 숨긴 채 사무실 출입문을 향해 몸을 돌리며 서둘러 나서는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "static medium frame at chest height, subdued and cool-toned ambient lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a casual outerwear jacket, a young Korean woman in her 20s, is captured at three-quarter angle from the right, looking off-screen, with the office exit toward which she turns occupying the middle-right background. Her torso angles toward the exit in mid-turn, one hand buried tight at her hip inside the jacket pocket. The pocket opening is turned partly away from the camera, covered by her inserted hand. The police station desk, with the existing case photographs on its surface, recedes behind her on the opposite side from the exit. The doorway faces partly toward the camera, remaining visible beyond her shoulder. Subdued ambient office illumination casts a dim, cool-toned, low-contrast light over the room, keeping the grounded environment intact.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영이 메모지가 들어간 겉옷 주머니에 손을 넣은 채 사무실 출입문을 향해 몸을 돌린 정지 순간", "혜수 책상위에는 살인사건 확대 사진들이 어지럽게 널려있다" ], "visual_inferences": [ "fixed_element suri_young_appearance: casual outerwear jacket with accessible side pockets, hand buried inside", "fixed_element office_ambience_lighting: subdued, restrained ambient illumination, dim and low-contrast", "the desk and case photographs serve as a spatial anchor in the background, contrasting with her route toward the doorway" ], "creative_decisions": [ "Chest-height medium frame to properly capture the concealed hand while maintaining the spatial relationship with the exit", "Cool-toned dim lighting to emphasize the chilling tension of the GPS discovery without breaking the realistic office mood" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "GPS map", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "case photograph", "usage_kind": "anchor", "source_phrase": "the existing case photographs on its surface" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "anchor", "source_phrase": "The police station desk, with the existing case photographs on its surface, recedes behind her" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "the office exit toward which she turns" }, { "owned_token": "filing cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "light fixture", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d3f6306d9adf7d98", "owned_hash": "d7ad311eb57ce7ed", "camera_direction_hash": "8ca451dc3a70d4ce", "owned_usage_hash": "9a062f9f5dbfd80f", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "case photograph", "violating_phrase": "the existing case photographs on its surface", "reason": "The prompt mentions the existing case photographs as a passive background detail on the desk surface, matching the anchor declaration without requesting a redraw.", "verdict": "anchor_reference" }, { "owned_object": "desk", "violating_phrase": "The police station desk, with the existing case photographs on its surface, recedes behind her", "reason": "The desk is used as a spatial and compositional anchor receding in the background behind the character, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "the office exit toward which she turns occupying the middle-right background", "reason": "The exit and doorway are mentioned solely as spatial reference points for the character's orientation and position in the background, matching the anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 21, "dependency_reason": "동일 씬 내 이전 샷(Shot 7)의 사무실 배경 공간(exact_background) 및 고정 요소(조명, 인물 외투) 연속성 유지", "_shot_index": 12, "visible_entities": [ "C01", "L13" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 12 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At chest height beside the route from desk to exit, a static medium frame catches Suri-young midway through turning toward the office door. Her torso angles toward the exit while one hand stays buried at her hip around the pocketed note; the doorway remains visible ahead rather than allowing her gesture to read as casual.", "lighting_mood": "Subdued ambient office illumination keeps the departure grounded and the concealed hand readable without specifying an unestablished 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": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "office exit toward which Suri-young turns", "screen_zone": "middle_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": [ "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": 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": "L13B01", "reference_usage": "exact_background", "owned_objects": [ "GPS map", "case photograph", "counter", "desk", "door", "filing cabinet", "light fixture", "monitor", "window", "workstation" ], "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": [ { "element_id": "suri_young_appearance", "element_type": "character_state", "character_name": "Suri-young", "description": "Suri-young is a young Korean woman wearing a casual outerwear jacket or coat with accessible side pockets. Her demographic features, hairstyle, and the fabric and color of her outerwear remain visually consistent across the sequence. The clothing is practical and matches the contemporary setting.", "applies_to_shots": [ 7, 12 ], "element_scope": "full", "source_facts": [ "수리영이 두 눈을 크게 뜨고 입을 살짝 벌린 굳은 얼굴 클로즈업", "수리영이 메모지가 들어간 겉옷 주머니에 손을 넣은 채" ], "visual_inferences": [ "The outerwear is appropriate for late winter or early spring in 2026 Korea, featuring a practical design that allows her to hide her hand inside the pocket." ], "creative_decisions": [ "Keep the clothing colors muted or neutral to align with the somber, tense mood of the police station and her shocking discovery." ], "confidence": "high" }, { "element_id": "office_ambience_lighting", "element_type": "environment_state", "character_name": "", "description": "The police station office interior is characterized by subdued, restrained ambient illumination. The lighting maintains a somewhat dim, low-contrast atmosphere. There are no overly bright or distracting unestablished directional light sources visible.", "applies_to_shots": [ 7, 12 ], "element_scope": "full", "source_facts": [ "Subdued ambient office illumination holds restrained contrast", "Subdued ambient office illumination keeps the departure grounded" ], "visual_inferences": [ "The office relies on standard but slightly dim overhead lighting, reinforcing a serious and gritty reality within the station." ], "creative_decisions": [ "Use a slightly cool color temperature for the ambient light to emphasize the chilling nature of the murder case and the missing person situation." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 21, "shot_index": 7, "ref_usage": "", "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": "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": "83e5d9e8acbaba19" }, { "scene_index": 22, "heading": "S#22. 경찰서 밖 / D", "beat_title": "수리영의 급출발", "representative_moment": "수리영(C01)이 경찰서 밖 도로에서 자전거를 타고 황급히 달려나가는 순간", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Track parallel at upper-body height, wide side composition, muted daylight", "t2i_prompt": "Photorealistic cinematic still. Captured at upper-body height, holding the figure in a wide side composition. C01O01 in a khaki sweatshirt, a Korean young woman, in left profile, seated on the bicycle saddle in mid-pedal with her torso pitched forward. She is gazing into the distance, her long black hair flowing back horizontally in the direction opposite to motion. Her hands grip the handlebar, the bicycle frame and front wheel visible below her. The road passes laterally beneath the wheels, while the daylight exterior of the police station remains anchored in the background. Natural daylight casts subdued brightness with muted colors and restrained contrast.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#22 action: '주차된 자전거를 타고 이내 달리는 수리영'", "shot description: 경찰서 건물 밖 도로 위에서 수리영이 자전거 안장에 앉아 상체를 앞으로 숙인 채 머리카락이 뒤로 흩날리는 측면" ], "visual_inferences": [ "mid-pedal posture to convey continuous cycling motion", "bicycle frame visible below to physically support the rider without floating" ], "creative_decisions": [ "wide side composition tracking parallel to emphasize horizontal speed and motion", "subdued brightness and restrained contrast for realistic daylight 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": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "car", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "entrance", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "guardrail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "police station", "usage_kind": "anchor", "source_phrase": "exterior of the police station remains anchored in the background" }, { "owned_token": "road", "usage_kind": "anchor", "source_phrase": "The road passes laterally beneath the wheels" }, { "owned_token": "sidewalk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "streetlight", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "traffic mirror", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "utility pole", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e9810d45fa7bc7ee", "owned_hash": "e4edcd4cec3beea8", "camera_direction_hash": "1a5acf3c95907ac1", "owned_usage_hash": "668aa2b809f4ca1a", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "police station", "violating_phrase": "exterior of the police station remains anchored in the background", "reason": "The police station is mentioned purely as an anchored background element, which matches the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "road", "violating_phrase": "The road passes laterally beneath the wheels", "reason": "The road is referenced as a spatial anchor passing beneath the wheels without any instruction to redraw or alter its composition, matching the producer's echo.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "Not applicable", "_shot_index": 1, "visible_entities": [ "C01", "L15" ], "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": "Track parallel to 수리영 from directly beside the bicycle at upper-body height, holding her in left-facing profile near the center while allowing the moving road and police station exterior to remain legible around her. Use a slightly wider side composition with her forward-pitched torso and trailing hair forming a strong horizontal line of motion.", "lighting_mood": "Natural daylight is rendered with subdued brightness, muted colors, and restrained contrast.", "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", "L15" ], "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": "L15B01", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "bollard", "car", "curb", "door", "entrance", "guardrail", "police station", "road", "sidewalk", "sign", "streetlight", "traffic mirror", "utility pole", "window" ], "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": [], "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": "L15B01", "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": "e0cae8ac3cbd5501" }, { "scene_index": 23, "heading": "S#23. 어느 PC방 / 실내", "beat_title": "PC방에서의 좌표 검색", "representative_moment": "어두운 PC방 안, 모니터 불빛에 얼굴이 비친 상태로 화면을 내려다보는 C01의 정면 상체.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "static medium framing at seated eye level, low-key monitor glow", "t2i_prompt": "Photorealistic cinematic still. A static camera placed at seated eye level, centering the upper body at medium distance. C01O01 in a khaki sweatshirt, a young East Asian woman with long black hair and a youthful oval face, facing the camera. She is seated leaning forward toward the monitor, looking down. The edge of the computer monitor is kept low in the foreground, its active screen face turned toward her and away from the camera, concealing its displayed content. The user-facing side of the computer workstation is partially visible around her. A dark interior setting illuminated primarily by the subdued glow of the computer monitor casting light on her face and upper body. This localized light source creates low-key contrast, highlighting only the immediate workstation area while the rest of the room remains in deep shadow.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#23 action: '모니터 화면에 앉아있는 수리영'", "Shot description: 어두운 PC방 안, 모니터 불빛이 비치는 수리영의 정면 상체" ], "visual_inferences": [ "fixed_element pc_cafe_lighting: A dark interior setting illuminated primarily by the subdued glow of a computer monitor, creating low-key contrast and keeping the room in deep shadow.", "The monitor face is turned away from the camera, focusing entirely on the subject's illuminated face." ], "creative_decisions": [ "Centering the subject's upper body with the monitor edge low in the foreground to create a confrontational frontal alignment without direct lens contact.", "Using deep shadow in the surrounding space to isolate the focused character." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "9f4f70d24de4bd0b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "985a8ee545946bfb", "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": [ "C01", "L16" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 23, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Place the static camera at 수리영's seated eye level, directly in front of her at medium distance, with her upper body centered and the monitor edge kept low in the foreground. The deliberate frontal alignment feels confrontational without having her look into the lens, while the surrounding PC café falls back around the concentrated figure.", "lighting_mood": "The dark PC café is lit on 수리영 primarily by the monitor glow, with low-key contrast and subdued surrounding brightness.", "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": "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": "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_cafe_lighting", "element_type": "environment_state", "character_name": "", "description": "A dark interior setting illuminated primarily by the subdued glow of a computer monitor. This localized light source creates low-key contrast, highlighting only the immediate workstation area while the rest of the room remains in deep shadow.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "어두운 PC방 안, 모니터 불빛", "The dark PC café is lit ... primarily by the monitor glow", "Monitor-lit ambient illumination remains subdued" ], "visual_inferences": [ "The monitor emits a localized digital light typical of computer screens", "The background is mostly obscured by shadows due to the low-key lighting" ], "creative_decisions": [ "Use the monitor glow to isolate the focal points from the unlit surrounding background" ], "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": "94c0e41bb7126b6d" }, { "scene_index": 23, "heading": "S#23. 어느 PC방 / 실내", "beat_title": "좌표가 찍힌 지도를 백팩에 챙겨넣는 수리영", "representative_moment": "C01의 두 손이 접힌 지도 종이(P04)를 열려 있는 백팩 안으로 밀어넣는 정지 순간", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "overhead high angle insert shot, subdued monitor-lit low-key contrast", "t2i_prompt": "Photorealistic cinematic still. Seen from an overhead high angle directly above the workstation, looking down. C01O01 in a khaki sweatshirt, a Korean woman in her 20s, leaning forward and looking down. Her two hands converge in the center of the frame, pressing P04, folded printed coastal maps, downward into an open backpack resting on the desk. The folded printed faces and map markings of P04 are partially visible to the camera, caught in mid-motion at the exact instant the papers pass through the receiving opening of the backpack beneath the hands. The dark interior setting is illuminated primarily by the subdued glow of an off-screen computer monitor, creating restrained color and low-key contrast that sharply separates the hands, the paper of P04, and the backpack opening, while leaving the rest of the workstation in deep shadow.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "어두운 PC방 안, 모니터 불빛", "여러 장의 지도를 접어 백팩에 넣는다", "두 손이 접힌 지도 종이들을 열려 있는 백팩 안으로 밀어넣은 정지 순간" ], "visual_inferences": [ "The monitor emits a localized digital light typical of computer screens", "The background is mostly obscured by shadows due to the low-key lighting", "Overhead angle allows a clear view into the backpack's opening as the maps go in" ], "creative_decisions": [ "Overhead insert framing to concentrate entirely on the interaction between the hands, P04, and the backpack", "Subdued cool monitor glow isolated to the central action area to maintain the tense, dark mood" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "prop" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "3a64ede8bd79ce9b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "bec159a0a3b6d4c3", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 23, "dependency_reason": "같은 씬의 이전 샷에서 확립된 모니터 조명 환경과 어두운 배경(pc_cafe_lighting)을 동일하게 유지", "_shot_index": 4, "visible_entities": [ "C01", "P04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 23, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From directly above and close to the workstation, angle downward so both hands converge on the open backpack while the folded maps cross the center of the frame. Hold at the instant the papers pass through the opening, keeping the maps and backpack below forty percent each while 수리영's hands provide the dominant human detail.", "lighting_mood": "Monitor-lit ambient illumination remains subdued, with restrained color and enough local contrast to separate hands, paper, and backpack opening.", "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", "P04" ], "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": "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_cafe_lighting", "element_type": "environment_state", "character_name": "", "description": "A dark interior setting illuminated primarily by the subdued glow of a computer monitor. This localized light source creates low-key contrast, highlighting only the immediate workstation area while the rest of the room remains in deep shadow.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "어두운 PC방 안, 모니터 불빛", "The dark PC café is lit ... primarily by the monitor glow", "Monitor-lit ambient illumination remains subdued" ], "visual_inferences": [ "The monitor emits a localized digital light typical of computer screens", "The background is mostly obscured by shadows due to the low-key lighting" ], "creative_decisions": [ "Use the monitor glow to isolate the focal points from the unlit surrounding background" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 23, "shot_index": 1, "ref_usage": "", "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": "prop", "id": "P04", "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": 23, "shot_indices": [ 4 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P04", "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": "8fa867b708c59016" }, { "scene_index": 24, "heading": "S#24. 해안가 / 소나기", "beat_title": "폭우 속 절박한 질주", "representative_moment": "비에 흠뻑 젖은 C01이 폭우를 뚫고 결연한 표정으로 자전거 핸들을 꽉 쥔 채 질주하는 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level tracking backward close framing, storm-dark ambient light with lightning highlights, shallow depth of field", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a soaked khaki sweatshirt, an Asian young woman with long black hair and a soft oval face, at three-quarter angle from the left, hunched forward, gaze fixed into the distance just beside the lens. Her rain-soaked face occupies most of the frame, showing hard facial tension. Hands firmly gripping the crosswise bicycle handlebar appear partially at the lower frame edge as a moving anchor. Wet hair and heavy raindrops are frozen in mid-air, trailing back in the direction opposite to the forward motion. Lit by storm-dark ambient light under heavy clouds with sharp lightning highlights and cool rain-muted colors, the dark coastal surroundings dissolved into a soft out-of-focus background.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#24 action: '천둥 번개와 함께 잔뜩 낀 먹구름이 세찬 비를 뿌린다. 자전거를 타고 빗속을 달리는 수리영.'", "shot description: '비에 흠뻑 젖은 수리영이 자전거 핸들을 꽉 쥔 채 앞을 노려보는 얼굴 클로즈업'" ], "visual_inferences": [ "mid-action freeze with hair and rain trailing backward to convey fast forward motion without sequential verbs", "hands and handlebars occupy the lower edge as a moving anchor due to close facial framing", "lightning flash emphasizing facial tension" ], "creative_decisions": [ "tracking backward camera angle keeping her face dominant in frame", "shallow depth of field blurring the background to follow skipped_close_framing policy while omitting L01 tag as requested", "cool rain-muted color temperature for the thunderstorm mood" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "957b359a69aa1484", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "ad46a20176e90582", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 9, "dependency_reason": "previous_shot_refs indicates Scene 9 as a location reference, though explicitly skipped in composition due to close framing", "_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": "Track backward just ahead of the bicycle at 수리영's eye level and slightly off its centerline, tightening until her rain-soaked face occupies most of the frame. Keep a narrow strip of the handlebars at the lower edge as a moving anchor while her fixed gaze passes just beside the lens.", "lighting_mood": "Storm-dark ambient light under heavy clouds is punctuated by the scene's lightning, with rain-muted color and hard facial tension.", "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": [], "previous_shot_refs": [ { "scene_index": 9, "shot_index": 1, "ref_usage": "", "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": "6a9bb622f6f834e5" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "엉망이 된 옥탑방과 수상한 발자국", "representative_moment": "해질 무렵, 누군가 헤집어 놓은 옥탑방 바닥에 진흙 묻은 발자국이 선명하게 찍혀 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Knee-height observer perspective, wide low-angle, deep shadows at dusk", "t2i_prompt": "Photorealistic cinematic still. A muddy footprint trail covers the lower-center foreground, past the existing overturned dining chairs in the middle-center midground, with the existing open living-room window in the upper-right background. Viewed from a knee-height observer perspective near the entrance door, looking slightly downward into [L04: ransacked rooftop interior]. The interior space is in severe disarray, with the existing scattered belongings seen on the floor near the table. Subdued low-key dusk ambient lighting casts deep shadows across the room, the dim natural light softly outlining the disturbed layout.", "outfit_assignments": [], "source_facts": [ "누군가 다녀간 흔적으로 보이는 발자국들", "집안에서 뭔가를 찾기 위해 가재도구를 온통 헤집은 것으로 보인다", "거실에 난 창은 열려있고, 식탁 의자들을 쓰러져있다", "muddy footprints as a receding path", "Overturned dining chairs interrupt the middle ground" ], "visual_inferences": [ "fixed_element ransacked_interior_state: The interior space is in severe disarray", "fixed_element subdued_dusk_lighting: deep shadows dominate the space", "Observer POV without a visible character in the frame" ], "creative_decisions": [ "Knee-height low angle to emphasize the muddy footprints leading inward", "Low-key contrast at dusk to create an eerie, suspenseful mood", "Framing the open window in the background to provide the only dim natural fill light" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002", "fsc_003" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "belonging", "usage_kind": "anchor", "source_phrase": "the existing scattered belongings seen on the floor near the table" }, { "owned_token": "chair", "usage_kind": "anchor", "source_phrase": "the existing overturned dining chairs in the middle-center midground" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "footprint", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "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": "anchor", "source_phrase": "the existing open living-room window in the upper-right background" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "34c7afd6d222b625", "owned_hash": "19affc080f4abd6d", "camera_direction_hash": "770b3cc3a65e9a52", "owned_usage_hash": "02d889bbd6517ec4", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "footprint", "violating_phrase": "A muddy footprint trail covers the lower-center foreground", "reason": "The prompt instructs drawing a new muddy footprint trail, which redraws the footprint object; however, this is permitted under visual state change as it is explicitly and structurally required by the shot intent ('진흙 묻은 발자국이 선명하게 찍혀 있다').", "verdict": "allowed_visual_state_change" }, { "owned_object": "chair", "violating_phrase": "the existing overturned dining chairs in the middle-center midground", "reason": "The prompt references the existing dining chairs solely as a spatial layout anchor, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "the existing open living-room window in the upper-right background", "reason": "The prompt references the existing window as a background landmark without proposing any redraw, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "near the entrance door", "reason": "The prompt mentions the entrance door purely as a spatial reference for the camera position, which the producer incorrectly declared as absent.", "verdict": "anchor_reference" }, { "owned_object": "belonging", "violating_phrase": "the existing scattered belongings seen on the floor near the table", "reason": "The prompt refers to the existing scattered belongings as a spatial reference, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "table", "violating_phrase": "near the table", "reason": "The prompt mentions the table purely as a spatial anchor to locate other objects, which the producer incorrectly declared as absent.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 18, "dependency_reason": "옥탑방 내부(L04)의 이전 배경 상태를 exact_background로 참조하여 연속성 유지", "_shot_index": 1, "visible_entities": [ "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Hold near the living-room entrance at knee height with a slightly downward wide angle, placing the muddy footprints as a receding path from the lower foreground toward the far side of the room. Overturned dining chairs interrupt the middle ground, while the open window and disturbed furnishings establish the searched interior before the forward track begins.", "lighting_mood": "Subdued ambient illumination appropriate to a rooftop room at dusk keeps the disturbed layout readable in low-key contrast.", "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": "muddy footprint trail", "screen_zone": "lower_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "open living-room window", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_003", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "overturned dining chairs", "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" ], "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", "TV stand", "bed", "belonging", "chair", "counter", "curtain", "door", "footprint", "lamp", "sink", "storage cabinet", "stovetop", "table", "wardrobe", "window" ], "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": [ { "element_id": "ransacked_interior_state", "element_type": "environment_state", "character_name": "", "description": "The interior space is in severe disarray, indicating a hasty search. Dining chairs lie overturned on the floor, and muddy footprints form a scattered path across the ground. Various household furnishings are displaced and left in a messy, disturbed condition.", "applies_to_shots": [ 1, 9 ], "element_scope": "full", "source_facts": [ "muddy footprints as a receding path", "Overturned dining chairs interrupt the middle ground", "disturbed furnishings", "the disturbed interior stays softly recognizable" ], "visual_inferences": [ "The displaced items are randomly scattered, reflecting chaotic movement rather than deliberate arrangement." ], "creative_decisions": [ "Keeping the overturned chairs and mess visible in the background even in tighter shots to maintain the underlying narrative tension." ], "confidence": "high" }, { "element_id": "subdued_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior is enveloped in subdued, low-key ambient lighting typical of dusk. Deep shadows dominate the space, creating a high-contrast, moody atmosphere that dimly outlines the shapes within the room.", "applies_to_shots": [ 1, 5, 9 ], "element_scope": "full", "source_facts": [ "해질 무렵", "Subdued ambient illumination appropriate to a rooftop room at dusk", "Subdued dusk-interior illumination", "Subdued dusk-interior ambient light" ], "visual_inferences": [ "The primary light source is the dim natural light coming from the open window, offering little fill light for the corners." ], "creative_decisions": [ "Using low-key lighting to isolate focal points and enhance the eerie, suspenseful mood of the scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 18, "shot_index": 11, "ref_usage": "", "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": "c2eb74cef0e643bd" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "벽에 꽂힌 피 묻은 사진을 발견하다", "representative_moment": "A tight focus on P02 pinned inside the red circle on the wall, isolating it from the rest of the dark room.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level insert framing with shallow focus and subdued low-key lighting.", "t2i_prompt": "Photorealistic cinematic still. P02, the existing blood-stained photograph at the center of the red circle on the wall, is secured by the pin in the middle-center midground. The vintage, black-and-white P02 print is noticeably crumpled, with its image-bearing front facing the camera. Subdued, low-key ambient dusk lighting envelopes the space, casting deep shadows around the edges, while the surrounding room falls into a soft blur, isolating P02.", "outfit_assignments": [], "source_facts": [ "벽에 또 붉은 원이 그려져있다. 그 가운데 꽂혀있는 피묻은 사진 한 장!", "7-80년대 쯤 보이는 시골 식당과 지붕위에 둥근 원이 설치된 ... 어린 여자애 둘(12살)이 앉아있다.", "해질 무렵" ], "visual_inferences": [ "fixed_element subdued_dusk_lighting: subdued dusk-interior illumination, low-key ambient, deep shadows", "fixed_element bloody_old_photograph: vintage, black-and-white, crumpled, heavy red blood stains" ], "creative_decisions": [ "Applied a shallow depth of field (soft blur on the surrounding room) to strictly isolate P02 and the red circle.", "Used the red circle to provide the primary color accent in the otherwise subdued, shadow-heavy dusk composition." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "prop", "background" ], "owned_object_usage": [ { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bloodstain", "usage_kind": "anchor", "source_phrase": "the existing blood-stained photograph" }, { "owned_token": "box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "circle", "usage_kind": "anchor", "source_phrase": "at the center of the red circle on the wall" }, { "owned_token": "clothing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drawer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "anchor", "source_phrase": "the existing blood-stained photograph" }, { "owned_token": "pin", "usage_kind": "anchor", "source_phrase": "secured by the pin" }, { "owned_token": "storage cabinet", "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": "ab22be0996db5c99", "owned_hash": "74051161e86a2aff", "camera_direction_hash": "5a639334907b6ef3", "owned_usage_hash": "d77564d61fadfe9b", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "photograph", "violating_phrase": "P02, the existing blood-stained photograph at the center of the red circle on the wall, is secured by the pin in the middle-center midground. The vintage, black-and-white P02 print is noticeably crumpled, with its image-bearing front facing the camera.", "reason": "The prompt references the existing photograph (P02) as a visual and composition anchor for focus and framing, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "bloodstain", "violating_phrase": "the existing blood-stained photograph", "reason": "The bloodstain is described as an attribute of the existing photograph to anchor the scene, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "circle", "violating_phrase": "at the center of the red circle on the wall", "reason": "The red circle is referenced solely as a spatial anchor to position the photograph, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "pin", "violating_phrase": "secured by the pin", "reason": "The pin is mentioned to establish the connection and position of the photograph, matching the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 25, "dependency_reason": "Same room, maintaining continuity of the background and dusk lighting from S25_Shot1.", "_shot_index": 5, "visible_entities": [ "P02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Pause the eye-level dolly-in at its tightest practical distance, centering the blood-stained photograph within the red circle while retaining enough surrounding wall to preserve scale. The photograph occupies roughly the central third rather than overwhelming the frame, and shallow focus suppresses the rest of the bedroom before the cut into memory.", "lighting_mood": "Subdued dusk-interior illumination isolates the black-and-white photograph while the red circle provides the restrained crimson accent.", "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": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "blood-stained photograph at the circle center", "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": [ "P02" ], "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", "bloodstain", "box", "circle", "clothing", "curtain", "door", "drawer", "photograph", "pin", "storage cabinet", "wardrobe", "window" ], "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": [ { "element_id": "subdued_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior is enveloped in subdued, low-key ambient lighting typical of dusk. Deep shadows dominate the space, creating a high-contrast, moody atmosphere that dimly outlines the shapes within the room.", "applies_to_shots": [ 1, 5, 9 ], "element_scope": "full", "source_facts": [ "해질 무렵", "Subdued ambient illumination appropriate to a rooftop room at dusk", "Subdued dusk-interior illumination", "Subdued dusk-interior ambient light" ], "visual_inferences": [ "The primary light source is the dim natural light coming from the open window, offering little fill light for the corners." ], "creative_decisions": [ "Using low-key lighting to isolate focal points and enhance the eerie, suspenseful mood of the scene." ], "confidence": "high" }, { "element_id": "bloody_old_photograph", "element_type": "persistent_prop", "character_name": "", "description": "A vintage, black-and-white photograph bearing heavy signs of damage. The paper is noticeably crumpled and wrinkled, with prominent, dark red blood stains smeared across its surface.", "applies_to_shots": [ 5, 9 ], "element_scope": "close", "source_facts": [ "피묻은 사진 한 장", "구겨진 피 묻은 낡은 흑백 사진", "blood-stained photograph", "crumpled photograph" ], "visual_inferences": [ "The edges of the photograph are frayed from age, and the blood stains look dried into the paper." ], "creative_decisions": [ "Ensuring the deep red of the blood contrasts sharply against the desaturated grayscale tones of the old photo for visual impact." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 25, "shot_index": 1, "ref_usage": "", "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": "L04B02", "policy": "required" }, { "kind": "prop", "id": "P02", "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": [ 5 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P02", "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": "30d50f8d3e8332e0" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "엉망이 된 집을 홀린 듯 떠남", "representative_moment": "수리영의 손이 구겨진 사진을 꽉 움켜쥔 타이트 샷", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Track close behind at waist height, isolating the clenched hand in the center-lower frame with subdued dusk-interior ambient light.", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a young Korean woman, with back to camera mid-stride, gaze off-screen, captured from close behind at waist height. Her torso and arms are in frame, her clenched fist isolated in the center-lower frame, tightly gripping P02. P02, a crumpled vintage black-and-white photograph already featuring the pre-existing bloodstains, is held with only part of its image-bearing front visible between her fingers. Subdued dusk-interior ambient light keeps the hand in low-key relief, highlighting the existing bloodstains as a restrained dark accent. The same room forms a softly recognizable background behind her movement toward the exit, where overturned dining chairs and scattered muddy footprints lie in the deep shadows.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "엉망이 된 집 안을 한번 획 둘러보는 수리영, 마치 이끌리듯 백팩을 메고 밖으로 나간다. 사진을 손에 꼭 쥔 채.", "Track close behind and slightly beside 수리영 at waist height, initially isolating her clenched hand as it carries the crumpled photograph near the center-lower frame.", "Subdued dusk-interior ambient light keeps the clenched hand in low-key relief with the photograph's stains as a restrained dark accent." ], "visual_inferences": [ "fixed_element: The interior space is in severe disarray, indicating a hasty search. Dining chairs lie overturned on the floor, and muddy footprints form a scattered path across the ground.", "fixed_element: A vintage, black-and-white photograph bearing heavy signs of damage. The paper is noticeably crumpled and wrinkled, with prominent, dark red blood stains smeared across its surface.", "fixed_element: The interior is enveloped in subdued, low-key ambient lighting typical of dusk. Deep shadows dominate the space." ], "creative_decisions": [ "Framed at waist height to isolate the hand and the photograph in the lower frame, strictly adhering to the spatial limits of an insert shot without fully showing the legs or head.", "Used 'the same room' to anchor the background according to zoom_in_detail continuity constraints, keeping the background softly recognizable to emphasize the tight focus." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background", "prop" ], "owned_object_usage": [ { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bloodstain", "usage_kind": "anchor", "source_phrase": "already featuring the pre-existing bloodstains" }, { "owned_token": "box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "circle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "clothing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drawer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "anchor", "source_phrase": "P02, a crumpled vintage black-and-white photograph" }, { "owned_token": "pin", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage cabinet", "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": "dad37a67cdf9114e", "owned_hash": "74051161e86a2aff", "camera_direction_hash": "839b662dd90c7b3e", "owned_usage_hash": "c9f175be632ec700", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bloodstain", "violating_phrase": "already featuring the pre-existing bloodstains", "reason": "The prompt refers to the pre-existing bloodstains as a background element to highlight without redesigning them, which matches the anchor usage declared by the producer.", "verdict": "anchor_reference" }, { "owned_object": "photograph", "violating_phrase": "P02, a crumpled vintage black-and-white photograph already featuring the pre-existing bloodstains, is held with only part of its image-bearing front visible between her fingers", "reason": "The prompt describes the physical interaction of holding the crumpled photograph (P02) in a clenched fist, satisfying all conditions for allowed prop contact.", "verdict": "allowed_prop_contact" } ] } } ], "scene_type": "normal", "dependent_scene_index": 25, "dependency_reason": "zoom_in_detail and exact_background ref_usage from S25_Shot5 requires keeping the same room environment and lighting consistent.", "_shot_index": 9, "visible_entities": [ "C01", "P02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "shot_index": 9 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Track close behind and slightly beside 수리영 at waist height, initially isolating her clenched hand as it carries the crumpled photograph near the center-lower frame. The photograph remains a small, partially obscured shape within her fist while the disturbed interior stays softly recognizable behind her movement toward the exit.", "lighting_mood": "Subdued dusk-interior ambient light keeps the clenched hand in low-key relief with the photograph's stains as a restrained dark accent.", "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", "P02" ], "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": "L04B02", "reference_usage": "exact_background", "owned_objects": [ "bed", "bloodstain", "box", "circle", "clothing", "curtain", "door", "drawer", "photograph", "pin", "storage cabinet", "wardrobe", "window" ], "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": [ { "element_id": "ransacked_interior_state", "element_type": "environment_state", "character_name": "", "description": "The interior space is in severe disarray, indicating a hasty search. Dining chairs lie overturned on the floor, and muddy footprints form a scattered path across the ground. Various household furnishings are displaced and left in a messy, disturbed condition.", "applies_to_shots": [ 1, 9 ], "element_scope": "full", "source_facts": [ "muddy footprints as a receding path", "Overturned dining chairs interrupt the middle ground", "disturbed furnishings", "the disturbed interior stays softly recognizable" ], "visual_inferences": [ "The displaced items are randomly scattered, reflecting chaotic movement rather than deliberate arrangement." ], "creative_decisions": [ "Keeping the overturned chairs and mess visible in the background even in tighter shots to maintain the underlying narrative tension." ], "confidence": "high" }, { "element_id": "subdued_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior is enveloped in subdued, low-key ambient lighting typical of dusk. Deep shadows dominate the space, creating a high-contrast, moody atmosphere that dimly outlines the shapes within the room.", "applies_to_shots": [ 1, 5, 9 ], "element_scope": "full", "source_facts": [ "해질 무렵", "Subdued ambient illumination appropriate to a rooftop room at dusk", "Subdued dusk-interior illumination", "Subdued dusk-interior ambient light" ], "visual_inferences": [ "The primary light source is the dim natural light coming from the open window, offering little fill light for the corners." ], "creative_decisions": [ "Using low-key lighting to isolate focal points and enhance the eerie, suspenseful mood of the scene." ], "confidence": "high" }, { "element_id": "bloody_old_photograph", "element_type": "persistent_prop", "character_name": "", "description": "A vintage, black-and-white photograph bearing heavy signs of damage. The paper is noticeably crumpled and wrinkled, with prominent, dark red blood stains smeared across its surface.", "applies_to_shots": [ 5, 9 ], "element_scope": "close", "source_facts": [ "피묻은 사진 한 장", "구겨진 피 묻은 낡은 흑백 사진", "blood-stained photograph", "crumpled photograph" ], "visual_inferences": [ "The edges of the photograph are frayed from age, and the blood stains look dried into the paper." ], "creative_decisions": [ "Ensuring the deep red of the blood contrasts sharply against the desaturated grayscale tones of the old photo for visual impact." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 25, "shot_index": 5, "ref_usage": "", "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": "L04B02", "policy": "required" }, { "kind": "prop", "id": "P02", "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": [ 9 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P02", "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": "736aba895c95aa17" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "지친 기다림 속 안개 낀 바다 응시", "representative_moment": "짙은 안개 낀 해질녘 포구, 수리영(C01)이 갯바위에 무릎을 모으고 앉아 하염없이 바다를 응시한다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide profile shot at seated shoulder height, muted natural colors with diffuse dusk lighting", "t2i_prompt": "Photorealistic cinematic still. Observer at seated shoulder height from the side, the figure's full seated silhouette anchored in the middle ground, leaving vast negative space on the right side of the frame. C01O01 in a khaki sweatshirt, a Korean young woman in her 20s with long black hair, seated in right profile on the dark rocky shore, her knees drawn together, her eyes gazing into the distance across the water. The deserted coastal harbor and wet rocky surfaces are covered in dense, thick fog at dusk. The sea and distant background boats are heavily obscured by the thick mist. Diffuse dusk light filtered through the fog creates subdued brightness, muted natural colors, and soft low-key separation between the isolated figure and the overwhelming gray landscape.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "비는 그치고 안개가 자욱하다", "텅 빈 포구. 안개로 인해 배는 보이지 않고", "갯바위에 앉아있는 수리영", "수리영이 갯바위 위에 무릎을 모으고 앉아 바다를 응시하는 측면" ], "visual_inferences": [ "fixed_element foggy_harbor_at_dusk: deserted harbor, thick fog obscuring distant elements", "wet rocky surfaces reflecting the recent heavy rain", "soft low-key separation highlighting the isolated figure against the fog" ], "creative_decisions": [ "wide framing to emphasize vast negative space and her stranded feeling within the harbor", "seated shoulder height observer position to align the camera with her gaze across the water" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "4b6f5a677ead064a", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "3f178c145e649c9a", "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": [ "C01", "L17" ], "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": "Hold static at 수리영's seated shoulder height from her side, keeping her gathered posture in the middle ground and the fog-covered sea extending across her gaze line. Frame wide enough to include the 갯바위 beneath her and negative space over the empty harbor, making her stillness feel stranded within the landscape.", "lighting_mood": "Diffuse dusk light filtered through dense fog creates subdued brightness, muted natural color, and soft low-key separation.", "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", "L17" ], "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": "foggy_harbor_at_dusk", "element_type": "environment_state", "character_name": "", "description": "A deserted coastal harbor and rocky shore covered in dense, thick fog at dusk. The overall atmosphere is muted with subdued natural colors and low-contrast diffuse lighting filtering through the heavy mist. The sea, background boats, and distant landscape are heavily obscured by the fog.", "applies_to_shots": [ 2, 5, 9 ], "element_scope": "full", "source_facts": [ "해질 무렵", "비는 그치고 안개가 자욱하다", "텅 빈 포구", "안개로 인해 배는 보이지 않고" ], "visual_inferences": [ "The ground and rocks remain slightly wet and dark from the recently stopped heavy rain.", "The lighting is flat and grayish due to the thick fog absorbing the remaining dusk light." ], "creative_decisions": [ "Use muted color palettes and soft focus on background elements to emphasize the feeling of isolation and vastness.", "Maintain low-key separation to ensure characters stand out against the overwhelming gray fog." ], "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": "4e53a81395d9e436" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "안개를 뚫고 도착한 배와 닻줄을 고정하는 인우", "representative_moment": "인우(C03)가 뱃머리에서 내려 말뚝에 굵은 닻줄(P08)을 팽팽하게 당기며 멈춰선 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Waist-height medium shot angled toward the boat's bow, diffuse dusk light in dense fog", "t2i_prompt": "Photorealistic cinematic still. Observed at waist height beside the mooring point, torso and arms occupying the middle third of the frame. C03O03 in a windproof fisherman jacket, an East Asian young man with natural black hair and balanced fine features, stands at three-quarter angle from the right, pausing mid-step with one foot planted on the dark, wet rocky shore, looking down at the lower foreground. His both hands are firmly tensioning P08, a thick mooring rope. The rope cuts a strong diagonal line from his gripped hands down toward a wooden mooring post positioned at the lower frame edge. Behind him, the wooden bow of the small boat is visible, angled partly toward the camera. The surrounding L17 is a deserted coastal harbor covered in dense thick fog at dusk. Diffuse dusk light filters through the heavy mist, casting flat, subdued natural colors with low-key separation, keeping the rope work grounded in restrained contrast.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O03" } ], "source_facts": [ "해질 무렵", "비는 그치고 안개가 자욱하다", "텅 빈 포구. 안개로 인해 배는 보이지 않고.", "배에서 내린 인우(24세. 남)는 닷줄을 고정시키고 있다." ], "visual_inferences": [ "fixed_element: A deserted coastal harbor and rocky shore covered in dense, thick fog at dusk.", "The ground and rocks remain slightly wet and dark from the recently stopped heavy rain.", "The mooring rope creates a strong diagonal line from the character's hands to the post in the lower foreground.", "The front of the boat anchors the background space behind the character for physical continuity." ], "creative_decisions": [ "Waist-height camera framing to emphasize the physical tension in the arms and the rope's diagonal placement.", "Muted color palette and low-contrast diffuse lighting to heighten the atmospheric density of the heavy fog." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "102b2063cfc126df", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "0156f7d16a1dc75a", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "독립적인 시공간의 일반 씬이나, 직전 샷(S26_Shot2)의 안개 낀 포구 배경 설정을 동일하게 유지함", "_shot_index": 5, "visible_entities": [ "C03", "L17", "P08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At waist height beside the mooring point, angle toward the boat's bow as 인우 pauses with one foot newly down and both hands tensioning the thick rope around the post. Keep 인우 in the middle third, the rope cutting diagonally toward the lower foreground post, and the front of the boat visible behind him for clear physical continuity.", "lighting_mood": "Diffuse dusk light in the fog keeps the rope work grounded and tactile with muted natural colors and restrained contrast.", "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": [ "C03", "L17", "P08" ], "allowed_outlook_pairs": [ { "character_id": "C03", "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": "foggy_harbor_at_dusk", "element_type": "environment_state", "character_name": "", "description": "A deserted coastal harbor and rocky shore covered in dense, thick fog at dusk. The overall atmosphere is muted with subdued natural colors and low-contrast diffuse lighting filtering through the heavy mist. The sea, background boats, and distant landscape are heavily obscured by the fog.", "applies_to_shots": [ 2, 5, 9 ], "element_scope": "full", "source_facts": [ "해질 무렵", "비는 그치고 안개가 자욱하다", "텅 빈 포구", "안개로 인해 배는 보이지 않고" ], "visual_inferences": [ "The ground and rocks remain slightly wet and dark from the recently stopped heavy rain.", "The lighting is flat and grayish due to the thick fog absorbing the remaining dusk light." ], "creative_decisions": [ "Use muted color palettes and soft focus on background elements to emphasize the feeling of isolation and vastness.", "Maintain low-key separation to ensure characters stand out against the overwhelming gray fog." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 26, "shot_index": 2, "ref_usage": "", "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": "C03O03", "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": "8c9c8b1b5cccbc15" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "음성 메시지를 남기는 수리영", "representative_moment": "C01이 바다를 향해 선 채 입가에 P01을 대고 음성 메시지를 남긴다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium side framing at chest height, soft dusk illumination diffused by heavy fog with muted colors", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, an Asian young woman with long black hair, stands in left profile on the right side of the frame. Her torso and head are visible against the soft dusk illumination. She is holding P01, a modern smartphone, raised near her lips with the screen facing her, gazing into the distant fog across the open sea that occupies the left negative space. A deserted coastal harbor covered in dense, thick fog stretches behind her, the heavy mist diffusing the subdued, flat grayish light and obscuring the distant boats and rocky shore.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 떠나기전, 바다를 바라보며 음성메시지를 남긴다.", "해질 무렵", "비는 그치고 안개가 자욱하다" ], "visual_inferences": [ "fixed_element foggy_harbor_at_dusk: A deserted coastal harbor and rocky shore covered in dense, thick fog at dusk.", "Hold static at C01's chest height from her side, framing her upper body on one side and leaving open sea space.", "left profile, gazing into the distance", "smartphone screen facing C01 while in use near her mouth" ], "creative_decisions": [ "Placed C01 on the right side of the frame in left profile to allow the misty sea to act as negative space on the left, visually representing her distant gaze.", "Used subdued low-key lighting and flat grayish tones to match the heavy fog and quiet tension." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "f7728830b209667b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "cdbef10ef33ed022", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 26, "dependency_reason": "이전 샷의 안개 낀 포구 배경과 해질녘 조명 상태를 연속성 있게 유지", "_shot_index": 9, "visible_entities": [ "C01", "L17", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 9 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Hold static at 수리영's chest height from her side, framing her upper body on one side and leaving open sea space in the direction of her gaze. Her smartphone remains close to her mouth near the center of the composition, while the fogged harbor recedes behind her profile.", "lighting_mood": "Soft dusk illumination diffused by fog holds 수리영 in subdued profile with muted color and quiet low-key tension.", "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", "L17", "P01" ], "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": "foggy_harbor_at_dusk", "element_type": "environment_state", "character_name": "", "description": "A deserted coastal harbor and rocky shore covered in dense, thick fog at dusk. The overall atmosphere is muted with subdued natural colors and low-contrast diffuse lighting filtering through the heavy mist. The sea, background boats, and distant landscape are heavily obscured by the fog.", "applies_to_shots": [ 2, 5, 9 ], "element_scope": "full", "source_facts": [ "해질 무렵", "비는 그치고 안개가 자욱하다", "텅 빈 포구", "안개로 인해 배는 보이지 않고" ], "visual_inferences": [ "The ground and rocks remain slightly wet and dark from the recently stopped heavy rain.", "The lighting is flat and grayish due to the thick fog absorbing the remaining dusk light." ], "creative_decisions": [ "Use muted color palettes and soft focus on background elements to emphasize the feeling of isolation and vastness.", "Maintain low-key separation to ensure characters stand out against the overwhelming gray fog." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 26, "shot_index": 2, "ref_usage": "", "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": "8ed0be7f531fad56" }, { "scene_index": 27, "heading": "S#27. 옥탑방 안 / 해질 무렵", "beat_title": "수리영의 녹음메시지 확인", "representative_moment": "혜수가 미간을 깊게 찌푸린 채 스마트폰을 내려다본다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level tight portrait framing with shallow depth of field and subdued low-key dusk lighting", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a dark leather detective jacket, a Korean middle-aged woman, facing the camera. Positioned directly before her at eye level, her deeply furrowed face occupies most of the frame in sharp shallow focus. She is looking down, shoulders curled, her eyes leading toward a smartphone held along the lower edge. The screen side of the phone faces her and angles away from the view, leaving its displayed content unreadable. [: soft, out-of-focus dim wall surface immediately behind her]. Subdued low-key dusk ambient light wraps her rigid expression in muted natural colors, isolating the tension between her brow and the unseen screen.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "S#27 action: '수리영의 녹음메시지를 확인하는 혜수'", "shot description: 혜수가 미간을 깊게 찌푸린 채 스마트폰을 내려다보는 얼굴 클로즈업", "DP directive: At eye level and tight portrait distance directly before 혜수... deeply furrowed face occupying most of the frame", "DP lighting: Dusk ambient light... subdued and low-key" ], "visual_inferences": [ "smartphone held along the lower edge with screen facing away from camera, maintaining spatial logic without revealing content", "background details are heavily blurred into a soft dim wall surface due to close framing scale" ], "creative_decisions": [ "isolated face composition against a shallow background to emphasize the psychological tension", "muted natural color palette to match the serious tone of the voice message discovery" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "TV stand", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "belonging", "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": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "footprint", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stovetop", "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": "032da9babd4f5985", "owned_hash": "19affc080f4abd6d", "camera_direction_hash": "1d47b417fea623f7", "owned_usage_hash": "4437c36fb7dc16a7", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 25, "dependency_reason": "S25_Shot5 provides exact background continuity for the rooftop room, though it is rendered as a soft out-of-focus frame-edge surface in this shot due to the skipped_close_framing mode.", "_shot_index": 3, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 27, "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 eye level and tight portrait distance directly before 혜수, the camera settles from its upward track with her deeply furrowed face occupying most of the frame. Her lowered eyes lead toward the smartphone held along the lower edge, while shallow focus isolates the tension between her brow and the unseen screen content.", "lighting_mood": "Dusk ambient light appropriate to the rooftop room is kept subdued and low-key, holding 혜수's rigid expression in muted natural color.", "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": "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": [], "previous_shot_refs": [ { "scene_index": 25, "shot_index": 5, "ref_usage": "", "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": "C02O02", "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": "a5d93da7324d12a5" }, { "scene_index": 28, "heading": "S#28. 바다 + 갑판 / 해질 무렵", "beat_title": "사진을 보고 경악하는 인우", "representative_moment": "인우(C03)가 사진을 보고 두 눈을 크게 뜨며 몸을 뒤로 뺀다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium framing from upper-torso height with subdued dusk lighting", "t2i_prompt": "Photorealistic cinematic still. Shot at upper-torso height beside the subject, head and chest in the frame, lower body implied beyond the bottom edge. C03O03 in a dark windproof jacket, an East Asian man in his 20s, occupies the center-left in right profile, chest recoiled backward, widened eyes gazing off-screen toward the right. In the lower-right foreground, the photograph remains an oblique foreground cue, its photographic face angled toward him and seen obliquely by the camera. The narrow deck behind him provides a spatial pathway toward the entrance door of the enclosed wheelhouse. The natural environment is illuminated by subdued dusk lighting, casting a dim, muted twilight tone over the sea and the boat.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O03" } ], "source_facts": [ "S#28 action: 수리영 shows the photograph from scene 25, and 인우 is shocked ('몰라. 모른다구.').", "shot description: 인우가 사진에 시선을 고정한 채 두 눈을 크게 뜨고 상체를 뒤로 뺀 측면 구도.", "DP instruction: Close beside 인우 at upper-torso height, his widened eyes read in profile as his chest recoils away from the photograph." ], "visual_inferences": [ "fixed_element dusk_ambient_light: Subdued dusk lighting casts a dim, muted twilight tone over the boat.", "fixed_element fishing_boat_deck_layout: The narrow deck connects to the entrance door of the control room cabin." ], "creative_decisions": [ "Adopted a medium framing from upper-torso height to tightly capture C03's widened eyes and recoil in profile.", "Positioned the photograph obliquely in the foreground to anchor his gaze and reaction.", "Utilized the subdued dusk lighting to enhance the pale tension on his face against the dim background." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "deck hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "entrance door of the enclosed wheelhouse" }, { "owned_token": "paper map", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "redraw", "source_phrase": "the photograph remains an oblique foreground cue, its photographic face angled toward him and seen obliquely by the camera" }, { "owned_token": "safety rail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheelhouse", "usage_kind": "anchor", "source_phrase": "entrance door of the enclosed wheelhouse" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "3260669bc807c69a", "owned_hash": "fa489f6d87863044", "camera_direction_hash": "00ad4fe50b99d04e", "owned_usage_hash": "7ac256587307ff10", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "entrance door of the enclosed wheelhouse", "reason": "The entrance door is mentioned purely as a spatial composition anchor behind the character, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "wheelhouse", "violating_phrase": "enclosed wheelhouse", "reason": "The wheelhouse is referenced solely as a spatial background anchor to define the pathway, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "photograph", "violating_phrase": "the photograph remains an oblique foreground cue, its photographic face angled toward him and seen obliquely by the camera", "reason": "The prompt specifies only the photograph's spatial placement and camera framing in the foreground rather than instructing a redraw or modification of its visual content, making it an anchor reference.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 25, "dependency_reason": "25씬에 등장한 사진을 수리영이 다시 보여주며 인우의 반응을 이끌어냄", "_shot_index": 5, "visible_entities": [ "C03", "L19" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 28, "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 인우 at upper-torso height, the camera holds perpendicular to his sightline so his widened eyes read in profile as his chest recoils away from the photograph. He occupies the center-left while the photograph remains a small oblique foreground cue at frame right, preserving the route behind him toward the control room for the coming move.", "lighting_mood": "Natural dusk light over the deck is subdued, keeping the palette muted while emphasizing the sudden pallor and tension in 인우's profile.", "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": [ "C03", "L19" ], "allowed_outlook_pairs": [ { "character_id": "C03", "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": "background_ref_attached", "bg_id": "L19B01", "reference_usage": "exact_background", "owned_objects": [ "deck hatch", "door", "paper map", "photograph", "safety rail", "wheelhouse", "window" ], "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": [ { "element_id": "dusk_ambient_light", "element_type": "environment_state", "character_name": "", "description": "The natural environment is illuminated by subdued dusk lighting. The ambient light casts a dim, muted twilight tone over the boat and the surrounding sea, creating a dark and atmospheric palette without bright or direct artificial light sources.", "applies_to_shots": [ 5, 8 ], "element_scope": "full", "source_facts": [ "Natural dusk light over the deck is subdued", "keeping the palette muted", "Dusk ambient light from outside" ], "visual_inferences": [ "The sky is darkening with cool twilight hues, casting deep shadows over the boat's exterior." ], "creative_decisions": [ "Utilize the dim ambient lighting to emphasize the tense, isolating mood of the characters out on the open water." ], "confidence": "high" }, { "element_id": "fishing_boat_deck_layout", "element_type": "environment_state", "character_name": "", "description": "The functional exterior deck of a small fishing boat floating on the sea. The narrow deck provides a direct spatial pathway that connects the open exterior space to the entrance doorway of the boat's enclosed control room cabin.", "applies_to_shots": [ 5, 8 ], "element_scope": "full", "source_facts": [ "preserving the route behind him toward the control room", "the deck and sea remaining visible around her silhouette", "across the entrance" ], "visual_inferences": [ "The boat has a utilitarian design with minimal decorative features, featuring a visible threshold between the outer deck and the inner control room." ], "creative_decisions": [ "Keep the spatial relationship between the deck and the control room tight and confined to visually support the physical blocking action that occurs in the scene." ], "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": "C03O03", "policy": "required" }, { "kind": "background", "id": "L19B01", "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": "b3da5921fcc14590" }, { "scene_index": 28, "heading": "S#28. 바다 + 갑판 / 해질 무렵", "beat_title": "조타실 입구를 막아선 수리영", "representative_moment": "C01이 조타실 입구를 두 팔을 벌려 막아서며 인우의 탈출을 차단하는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide full-body shot from chest height looking outward, muted dusk twilight lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a Korean early 20s woman with long black hair, an oval youthful face, soft facial features and a natural skin tone, stands in the middle-center midground, blocking the control room entrance also in the middle-center midground. Shot from inside the wheelhouse at chest height looking outward, her full body fits within the wide framing as she is facing the camera. Her arms are spread wide, bracing the sides of the door threshold to cut off the escape route, and her eyes are fixed fiercely forward. Through the open entrance behind her, the functional exterior deck of the small fishing boat and the surrounding sea provide a direct spatial pathway to the outside. The natural environment is illuminated by subdued dusk lighting, casting a dim, muted twilight tone over the boat without bright artificial light sources. The dusk ambient light from outside traces a rim-lit outline along her shoulders and extended arms, anchoring her silhouetted volume against the darker interior.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 조타실로 따라들어가 막으면서", "수리영 plants herself centrally across the entrance", "Her extended arms reach toward both sides of the opening", "The deck and sea remaining visible around her silhouette" ], "visual_inferences": [ "fixed_element dusk_ambient_light", "fixed_element fishing_boat_deck_layout", "Camera positioned inside looking out to capture the blocked path", "She braces the door frame to physically block passage" ], "creative_decisions": [ "Wide framing from chest height to ensure her entire blocking posture and the background deck are geometrically consistent in frame", "Rim-lighting on her arms and shoulders to preserve body volume and separate her dark shape from the dim interior" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "deck", "usage_kind": "anchor", "source_phrase": "functional exterior deck of the small fishing boat" }, { "owned_token": "deck hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "sides of the door threshold" }, { "owned_token": "helm console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steering wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheelhouse", "usage_kind": "anchor", "source_phrase": "inside the wheelhouse" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "windscreen", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1add88bd5dc3f75e", "owned_hash": "4a17a723881eb667", "camera_direction_hash": "231a7e225dec2ce2", "owned_usage_hash": "4f0b8ce7896550c8", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "deck", "violating_phrase": "exterior deck of the small fishing boat", "reason": "The exterior deck is mentioned as a background element to establish the spatial pathway, which is an anchor reference with no redraw instruction, matching the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "sides of the door threshold", "reason": "The door threshold is used as a spatial anchor for the character's bracing action without instructions to modify or redraw the door itself, aligning with the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "wheelhouse", "violating_phrase": "inside the wheelhouse", "reason": "The wheelhouse is used strictly as a camera position anchor to define the shot's perspective, matching the producer's echo.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 28, "dependency_reason": "동일 씬 앞 샷(Shot 5)의 해질녘 어선 갑판 및 조명 연속성(fixed_elements)을 직접적으로 이어받아 동일 공간의 다른 앵글을 유지", "_shot_index": 8, "visible_entities": [ "C01", "L19" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 28, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From inside the control room at chest height, the camera looks outward in a static full-body composition as 수리영 plants herself centrally across the entrance. Her extended arms reach toward both sides of the opening, with the deck and sea remaining visible around her silhouette to clarify that she has physically cut off 인우's escape route.", "lighting_mood": "Dusk ambient light from outside outlines 수리영 against the more subdued control-room interior without introducing an additional light 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": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "control room entrance", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영 blocking the entrance", "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", "L19" ], "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": "L19B02", "reference_usage": "exact_background", "owned_objects": [ "deck", "deck hatch", "door", "helm console", "rail", "steering wheel", "wheelhouse", "window", "windscreen" ], "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": [ { "element_id": "dusk_ambient_light", "element_type": "environment_state", "character_name": "", "description": "The natural environment is illuminated by subdued dusk lighting. The ambient light casts a dim, muted twilight tone over the boat and the surrounding sea, creating a dark and atmospheric palette without bright or direct artificial light sources.", "applies_to_shots": [ 5, 8 ], "element_scope": "full", "source_facts": [ "Natural dusk light over the deck is subdued", "keeping the palette muted", "Dusk ambient light from outside" ], "visual_inferences": [ "The sky is darkening with cool twilight hues, casting deep shadows over the boat's exterior." ], "creative_decisions": [ "Utilize the dim ambient lighting to emphasize the tense, isolating mood of the characters out on the open water." ], "confidence": "high" }, { "element_id": "fishing_boat_deck_layout", "element_type": "environment_state", "character_name": "", "description": "The functional exterior deck of a small fishing boat floating on the sea. The narrow deck provides a direct spatial pathway that connects the open exterior space to the entrance doorway of the boat's enclosed control room cabin.", "applies_to_shots": [ 5, 8 ], "element_scope": "full", "source_facts": [ "preserving the route behind him toward the control room", "the deck and sea remaining visible around her silhouette", "across the entrance" ], "visual_inferences": [ "The boat has a utilitarian design with minimal decorative features, featuring a visible threshold between the outer deck and the inner control room." ], "creative_decisions": [ "Keep the spatial relationship between the deck and the control room tight and confined to visually support the physical blocking action that occurs in the scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 28, "shot_index": 5, "ref_usage": "", "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": "L19B02", "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": "f85221e4dd11e740" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "사진 속 소녀가 엄마임을 확인하는 수리영", "representative_moment": "C01의 손가락이 P02 사진 속 아이의 얼굴을 가리키는 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Insert framing just above hand level angling down, narrow focus plane on the fingertip and photo", "t2i_prompt": "Photorealistic cinematic still. Just above hand level, the camera angles down on C01 in a khaki sweatshirt, an Asian woman in her 20s, her pointing hand positioned in the middle-left foreground, pointing toward the child's face in the photograph. The index finger of C01 enters from the left edge and touches the center of P02, a hand-held, crumpled, and bloodstained old black-and-white photographic print. The aged image side of P02 faces the camera, revealing the printed face of a 12-year-old Korean girl exactly where the fingertip meets the paper. The fingertip and the indicated face share a narrow focus plane, while the photograph remains below forty percent of the frame and the surrounding hand provides a human scale reference. Subdued dusk ambient lighting inside the control room maintains a muted color palette and soft shadows, casting a consistently dim atmosphere across the space and keeping the photograph legible against a soft, out-of-focus background.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "C01's index finger touches the child's face in the aged photograph", "P02 is a crumpled, bloodstained old black-and-white photo", "Camera angles down just above hand level, index finger enters from left" ], "visual_inferences": [ "fixed_element control_room_dusk_lighting: subdued dusk ambient lighting, muted colors", "Insert scale requires shallow depth of field, keeping background elements blurred and out of focus" ], "creative_decisions": [ "Framed the photograph exactly facing the camera to emphasize the printed face beneath the finger", "Used a narrow focus plane to direct tension purely onto the interaction between the finger and the paper" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character", "prop", "background" ], "owned_object_usage": [ { "owned_token": "compass", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doorway", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gauge", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photograph", "usage_kind": "redraw", "source_phrase": "the child's face in the photograph" }, { "owned_token": "seat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steering wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "4ed951dcccc90d3c", "owned_hash": "e11c8243e2c2a449", "camera_direction_hash": "13d8beedb0e51a49", "owned_usage_hash": "d9acf27b4efbe7d1", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "photograph", "violating_phrase": "revealing the printed face of a 12-year-old Korean girl exactly where the fingertip meets the paper", "reason": "The printed content on the photograph is explicitly required by the shot_intent to depict the character pointing to the child's face in the photo, satisfying the allowed_visual_state_change exception.", "verdict": "allowed_visual_state_change" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct visual dependency on prior scenes for this specific insert moment", "_shot_index": 3, "visible_entities": [ "C01", "P02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 29, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Just above hand level, the camera angles down so 수리영's index finger enters from the left and meets the child's face near the center of the aged photograph. The fingertip and indicated face share the narrow focus plane, while the photograph remains below forty percent of the frame and the surrounding hand provides a human scale reference.", "lighting_mood": "Subdued dusk ambient light inside the control room keeps the photograph legible while preserving muted color and low-key tension.", "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": "points_to_anchor", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "points_to", "gesture_target_label": "child's face in the photograph", "label": "수리영's pointing hand", "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", "P02" ], "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 insert identifies 수리영 only through her pointing hand, so an outlook reference would not align reliably with the visible anatomy." } ], "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": "L20B01", "reference_usage": "exact_background", "owned_objects": [ "compass", "console", "doorway", "gauge", "photograph", "seat", "steering wheel", "window" ], "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": [ { "element_id": "control_room_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the control room is illuminated by subdued dusk ambient lighting. The low-key light maintains a muted color palette and soft shadows, establishing a consistently dim atmosphere across the space.", "applies_to_shots": [ 3, 4, 8 ], "element_scope": "full", "source_facts": [ "Subdued dusk ambient light inside the control room (Shot 3)", "Subdued dusk light (Shot 4)", "Low-key dusk ambient light in the control room (Shot 8)" ], "visual_inferences": [ "The lighting casts gentle shadows and keeps overall scene saturation low" ], "creative_decisions": [ "Maintaining a consistent low-light mood to emphasize dramatic tension over bright environmental details" ], "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": "L20B01", "policy": "required" }, { "kind": "prop", "id": "P02", "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": [ 3 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P02", "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": "498aeee2d088c4ff" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "사색이 된 인우와 붉은 원 환시", "representative_moment": "충격으로 하얗게 질려 입술을 벌린 C03의 얼굴이 프레임을 가득 채우고, 등 뒤의 흐릿한 공간에 붉은 원이 어른거린다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level close framing with a tight dolly-in feel, low-key dusk lighting and desaturated tones", "t2i_prompt": "Photorealistic cinematic still. C03O03 in a dark windproof fisher jacket, a Korean mid-20s man, facing the camera, occupying most of the frame from the neck up. His neck is recoiled, lips slightly parted, and his eyes are fixed just off the lens, looking off-screen. His face maintains a drained, starkly pale complexion visibly lacking in color. Subdued dusk ambient light casts gentle, soft shadows across his skin, keeping the overall scene saturation low. Immediately behind his head, the close frame-edge surface is softly out of focus, revealing dim circular forms including a restrained red circle acting as a muted accent against the dark void.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O03" } ], "source_facts": [ "사색이 된 인우 (Scene text)", "his pale face and parted lips (Shot description)", "red circle held as a restrained background accent (Shot description)", "facing the camera, neck recoiled, lips parted, gaze just off the lens (DP directions)" ], "visual_inferences": [ "fixed_element control_room_dusk_lighting: subdued dusk ambient lighting, low-key light, soft shadows", "fixed_element inu_pale_complexion: pale, drained complexion lacking in color", "the circular forms are rendered softly out of focus due to the close framing on his face" ], "creative_decisions": [ "Strict close framing isolates the character's psychological shock, removing the wider room context as dictated by the skipped background policy", "Desaturated skin tone contrasting slightly with the muted red accent in the background to build visual tension" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "circle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "compass", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gauge", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "helm console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "helm seat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "speaker", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steering wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ventilation grille", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "ac12a24e5feb1dfc", "owned_hash": "b247148db2a1a85c", "camera_direction_hash": "fe22b4900262acf8", "owned_usage_hash": "9b57691b49e04b1b", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 29, "dependency_reason": "Carries over the subdued dusk lighting and Inu's pale, drained complexion established in the previous shots of the same location.", "_shot_index": 4, "visible_entities": [ "C03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 29, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At 인우's eye level directly before him, the dolly-in tightens slightly until his pale face and parted lips occupy most of the frame. His gaze passes just off the lens toward 수리영, while the control room's circular forms remain soft behind him and prepare the lateral transition beyond his face.", "lighting_mood": "Subdued dusk light keeps 인우's face drained and low in saturation, with the red circle held as a restrained background accent.", "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": [ "C03" ], "allowed_outlook_pairs": [ { "character_id": "C03", "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": "skipped_close_framing", "bg_id": "L20B02", "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": "control_room_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the control room is illuminated by subdued dusk ambient lighting. The low-key light maintains a muted color palette and soft shadows, establishing a consistently dim atmosphere across the space.", "applies_to_shots": [ 3, 4, 8 ], "element_scope": "full", "source_facts": [ "Subdued dusk ambient light inside the control room (Shot 3)", "Subdued dusk light (Shot 4)", "Low-key dusk ambient light in the control room (Shot 8)" ], "visual_inferences": [ "The lighting casts gentle shadows and keeps overall scene saturation low" ], "creative_decisions": [ "Maintaining a consistent low-light mood to emphasize dramatic tension over bright environmental details" ], "confidence": "high" }, { "element_id": "inu_pale_complexion", "element_type": "character_state", "character_name": "인우", "description": "The man maintains a pale, drained complexion on his face. His skin tone is visibly lacking in color, physically reflecting his state of sudden, deep psychological shock.", "applies_to_shots": [ 4, 8 ], "element_scope": "close", "source_facts": [ "사색이 된 인우 (Scene text)", "his pale face and parted lips (Shot 4)", "face drained and low in saturation (Shot 4)" ], "visual_inferences": [ "The stark paleness of his face persists into the following moment as his reaction continues unabated" ], "creative_decisions": [ "Focusing on his desaturated skin tone to visually manifest his internal fear and realization" ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 29, "shot_index": 3, "ref_usage": "", "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": "C03O03", "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": "ea9c436e48be76cc" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "의식을 잃는 수리영과 당황한 인우", "representative_moment": "의식이 혼미해진 C01이 C03의 어깨에 쓰러지며 맞닿는 찰나.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "At shoulder height close framing, shallow focus on the physical contact point, low-key dusk ambient light", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a khaki sweatshirt, a Korean young woman with her forehead touching C03O03's shoulder, is positioned in the middle-left foreground. C03O03 in a windproof jacket, a Korean young man supporting the contact point, is framed in the middle-right foreground. At shoulder height beside him, C03O03 is seen from over the shoulder, his shoulder braced under impact, his pale and drained complexion visible in partial profile with eyes fixed on C01O01. C01O01 appears at a three-quarter angle from the right, her forehead pressed against his shoulder, looking down with half-closed eyes. The tight close framing isolates the point of contact, keeping only C01O01's brow and C03O03's shoulder fully visible in shallow focus, while the rest of their bodies extend beyond the frame edges. The background surface is heavily blurred, illuminated by [: subdued dusk ambient lighting, maintaining a dim, low-key muted atmosphere with soft shadows].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C03", "outlook_id": "O03" } ], "source_facts": [ "의식이 혼미해지는 수리영. 수리영을 노려보는 인우의 어깨에 그대로 쓰러진다.", "Her half-closed eyes and brow occupy the center-left while C03's shoulder and partial profile anchor the center-right.", "Low-key dusk ambient light in the control room is held soft and subdued." ], "visual_inferences": [ "fixed_element inu_pale_complexion: pale, drained complexion reflecting shock", "fixed_element control_room_dusk_lighting: subdued dusk ambient lighting maintaining a dim atmosphere" ], "creative_decisions": [ "Applying shallow depth of field to heavily blur the background and strictly isolate the contact point between C01's forehead and C03's shoulder", "Using partial profile and close-up framing to avoid rendering full bodies, strictly adhering to close framing rules" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "circle", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "compass", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gauge", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "helm console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "helm seat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "speaker", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "steering wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ventilation grille", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "3f8ecf5ecc8ab24a", "owned_hash": "b247148db2a1a85c", "camera_direction_hash": "8f839d83562bdeb4", "owned_usage_hash": "9b57691b49e04b1b", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 29, "dependency_reason": "같은 공간(조타실)의 해질녘 조명과 인우의 창백한 얼굴 상태가 연속됨", "_shot_index": 8, "visible_entities": [ "C01", "C03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 29, "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 인우, the handheld camera steadies for the exact instant 수리영's forehead comes to rest against his shoulder. Her half-closed eyes and brow occupy the center-left while 인우's shoulder and partial profile anchor the center-right, keeping both bodies within the same shallow focus plane at the point of contact.", "lighting_mood": "Low-key dusk ambient light in the control room is held soft and subdued, concentrating attention on the physical contact and 수리영's fading awareness.", "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 with forehead touching 인우's shoulder", "screen_zone": "middle_left", "target_id": "C01", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "인우 supporting the contact point", "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": "O01" }, { "character_id": "C03", "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": "skipped_close_framing", "bg_id": "L20B02", "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": "control_room_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the control room is illuminated by subdued dusk ambient lighting. The low-key light maintains a muted color palette and soft shadows, establishing a consistently dim atmosphere across the space.", "applies_to_shots": [ 3, 4, 8 ], "element_scope": "full", "source_facts": [ "Subdued dusk ambient light inside the control room (Shot 3)", "Subdued dusk light (Shot 4)", "Low-key dusk ambient light in the control room (Shot 8)" ], "visual_inferences": [ "The lighting casts gentle shadows and keeps overall scene saturation low" ], "creative_decisions": [ "Maintaining a consistent low-light mood to emphasize dramatic tension over bright environmental details" ], "confidence": "high" }, { "element_id": "inu_pale_complexion", "element_type": "character_state", "character_name": "인우", "description": "The man maintains a pale, drained complexion on his face. His skin tone is visibly lacking in color, physically reflecting his state of sudden, deep psychological shock.", "applies_to_shots": [ 4, 8 ], "element_scope": "close", "source_facts": [ "사색이 된 인우 (Scene text)", "his pale face and parted lips (Shot 4)", "face drained and low in saturation (Shot 4)" ], "visual_inferences": [ "The stark paleness of his face persists into the following moment as his reaction continues unabated" ], "creative_decisions": [ "Focusing on his desaturated skin tone to visually manifest his internal fear and realization" ], "confidence": "medium" } ], "previous_shot_refs": [ { "scene_index": 29, "shot_index": 4, "ref_usage": "", "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": "C03O03", "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": "4483ac62d1032072" }, { "scene_index": 30, "heading": "S#30. 바다 / 해질 무렵", "beat_title": "안개 속으로 사라지는 배", "representative_moment": "A small fishing boat disappears into a thick wall of fog on the open sea under a golden sunset.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "distant wide static position, restrained dusk lighting with warm sunset tones", "t2i_prompt": "Photorealistic cinematic still. [L18: Distant wide view facing the horizon across the open sea. The descending sun casts a long golden reflection toward the lower-center frame, flanked by dark water under restrained dusk lighting. A dense vertical wall of fog rests directly on the surface, muting the golden light.] A small silhouette of a fishing boat entering the opaque fog wall, its bow already obscured by the heavy mist.", "outfit_assignments": [], "source_facts": [ "S#30 action: '서서히 지는 해가 바다위에 황금빛 그림자를 드리운다.'", "S#30 action: '세상과 경계를 지우는 것 같은 짙은 안개 띠.'", "S#30 action: '서서히 안개 안으로 들어가는 인우의 배.'" ], "visual_inferences": [ "fixed_element body_full_pose: The fog forms a distinct, opaque vertical barrier resting directly on the water surface.", "distant wide framing captures the vastness of the sea, the sunset reflection, and the fog boundary" ], "creative_decisions": [ "restrained overall exposure to emphasize the stark contrast between the warm golden reflection path and the cold, imposing fog wall", "rendering the boat as a small silhouette to evoke a sense of mystery and being swallowed by the vast natural elements" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "buoy", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cloud", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "island", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mudflat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sea", "usage_kind": "anchor", "source_phrase": "across the open sea" }, { "owned_token": "shoreline", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sun", "usage_kind": "anchor", "source_phrase": "The descending sun casts a long golden reflection" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "173b682cf1ba0762", "owned_hash": "fe7d97dddfba59d1", "camera_direction_hash": "dca435f7808ae110", "owned_usage_hash": "a0bc3411d45e6887", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "sea", "violating_phrase": "across the open sea", "reason": "The prompt refers to 'the open sea' as a spatial reference for the camera direction, which matches the producer's declaration of an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "sun", "violating_phrase": "The descending sun casts a long golden reflection", "reason": "The prompt describes the sun and its golden reflection to establish lighting and composition, which serves as an anchor reference to the background's existing element, matching the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 1, "visible_entities": [ "L18" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 30, "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 distant wide position above water level, the static camera faces the open sea and horizon with the descending sun set above its long reflection. The golden path runs from the horizon toward the lower center, leaving broad areas of open water on either side before the subsequent horizontal pan.", "lighting_mood": "Golden sunset light stretches across the sea while the overall exposure remains restrained and 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": [ "L18" ], "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": "L18B02", "reference_usage": "exact_background", "owned_objects": [ "buoy", "cloud", "island", "mudflat", "sea", "shoreline", "sun" ], "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": [ { "element_id": "sunset_sea_and_fog_boundary", "element_type": "environment_state", "character_name": "", "description": "An open sea surface under the restrained, subdued lighting of dusk. A lingering golden sunset tone casts a long reflection across the water, contrasting with a thick, dense wall of fog that creates a stark physical boundary where the golden light becomes muted and visibility is heavily obscured.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "서서히 지는 해가 바다위에 황금빛 그림자를 드리운다.", "세상과 경계를 지우는 것 같은 짙은 안개 띠.", "Golden sunset light stretches across the sea while the overall exposure remains restrained and subdued.", "Dusk light becomes muted at the dense fog boundary" ], "visual_inferences": [ "The fog forms a distinct, opaque vertical barrier resting directly on the water surface.", "The water outside the golden reflection path appears dark, emphasizing the fading light." ], "creative_decisions": [ "Maintaining a melancholic and mysterious atmosphere by contrasting the warm trace of the sunset with the cold, imposing presence of the fog wall." ], "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": "background", "id": "L18B02", "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": "ccbb9cc647decd9c" }, { "scene_index": 30, "heading": "S#30. 바다 / 해질 무렵", "beat_title": "안개 속으로 사라지는 어선", "representative_moment": "황금빛 낙조가 드리운 바다 위, 소형 어선의 뱃머리가 짙은 안개 장벽 속으로 절반쯤 잠겨들어간다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "perpendicular side view tracking shot at boat height, wide framing", "t2i_prompt": "Photorealistic cinematic still. A small fishing boat in the middle-left midground, moving toward the dense fog wall in the middle-right background. Seen from a perpendicular side view at boat height, the stern and midsection of the hull remain readable on the left, while the bow crosses the stark physical boundary of the fog and becomes half-obscured. An open sea surface lies below, under the restrained, subdued lighting of dusk. A lingering golden sunset tone casts a long reflection across the dark water, sharply contrasting with the opaque vertical barrier where the golden light becomes muted. Focus is on the wooden hull cutting into the dense fog, marking the crossing into obscurity.", "outfit_assignments": [], "source_facts": [ "서서히 지는 해가 바다위에 황금빛 그림자를 드리운다.", "세상과 경계를 지우는 것 같은 짙은 안개 띠.", "서서히 안개 안으로 들어가는 인우의 배.", "Shot description: 어선의 뱃머리가 짙은 안개벽에 절반쯤 가려진 측면 구도" ], "visual_inferences": [ "fixed_element sunset_sea_and_fog_boundary: The fog forms a distinct, opaque vertical barrier resting directly on the water surface.", "fixed_element sunset_sea_and_fog_boundary: The water outside the golden reflection path appears dark, emphasizing the fading light.", "Perpendicular side view tracking parallel to the vessel to clearly show the bow crossing into obscurity." ], "creative_decisions": [ "Wide framing scale to preserve the boat's scale against the surrounding sea and fog.", "Restrained dusk lighting with a single golden reflection path to maximize the contrast between the sunset and the imposing fog wall." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "antenna", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "buoy", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cabin", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fishing boat", "usage_kind": "anchor", "source_phrase": "A small fishing boat in the middle-left midground" }, { "owned_token": "fog", "usage_kind": "anchor", "source_phrase": "the dense fog wall in the middle-right background" }, { "owned_token": "hull", "usage_kind": "anchor", "source_phrase": "the midsection of the hull remain readable" }, { "owned_token": "mast", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "navigation light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "radar", "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": "An open sea surface lies below" }, { "owned_token": "winch", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "2d9fef2e956fef55", "owned_hash": "e6e9121986510a10", "camera_direction_hash": "ec61b423ad10f27f", "owned_usage_hash": "76aa1e102b15e359", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "fishing boat", "violating_phrase": "A small fishing boat in the middle-left midground", "reason": "The prompt references the existing fishing boat as a compositional anchor to define its position and motion, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "fog", "violating_phrase": "the dense fog wall in the middle-right background", "reason": "The prompt mentions the fog wall as a spatial boundary for composition and lighting reference, aligning with the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "hull", "violating_phrase": "the midsection of the hull remain readable", "reason": "The prompt refers to the hull of the existing boat to describe its partial visibility through the fog, acting as an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "sea", "violating_phrase": "An open sea surface lies below", "reason": "The prompt references the sea surface as a background environment element to describe dusk lighting reflections, aligning with the anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 30, "dependency_reason": "동일 씬 앞쪽 Shot 1의 바다 및 안개 배경 톤과 정확한 공간적 연속성을 유지하기 위함 (ref_usage: exact_background).", "_shot_index": 3, "visible_entities": [ "L18" ], "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": "At boat height and medium-wide distance, the camera tracks parallel to the fishing boat in a perpendicular side view as its bow advances into the dense fog wall. The stern and midsection remain readable toward frame left while the bow crosses center-right and becomes half obscured, preserving the boat's scale against the surrounding sea.", "lighting_mood": "Dusk light becomes muted at the dense fog boundary, retaining only a restrained trace of the scene's golden sunset tone.", "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": "dense fog wall", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "dense fog wall", "label": "small fishing boat", "screen_zone": "middle_left", "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": [ "L18" ], "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": "L18B01", "reference_usage": "exact_background", "owned_objects": [ "antenna", "buoy", "cabin", "deck", "fishing boat", "fog", "hull", "mast", "navigation light", "radar", "railing", "rope", "sea", "winch" ], "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": [ { "element_id": "sunset_sea_and_fog_boundary", "element_type": "environment_state", "character_name": "", "description": "An open sea surface under the restrained, subdued lighting of dusk. A lingering golden sunset tone casts a long reflection across the water, contrasting with a thick, dense wall of fog that creates a stark physical boundary where the golden light becomes muted and visibility is heavily obscured.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "서서히 지는 해가 바다위에 황금빛 그림자를 드리운다.", "세상과 경계를 지우는 것 같은 짙은 안개 띠.", "Golden sunset light stretches across the sea while the overall exposure remains restrained and subdued.", "Dusk light becomes muted at the dense fog boundary" ], "visual_inferences": [ "The fog forms a distinct, opaque vertical barrier resting directly on the water surface.", "The water outside the golden reflection path appears dark, emphasizing the fading light." ], "creative_decisions": [ "Maintaining a melancholic and mysterious atmosphere by contrasting the warm trace of the sunset with the cold, imposing presence of the fog wall." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 30, "shot_index": 1, "ref_usage": "", "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": "L18B01", "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": "10ad757c079c5b6b" } ], "consumed_bg_catalog_hash": "609cf360bee5067f", "consumed_shot_binding_hash": "aa8624298cd2588e" }, "step_id": "scene_detail", "run_id": "90c0d882-0224-4821-aa27-b45c68ac28db", "resolved_model": "gemini-pro", "updated_at": "2026-07-11T02:56:33.886588+00:00", "project_config_snapshot": {} }