{ "status": "completed", "completed_count": 55, "applicable_count": 55, "failed_count": 0, "schema_version": 13, "config_hash": "187e36a52f3faa6a", "data": { "scenes": [ { "scene_index": 1, "heading": "S#1. 숲속 / 해 질 무렵", "beat_title": "남자가 숲속에서 무언가를 피해 필사적으로 도망치고 있는 상태", "representative_moment": "짙은 그림자가 깔린 숲의 수풀을 헤치며 달리는 자세로 한쪽 발이 공중에 떠 있는 찢어진 옷차림의 한국인 남자 전신.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Chest-height wide tracking shot, muted gray-green dusk lighting with bruised-gold highlights", "t2i_prompt": "Photorealistic cinematic still. A tracking camera at chest height retreats along a muddy, uneven narrow forest path. C08O08 in heavily torn clothing deeply stained with dark blood, an Asian man, is captured in mid-stride crashing forward through the dark, tangled wet forest undergrowth. He is positioned in the center, angled at three-quarter left, facing forward with eyes looking off-screen in panic. He is visibly wounded with a pained expression, showing the variant features carried from stable_traits — pale complexion, red irises, sharp teeth visible, and a rough facial expression. He is missing one shoe, leaving his bare foot airborne, and has a large bleeding bite mark on his thigh, his torn clothes flowing back in the direction opposite to his forward motion. The forest floor below him is covered with scattered rough dirt clumps and crushed wet leaves. Foreground branches partially veil the lens at the left and right edges. The entire environment is engulfed in deep shadows lit by muted gray-green dusk light, with small bruised-gold highlights catching the torn fabric and wet leaves. The narrow path runs from the lower center foreground toward the shadowed rear of the frame, disappearing into the thickening dusk forest haze.", "outfit_assignments": [ { "character_id": "C08", "outlook_id": "O08" } ], "source_facts": [ "S#1 action: '우거진 수풀을 헤치며 도망치는. 벗겨진 한쪽 신발. 허벅지에 커다란 이빨 자국. 피가 흐르고.'", "S#1 emotion: '잔뜽 겁에 질린 남자 표정'", "shot description: '짙은 그림자가 깔린 숲의 수풀을 헤치며 달리는 자세로 한쪽 발이 공중에 떠 있는 찢어진 옷차림의 한국인 남자 전신.'" ], "visual_inferences": [ "fixed_element man_injured_outfit: deeply stained dark blood on torn clothing, missing one shoe with bare foot, large bleeding bite mark on the thigh", "fixed_element forest_dusk_ground: scattered rough dirt clumps, crushed wet leaves, deep shadows lit by muted dusk light", "mid-stride freeze preserves the intense forward motion with clothes flowing backward" ], "creative_decisions": [ "Chest-height tracking camera framed wide to capture the frantic movement through the narrow path", "Muted gray-green dusk light with bruised-gold highlights to emphasize the eerie, threatening atmosphere of the forest", "Foreground branches veiling the lens edges to enhance the claustrophobic feeling of being chased" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "4b8ab9ffd8dc3d28", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "99c79b0276464879", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "None", "_shot_index": 1, "visible_entities": [ "C08", "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": "A low, chest-height tracking camera retreats along the narrow forest path about 10 feet ahead of the fleeing man, keeping his full body in frame as he crashes through the undergrowth. Foreground branches partially veil the lens at the left and right edges, while his airborne foot and torn clothes cut through the center of the wide frame with the path receding behind him.", "lighting_mood": "Muted gray-green dusk light is swallowed by deep forest shadows, with small bruised-gold highlights catching torn fabric and wet leaves.", "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": [ "C08", "L01" ], "allowed_outlook_pairs": [ { "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": [], "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_injured_outfit", "element_type": "character_state", "character_name": "Korean man", "description": "The Korean man wears heavily torn clothing that is deeply stained with dark blood. He is missing one shoe, leaving his foot bare, and has a large bleeding bite mark on his thigh.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "찢어진 옷차림의 한국인 남자", "벗겨진 한쪽 신발", "허벅지에 커다란 이빨 자국. 피가 흐르고", "blood darkens the man's clothes" ], "visual_inferences": [ "His bare foot and the bite mark on his thigh remain as constant physical traits even as he falls to the ground." ], "creative_decisions": [ "The blood is rendered as dark and wet to emphasize the freshness of the wound in the dim dusk light." ], "confidence": "high" }, { "element_id": "forest_dusk_ground", "element_type": "environment_state", "character_name": "", "description": "The forest floor is covered with scattered rough dirt clumps, crushed wet leaves, and undergrowth. The entire environment is engulfed in deep shadows lit only by the muted and ashen light of dusk.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "해 질 무렵", "우거진 수풀", "짙은 그림자가 깔린 숲", "dirt clumps and crushed leaves", "Muted gray-green dusk light is swallowed by deep forest shadows" ], "visual_inferences": [ "The consistency of the leaf-littered dirt ground remains identical between the running sequence and the ground-level struggle." ], "creative_decisions": [ "A cold, ashen color palette with heavy shadows is used to amplify the desperate and terrifying atmosphere of the forest." ], "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": "C08O08", "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": "e1472c3a9d46a591" }, { "scene_index": 1, "heading": "S#1. 숲속 / 해 질 무렵", "beat_title": "검은 실루엣의 기습과 제압", "representative_moment": "C10O10 suddenly pins the severely injured C08O08 to the ground in the dusky, shadowed forest.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low observer camera at ground level beside the shoulder, angled slightly upward, medium framing, underlit ashen lighting with a cold rim", "t2i_prompt": "Photorealistic cinematic still. Low observer camera at ground level beside the fallen man's shoulder, angled slightly upward. Scattered rough dirt clumps and mud-streaked crushed wet leaves fill the extreme foreground edge. C08O08 in heavily torn clothing deeply stained with dark blood, an East Asian man, lies prone across the lower frame in left profile, severely injured with a pained expression. His pale skin, rough facial features, red irises, and sharp teeth are visible, gaze directed down at the dirt. A large bleeding bite mark on his thigh and his bare foot, missing one shoe, extend into the frame. C10O10 in a dark ragged coat, a shadow-like figure, crouches over him in the middle ground at a three-quarter angle from the right. C10O10 appears as a dark mass and black humanoid outline with faint edges and no visible facial details, exerting sudden firm force, both hands locked onto C08O08's shoulders, pinning the man down from above against the shared forest floor, gaze fixed firmly on the fallen man. The environment is engulfed in deep forest shadows lit by an underlit ashen dusk. A thin cold rim of light traces C10O10's hands, contrasting with the dark wet blood accent on C08O08's clothes. The dark tree line forms a nearly black shadow mass in the rear background.", "outfit_assignments": [ { "character_id": "C08", "outlook_id": "O08" }, { "character_id": "C10", "outlook_id": "O10" } ], "source_facts": [ "해 질 무렵 우거진 수풀", "벗겨진 한쪽 신발. 허벅지에 커다란 이빨 자국. 피가 흐르고", "잔뜩 겁에 질린 남자 표정", "검은 인영이 남자를 덮친다" ], "visual_inferences": [ "fixed_element man_injured_outfit: heavily torn clothing deeply stained with dark blood, missing shoe and bite mark visible", "fixed_element forest_dusk_ground: rough dirt clumps, crushed wet leaves, deep shadows", "The black silhouette exerts sudden firm force, pinning the man from above in a sudden hard stillness" ], "creative_decisions": [ "Low ground-level camera beside the shoulder angled upward to emphasize the weight of the attacker and tactile ground textures", "Cold rim light focused on the silhouette's gripping hands to highlight the physical contact", "Muted red accent placed exclusively on the blood to contrast heavily with the ashen gray palette" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "6017be787b4a11db", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "d55e08da65ae53a2", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 1, "dependency_reason": "Same continuous scene, exact background and continuity elements carried over from earlier in the sequence.", "_shot_index": 4, "visible_entities": [ "C08", "C10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 1, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera has fallen to ground level beside the man's shoulder, only about 2 feet away, angled slightly upward through dirt clumps and crushed leaves. The fallen man lies across the lower frame while the black silhouette crouches over him in the middle ground, both hands locked onto his shoulders in a sudden, hard stillness.", "lighting_mood": "The image is underlit and ashen, with only a thin cold rim of dusk on the silhouette’s hands and a muted red accent where blood darkens the man’s clothes.", "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": [ "C08", "C10" ], "allowed_outlook_pairs": [ { "character_id": "C08", "outlook_id": "O08" }, { "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_injured_outfit", "element_type": "character_state", "character_name": "Korean man", "description": "The Korean man wears heavily torn clothing that is deeply stained with dark blood. He is missing one shoe, leaving his foot bare, and has a large bleeding bite mark on his thigh.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "찢어진 옷차림의 한국인 남자", "벗겨진 한쪽 신발", "허벅지에 커다란 이빨 자국. 피가 흐르고", "blood darkens the man's clothes" ], "visual_inferences": [ "His bare foot and the bite mark on his thigh remain as constant physical traits even as he falls to the ground." ], "creative_decisions": [ "The blood is rendered as dark and wet to emphasize the freshness of the wound in the dim dusk light." ], "confidence": "high" }, { "element_id": "forest_dusk_ground", "element_type": "environment_state", "character_name": "", "description": "The forest floor is covered with scattered rough dirt clumps, crushed wet leaves, and undergrowth. The entire environment is engulfed in deep shadows lit only by the muted and ashen light of dusk.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "해 질 무렵", "우거진 수풀", "짙은 그림자가 깔린 숲", "dirt clumps and crushed leaves", "Muted gray-green dusk light is swallowed by deep forest shadows" ], "visual_inferences": [ "The consistency of the leaf-littered dirt ground remains identical between the running sequence and the ground-level struggle." ], "creative_decisions": [ "A cold, ashen color palette with heavy shadows is used to amplify the desperate and terrifying atmosphere of the forest." ], "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": "C08O08", "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": "ec6631b5e7116680" }, { "scene_index": 2, "heading": "S#2. 도시 뒷골목 / N", "beat_title": "치명적인 흡혈 공격", "representative_moment": "어두운 골목 전봇대 옆에서 남자가 쓰러져가는 여자의 목을 물어뜯고 있는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "voyeur perspective, close framing from shoulder height, weak jaundiced sodium lighting", "t2i_prompt": "Photorealistic cinematic still. Close framing from shoulder height beside the utility pole, concentrating on the intersection of a neck and a mouth with the upper faces mostly cropped out. C09 in a burgundy cardigan, an East Asian woman, occupies the center-right at a three-quarter angle from the right, her neck arched sideways, severely injured with eyes closed. C08 in a blood-stained hiking jacket, an East Asian man showing pale skin, faint red eyes, and a rough facial expression, enters from the left in left profile, looking down. C08's jaw remains firmly locked onto C09's exposed neck, his sharp teeth embedded deep at the localized bite wound. Thick, dark saturated red blood constantly coats the torn flesh, beading along the skin as a sharp visual accent. [: A damp, grime-streaked utility pole edge bounds the left foreground, and a barely lit stained alley wall extends diagonally away behind the woman's shoulder]. A weak sodium streetlamp casts dim jaundiced light and deep cold blue-gray shadows across the skin, leaving the dark red blood as the single heavily saturated element in the frame.", "outfit_assignments": [ { "character_id": "C08", "outlook_id": "O08" }, { "character_id": "C09", "outlook_id": "O09" } ], "source_facts": [ "weak sodium streetlamp creates jaundiced skin tones", "thick, dark saturated red blood bleeds from the torn flesh", "man's lowered mouth enters from the left, teeth embedded at the wound", "utility pole and alley walls", "woman's exposed neck", "most of both faces are cropped out" ], "visual_inferences": [ "faces mostly cropped out due to close framing on the bite contact point", "utility pole edge compresses the space tightly on the left", "the blood acts as a sharp visual accent against pale skin and deep blue-gray shadows" ], "creative_decisions": [ "close cropping to emphasize the violence of the bite while hiding full identities", "using cold blue-gray shadows to contrast starkly with the jaundiced light and saturated red blood", "voyeuristic shoulder-height angle to heighten the dread" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "5d5b24e531f1815b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "ae53491462be1f74", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No explicit prior visual scene dependency required for this isolated alley attack moment.", "_shot_index": 3, "visible_entities": [ "C08", "C09" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 2, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From shoulder height beside the utility pole, the camera has pushed in tighter than the stage baseline until most of both faces are cropped out. The woman’s exposed neck occupies the center-right of the close frame, while the man's lowered mouth enters from the left, his teeth embedded at the wound and the dark red blood gathering in a sharp, sickening accent.", "lighting_mood": "A weak sodium streetlamp creates jaundiced skin tones and deep black pockets, while the blood reads as the only saturated red in the frame.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C08", "C09" ], "allowed_outlook_pairs": [ { "character_id": "C08", "outlook_id": "O08" }, { "character_id": "C09", "outlook_id": "O09" } ], "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": "base_id_required", "reason": "Only the lowered mouth and partial head are visible in an action close-up, making base identity sufficient while avoiding over-specific facial framing." }, { "subject_id": "C09", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The shot crops out most facial identity and centers on the neck wound and mouth contact, so a full outlook reference would not align reliably." } ], "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": "alley_utility_pole_and_lighting", "element_type": "environment_state", "character_name": "", "description": "A dark city back alley featuring a prominent utility pole. The environment is illuminated by a weak, sickly sodium streetlamp that casts dim, jaundiced light and creates deep, cold blue-gray shadows across the pavement and surrounding walls.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "도심 희미한 가로등 불빛", "전봇대", "weak sodium streetlamp creates jaundiced skin tones", "utility pole and alley walls", "sickly streetlamp falloff" ], "visual_inferences": [ "The utility pole stands near the center of the scene's action, anchoring the visual space of the narrow alley." ], "creative_decisions": [ "Using a high-contrast mix of jaundiced yellow light and cold blue-gray shadows to emphasize a stark, ominous nighttime atmosphere." ], "confidence": "high" }, { "element_id": "korean_woman_bleeding_neck", "element_type": "character_state", "character_name": "한국인 여자", "description": "The Korean woman has a severe, localized bite wound on her exposed neck. Thick, dark saturated red blood constantly bleeds from the torn flesh, creating a sickening and sharp visual accent against her skin.", "applies_to_shots": [ 3, 5 ], "element_scope": "close", "source_facts": [ "목의 살점을 뜯는 이빨", "여자의 목덜미에 깊게 박힌... 붉은 핏자국", "spreading dark blood", "blood gathering in a sharp, sickening accent" ], "visual_inferences": [ "The fatal bite wound is clearly localized on the side of her neck, remaining the primary source of the dark blood pooling around her." ], "creative_decisions": [ "Keeping the dark red blood as the only heavily saturated color on her body to draw immediate attention to the fatal injury." ], "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": "C08", "policy": "required" }, { "kind": "character", "id": "C09", "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": "b89d44abedce0026" }, { "scene_index": 2, "heading": "S#2. 도시 뒷골목 / N", "beat_title": "골목 바닥에 쓰러진 여자", "representative_moment": "피가 고인 골목길 바닥에 싸늘하게 식어가는 C09의 모습을 멀리서 담은 서늘한 전경.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide, low-angle ground-level composition with cold blue-gray night lighting and sickly sodium falloff", "t2i_prompt": "Photorealistic cinematic still. A wide, cold composition from a low observer camera on the alley floor. C09O09 in a burgundy cardigan, an adult Asian woman, lies motionless in right profile, sprawled off-center on the wet-looking, grimy, blood-darkened pavement, eyes closed, lifeless, with black hair splayed. A severe, localized bite wound is visible on her exposed neck, surrounded by thick, dark saturated red blood forming an irregular, subdued but ominous red-black pool. Behind her, a prominent utility pole base, stained and dimly lit, faces the camera from the background-left, partly cropped by darkness. Sooty, damp alley walls taper toward the background, closing the narrow space into empty negative space around her small figure. The environment is washed in thin blue-gray night with a weak, sickly sodium streetlamp falloff, casting dim, jaundiced light on the skin and deep cold blue-gray shadows across the pavement.", "outfit_assignments": [ { "character_id": "C09", "outlook_id": "O09" } ], "source_facts": [ "도심 희미한 가로등 불빛", "푹 힘없이 쓰러지는 여자", "여자의 시신을 비추는 넓고 서늘한 구도", "검붉은 핏자국이 번진 어두운 골목길 바닥" ], "visual_inferences": [ "fixed_element korean_woman_bleeding_neck: severe, localized bite wound on exposed neck, thick dark saturated red blood", "fixed_element alley_utility_pole_and_lighting: weak sickly sodium streetlamp, deep cold blue-gray shadows", "camera positioned low on the ground to emphasize the smallness and abandonment of the corpse in the wide negative space" ], "creative_decisions": [ "Applied a wide framing scale to capture both the isolated body and the receding alley walls for spatial tension", "Contrasted the jaundiced light from the sodium lamp on the wound against the overall cold blue-gray ambient shadows" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "cd78999ace3f8d95", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "c7ac4bf27969c947", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 2, "dependency_reason": "동일 씬 내 이전 샷(Shot 3)의 골목길 배경 및 조명 연속성 유지(exact_background)", "_shot_index": 5, "visible_entities": [ "C09", "L02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 2, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera settles low and several meters back on the alley floor after the dolly-out, leaving a wide, cold composition with the woman’s body small against the pavement. Her sprawled form lies off-center amid spreading dark blood, while the utility pole and alley walls recede into empty negative space around her.", "lighting_mood": "The alley is washed in thin blue-gray night with a sickly streetlamp falloff, leaving the blood as a subdued but ominous red-black pool.", "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", "L02" ], "allowed_outlook_pairs": [ { "character_id": "C09", "outlook_id": "O09" } ], "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": "alley_utility_pole_and_lighting", "element_type": "environment_state", "character_name": "", "description": "A dark city back alley featuring a prominent utility pole. The environment is illuminated by a weak, sickly sodium streetlamp that casts dim, jaundiced light and creates deep, cold blue-gray shadows across the pavement and surrounding walls.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "도심 희미한 가로등 불빛", "전봇대", "weak sodium streetlamp creates jaundiced skin tones", "utility pole and alley walls", "sickly streetlamp falloff" ], "visual_inferences": [ "The utility pole stands near the center of the scene's action, anchoring the visual space of the narrow alley." ], "creative_decisions": [ "Using a high-contrast mix of jaundiced yellow light and cold blue-gray shadows to emphasize a stark, ominous nighttime atmosphere." ], "confidence": "high" }, { "element_id": "korean_woman_bleeding_neck", "element_type": "character_state", "character_name": "한국인 여자", "description": "The Korean woman has a severe, localized bite wound on her exposed neck. Thick, dark saturated red blood constantly bleeds from the torn flesh, creating a sickening and sharp visual accent against her skin.", "applies_to_shots": [ 3, 5 ], "element_scope": "close", "source_facts": [ "목의 살점을 뜯는 이빨", "여자의 목덜미에 깊게 박힌... 붉은 핏자국", "spreading dark blood", "blood gathering in a sharp, sickening accent" ], "visual_inferences": [ "The fatal bite wound is clearly localized on the side of her neck, remaining the primary source of the dark blood pooling around her." ], "creative_decisions": [ "Keeping the dark red blood as the only heavily saturated color on her body to draw immediate attention to the fatal injury." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 2, "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": [], "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": "5cb4489edb595b88" }, { "scene_index": 3, "heading": "S#3. 현장 / D", "beat_title": "뉴스 화면 속 사건 현장을 지휘하는 혜수", "representative_moment": "뉴스 화면 속에서 C02가 폴리스 라인 뒤에서 현장을 향해 꼿꼿하게 팔을 뻗어 지휘하고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest height wide composition, flattened news-video exposure", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a navy safari jacket, a Korean woman in her 40s with sharp eyes and distinct features, stands in the middle-center midground at a three-quarter angle from the left, with her arm fully extended pointing toward the crime scene in the upper-right background, looking off-screen. A Korean male detective in a worn jacket is seen over the shoulder from behind, bracing behind the tape, looking off-screen. Several other Korean male detectives in dark jackets are clustered in formation at a three-quarter angle from the right, forming a tense wall around her, looking off-screen. The composition holds at chest height. A taut police line tape cuts a shallow diagonal across the lower foreground, its printed side facing the camera with warning text partially readable, acting as a shared boundary for the group. The crime scene area behind them is an active controlled space. Daylight is flattened by news-video exposure, the visible image surface carrying slightly compressed field footage with faint broadcast softness. Muted grays, procedural blue shadows, and a restrained red accent from small evidence markers near the controlled area complete the scene.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "Shot 2: 사건 현장을 향해 팔을 꼿꼿하게 뻗은 자세로 멈춰 선 혜수와 그 주위를 둘러싼 형사들의 굳건한 모습.", "지역 뉴스 화면. 기자는 보이지 않고 멘트만 흘러나온다." ], "visual_inferences": [ "Korean male detectives in dark jackets clustered in formation to block access to the scene", "Police tape acting as a shared physical boundary separating the public view from the active scene" ], "creative_decisions": [ "Applied muted grays, procedural blue shadows, and faint broadcast softness to faithfully emulate the flattened regional news video texture", "Arranged the group dynamically at chest height, using the police tape to anchor all characters within the same foreground space" ], "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": "28e8ceb29c3853c8", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "6d15f23946bf09fd", "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": [ "C02", "L03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 3, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Presented as regional news footage, the camera holds at chest height near the front edge of the detective group, about 3 meters from Hyesu, in a medium-wide frame. Hyesu occupies the center with her arm extended toward the crime scene off-frame, while the detectives form a tense wall around her and police tape cuts a shallow line across the lower foreground.", "lighting_mood": "Daylight is flattened by news-video exposure, with muted grays, procedural blue shadows, and a restrained red accent from small evidence markers near the controlled area.", "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": "crime scene", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "points_to", "gesture_target_label": "crime scene", "label": "Hyesu", "screen_zone": "middle_center", "target_id": "C02", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02", "L03" ], "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": [], "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": [], "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": "327a8b6db9a8f34c" }, { "scene_index": 4, "heading": "S#4. 옥탑방 / D", "beat_title": "불안감이 맴도는 적막한 옥탑방 전경", "representative_moment": "A high overhead crane view captures the weathered rooftop and laundry lifting in the foggy coastal wind.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "High overhead crane view, wide framing, diffuse blue-gray daylight with a dull-gold fog haze", "t2i_prompt": "Photorealistic cinematic still. A high overhead crane view looking steeply downward, framing a weathered rooftop room and its faded roof. Beside it, a clothesline holds pale laundry slightly lifted by the sea wind above the damp rooftop floor. Concrete parapet edges border the space, as fog-softened coastal low-rise buildings fall away into the gray background. Diffuse coastal daylight turns the scene blue-gray and tactile, with a faint dull-gold haze trapped in the thick, low fog, leaving no clean highlights.", "outfit_assignments": [], "source_facts": [ "shot description: A high overhead crane view hangs steeply above the old villa rooftop, slightly off-center from the roofline", "shot description: laundry lifts in the sea wind", "shot description: fog-softened coastal buildings falling away into the gray background" ], "visual_inferences": [ "No characters are visible; focus is entirely on the atmospheric environment and architectural textures", "The steep angle emphasizes the isolation of the rooftop room" ], "creative_decisions": [ "Use wide framing and diffuse blue-gray lighting with a faint dull-gold haze to underscore the ominous stillness and isolation.", "Emphasize the motion of the laundry as the sole dynamic element in a suspended, quiet space." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "clothesline", "usage_kind": "anchor", "source_phrase": "a clothesline holds" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "laundry", "usage_kind": "redraw", "source_phrase": "pale laundry slightly lifted by the sea wind" }, { "owned_token": "metal door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "anchor", "source_phrase": "Concrete parapet edges" }, { "owned_token": "roof", "usage_kind": "anchor", "source_phrase": "faded roof" }, { "owned_token": "rooftop floor", "usage_kind": "anchor", "source_phrase": "damp rooftop floor" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "weathered rooftop room" }, { "owned_token": "stair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stairwell", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "threshold", "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": "eb6a02054fa73c28", "owned_hash": "8e0827c5a1585e6b", "camera_direction_hash": "b3632531d155cbcb", "owned_usage_hash": "c31ffdd0ea8c5880", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "clothesline", "violating_phrase": "a clothesline holds", "reason": "The prompt references the existing clothesline to establish the scene layout, acting as an anchor rather than a redraw instruction.", "verdict": "anchor_reference" }, { "owned_object": "laundry", "violating_phrase": "pale laundry slightly lifted by the sea wind", "reason": "The prompt describes the wind's effect on the existing laundry as a spatial/state anchor, not instructing to redraw or add new laundry, making it an anchor reference despite the producer's redraw classification.", "verdict": "anchor_reference" }, { "owned_object": "parapet", "violating_phrase": "Concrete parapet edges", "reason": "The prompt refers to the concrete parapet edges to define the spatial boundaries of the pre-existing shot.", "verdict": "anchor_reference" }, { "owned_object": "roof", "violating_phrase": "faded roof", "reason": "The prompt mentions the faded roof as a framing element of the pre-existing environment.", "verdict": "anchor_reference" }, { "owned_object": "rooftop floor", "violating_phrase": "damp rooftop floor", "reason": "The prompt mentions the damp rooftop floor to describe the atmosphere and layout of the pre-existing background.", "verdict": "anchor_reference" }, { "owned_object": "rooftop room", "violating_phrase": "weathered rooftop room", "reason": "The prompt frames the pre-existing weathered rooftop room as a composition anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "This is an independent establishing shot introducing the rooftop environment without visual dependency on prior scenes.", "_shot_index": 1, "visible_entities": [ "L04" ], "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": "A high overhead crane view hangs steeply above the old villa rooftop, slightly off-center from the roofline, close enough to read the rooftop room and the laundry line. The frame feels almost suspended as the laundry lifts in the sea wind, with fog-softened coastal buildings falling away into the gray background.", "lighting_mood": "Diffuse coastal daylight turns the city blue-gray and tactile, with a faint dull-gold haze trapped in the fog and no clean highlights.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "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": "L04B02", "reference_usage": "exact_background", "owned_objects": [ "clothesline", "deck", "drying platform", "laundry", "metal door", "parapet", "roof", "rooftop floor", "rooftop room", "stair", "stairwell", "threshold", "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": "L04B02", "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": "13bfcb6dd3b48ade" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "뉴스를 주시하는 어두운 표정의 민숙", "representative_moment": "앞치마를 두른 C03이 의자 끝에 앉아 굳은 표정으로 프레임 밖의 TV 화면을 응시하고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "tight close framing at seated eye level, cold TV flicker contrasting with soft warm background steam", "t2i_prompt": "Photorealistic cinematic still. C03O03 in a knit top and apron, an East Asian woman in her 40s, at three-quarter angle from the right, perched on a chair edge with her eyes fixed off-screen. Her apron-framed face with a heavy, dark expression fills the tight center-left of the composition. Cold, unseen TV flicker washes over her features, draining her face into heavy gray shadows, while thin daylight filters from the background. [: The cramped living space frame edge recedes softly behind her, featuring the curved side of a running sink faucet and the rim of a simmering pot in shallow depth of field, the rising steam faintly catching gold highlights at the edges.]", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O03" } ], "source_facts": [ "S#5 action: '앞치마를 두른 채 TV 뉴스를 뚫어지게 바라보는 민숙(수리영 엄마. 40세) 표정이 어둡다.'", "shot description: '앞치마를 멘 채 거실 TV 화면에 시선을 고정한 민숙의 무겁고 굳은 얼굴.'" ], "visual_inferences": [ "C03 seated pose: perched on chair edge, three-quarter right angle", "unseen TV screen implied by cold flicker reflecting on her face", "background features sink and steaming pot in soft focus" ], "creative_decisions": [ "close framing to emphasize Minsuk's dark expression and the cold TV lighting", "shallow depth of field to soften the background kitchenette details, adhering to the skipped_close_framing policy without relying on the absent L05 background reference", "used generic [L00: ...] block for the frame edge surfaces" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "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": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lock", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "plant", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pot", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shoe step", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stove", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "362fe8797bd12bb0", "owned_hash": "2d8d6b0a61824cd1", "camera_direction_hash": "161a483ac712261b", "owned_usage_hash": "353949352c1466cb", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "N/A", "_shot_index": 1, "visible_entities": [ "C03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Inside the rooftop room, the camera sits at seated eye level about 1.5 meters in front of Minsuk, slightly off the television axis so the screen itself is only implied by her fixed stare and reflected flicker. Her apron-framed face fills the tight composition center-left, with the sink and steaming pot receding softly behind her in the cramped living space.", "lighting_mood": "Thin window daylight mixes with cold TV flicker, draining Minsuk’s face into gray shadows while the kitchen steam glows faintly gold at the edges.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "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": 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": "L05B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "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": "1b71daa7ed26d873" }, { "scene_index": 5, "heading": "S#5. 옥탑방 안 / 실내", "beat_title": "딸의 시선에 머문 선명한 표식", "representative_moment": "민숙의 창백한 손목 안쪽에 새겨진 붉은 원형 표식이 클로즈업으로 또렷하게 보인다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "macro dolly-in at wrist height, extremely shallow depth of field", "t2i_prompt": "Photorealistic cinematic still. A macro insert shot at wrist height. The camera focuses entirely on the extended pale inner wrist of C03 in a knit and apron, an Asian middle-aged woman. The inner side of the wrist faces the camera directly in the lower-center frame, with her body extending off-screen and faces completely excluded. P02, a dark ring-like mark, sits sharply in focus on the colorless skin, illuminated by a narrow strip of warm indoor light that makes the mark read as a dense red-black omen. Behind the wrist, the cramped apartment entryway, featuring a blurred door and shoe step, dissolves into a soft, warm out-of-focus background.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O03" } ], "source_facts": [ "S#5 action: '민숙 손목에 보이는 작은 원(이하 표식)을 본다. 오늘따라 유난히 또렷해 보이지만...'", "shot description: Minsuk's pale inner wrist skin with circular mark", "DP instruction: inner wrist faces camera directly, showing the dark ring-like mark clearly" ], "visual_inferences": [ "macro focus isolates the wrist entirely, excluding the face to emphasize the mark (insert framing scale)", "apartment entryway in the background is deeply blurred to maintain focus on the wrist while providing spatial context" ], "creative_decisions": [ "warm narrow strip of light contrasting with pale skin to make the mark pop visually as a dense red-black omen", "shallow depth of field blurring the door and shoe step to give a cramped but soft domestic feel without distracting from the primary subject" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "prop", "background" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "counter", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "a blurred door and shoe step" }, { "owned_token": "faucet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "lock", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "plant", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pot", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shoe step", "usage_kind": "anchor", "source_phrase": "a blurred door and shoe step" }, { "owned_token": "sink", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stove", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "62c00a9c323fde2f", "owned_hash": "2d8d6b0a61824cd1", "camera_direction_hash": "70fccc3bbc07e53c", "owned_usage_hash": "4046c5d9188492f5", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "door", "violating_phrase": "a blurred door and shoe step", "reason": "The prompt mentions 'door' as a blurred background element to establish context, which matches the producer's echo of 'anchor' with no instruction to redraw.", "verdict": "anchor_reference" }, { "owned_object": "shoe step", "violating_phrase": "a blurred door and shoe step", "reason": "The prompt mentions 'shoe step' as part of the out-of-focus background entryway, matching the producer's echo of 'anchor' and avoiding any redraw instructions.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "같은 옥탑방 배경을 유지하기 위해 앞선 샷의 정확한 배경 정보(exact_background) 참조", "_shot_index": 8, "visible_entities": [ "C03", "P02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 5, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At wrist height beside Minsuk, the camera completes a macro dolly-in from only 20 to 30 centimeters away, excluding both faces. The inner wrist lies across the lower-center of the insert, the dark circular mark held in sharp focus while the apartment entryway dissolves into a soft, cramped blur behind it.", "lighting_mood": "A narrow strip of warm indoor light grazes the skin, making the mark read as a dense red-black omen against otherwise colorless flesh.", "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", "P02" ], "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": [ { "subject_id": "C03", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The image isolates only Minsuk’s wrist and mark with both faces excluded, so a base character reference is enough and outlook-specific facial reference should be avoided." } ], "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": "L05B01", "reference_usage": "exact_background", "owned_objects": [ "TV", "chair", "counter", "dining table", "door", "faucet", "lock", "plant", "pot", "shoe step", "sink", "stove", "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": 5, "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", "id": "C03", "policy": "required" }, { "kind": "background", "id": "L05B01", "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": 5, "shot_indices": [ 8 ], "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": "74b789736813387f" }, { "scene_index": 6, "heading": "S#6. 해안가 도로 / D", "beat_title": "해안가 도로를 달리는 수리영", "representative_moment": "수리영이 바람을 맞으며 자전거 페달을 힘차게 밟고 해안가 도로를 달린다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "side-tracking wide profile shot, bright coastal daylight softened by mist", "t2i_prompt": "Photorealistic cinematic still. C01O01, a Korean young woman in her early 20s in a gray casual windbreaker, appears as Suriyeong on a bicycle in the middle-center midground moving toward screen right along the coastal road, while the coastal road direction stretches across the middle-right background. Seen in right profile, seated in mid-pedal freeze with one foot driving the down stroke, she is wearing earphones and gazing into the distance. Her hands grip the handlebar, the bicycle frame and wheels clearly visible below her. Her dark long straight hair and jacket collar are blown backward by the sea wind in the direction opposite to her motion. A weathered guardrail forms a horizontal band behind her, separating the rider from the open pale sea. Bright coastal daylight is softened by mist, featuring muted blue-gray sea tones and a clean gold edge catching her windblown hair in the wide frame.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#6 action: Suriyeong rides her bicycle along the coastal road, wearing earphones.", "shot description: energetic side profile of Suriyeong pedaling vigorously, hair and clothes blowing backward in the wind." ], "visual_inferences": [ "bicycle frame and wheels must be fully visible to support the riding posture", "guardrail acts as a midground horizontal anchor separating the road from the sea", "mid-pedal freeze to represent active motion without blurring" ], "creative_decisions": [ "side-tracking wide composition to emphasize speed, direction, and the coastal environment", "mist-softened daylight with a gold edge light on the hair to highlight the bright mood against muted sea tones" ], "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": "a9ee3ef0f90e5979", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "113143f9f9bd5ca3", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "none", "_shot_index": 1, "visible_entities": [ "C01", "L06" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 6, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "A side-tracking camera moves parallel to Suriyeong about 6 feet from her bicycle at chest height, holding a clean perpendicular profile as she rides along the coastal road. She and the bicycle run through the middle of the wide frame, hair and collar blown backward by sea wind while the road and guardrail streak horizontally behind her.", "lighting_mood": "Bright coastal daylight is softened by mist, with muted blue-gray sea tones and a clean gold edge on her windblown hair.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "coastal road direction", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "screen right along the coastal road", "label": "Suriyeong on bicycle", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L06" ], "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": "9a4e992823b895b7" }, { "scene_index": 7, "heading": "S#7. 유기견 보호원 / D", "beat_title": "맹견을 압도하는 매서운 눈빛", "representative_moment": "수리영(C01)이 미동조차 없이 매서운 눈빛으로 프레임 정중앙에서 카메라(맹견의 시선)를 똑바로 응시한다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Extreme close-up from dog-eye height, cold fluorescent lighting with deep gray damp shadows", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, an East Asian young woman in her early 20s with long straight black hair, a round face contour, and distinct dark eyes, positions her eyes in the middle-center foreground, looking toward the Rottweiler just outside the frame. She faces the camera directly, chin locked forward, the tight extreme close-up framing isolating her intense, motionless stare. Captured from a low observer position at dog-eye height, the cold fluorescent shelter light flattens the skin across her cheeks, flanked by deep gray, damp shadows under the brow. Tiny wet highlights catch the center of her unblinking pupils. [: Soft vertical dark strips of worn metallic kennel cages blur at the extreme left and right frame edges, compressing the dim, heavy utilitarian atmosphere of the aisle into indistinct periphery shadows.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Suriyeong staring down the fierce eyes of the Rottweiler.", "Cold fluorescent shelter light flattens the skin while deep gray shadows and tiny wet eye highlights make her stare feel unnervingly still.", "The push-in resolves into an extreme close-up on Suriyeong’s eyes, holding almost motionless at the end of the confrontation axis." ], "visual_inferences": [ "fixed_element shelter_kennel_environment: Cold fluorescent light flattens the aisle while damp shadows under the cages add heavy utilitarian atmosphere.", "The dog-eye height observer POV frames her face looking slightly downward but piercing directly forward toward the lens.", "Kennel bars act as out-of-focus vertical frames on the left and right edges, isolating the subject's face." ], "creative_decisions": [ "Extreme close-up focusing purely on the eyes, brow, and upper cheeks to emphasize her unyielding dominance over the off-screen dog.", "Using shallow depth of field to blur the kennel bars at the frame edges, adhering to the skipped_close_framing background binding." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "31b2ebd3ff228491", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "f6e8195fa4402031", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "N/A", "_shot_index": 2, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 7, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From dog-eye height in the kennel aisle, the push-in resolves into an extreme close-up on Suriyeong’s eyes, holding almost motionless at the end of the confrontation axis. Her eyes sit dead center in the narrow frame, facing the lens as if it were the Rottweiler’s stare, while the kennel bars fall into dark vertical blur at the edges.", "lighting_mood": "Cold fluorescent shelter light flattens the skin while deep gray shadows and tiny wet eye highlights make her stare feel unnervingly still.", "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": "looks_toward", "gesture_target_label": "Rottweiler just outside frame", "label": "Suriyeong's eyes", "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": 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_kennel_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of the animal shelter features cold, flat gray fluorescent lighting that casts deep, damp shadows. Rows of metal kennel cages and bars define the structural background of the aisle, giving the space a heavy and utilitarian atmosphere.", "applies_to_shots": [ 2, 5 ], "element_scope": "full", "source_facts": [ "Cold fluorescent shelter light flattens the skin while deep gray shadows...", "kennel bars fall into dark vertical blur at the edges", "Cold gray fluorescent light flattens the aisle while damp shadows under the cages...", "cage line receding along the left edge" ], "visual_inferences": [ "The metal cages run continuously along the aisle, creating consistent vertical and receding lines.", "The shelter environment feels slightly worn and inherently damp." ], "creative_decisions": [ "Maintaining a consistent cold gray color palette and flat lighting across the shots to emphasize the dreary, exhausted mood of the shelter aisle." ], "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": "00a6f69a05fa8b24" }, { "scene_index": 7, "heading": "S#7. 유기견 보호원 / D", "beat_title": "답답함과 한숨", "representative_moment": "수리영(C01)이 보호소 복도에 앉아 휴대전화를 귀에 댄 채 굳은 표정으로 바닥을 응시하고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium upper-body frame at chest height, cold gray flat fluorescent lighting", "t2i_prompt": "Photorealistic cinematic still. At chest height in the aisle, a medium upper-body frame. C01O01 in a grey windbreaker, a Korean woman in her early 20s with long straight black hair, sitting at a three-quarter angle from the left slightly right of center, a smartphone pressed to her ear, looking down with lips firmly closed. The interior of the animal shelter features cold, flat gray fluorescent lighting from dull ceiling strips that casts deep, damp shadows. Rows of metal kennel cages with closed, crowded bars face partially toward the lens and recede into a dark vertical blur along the left edge. A strip of scuffed, wet gray concrete floor occupies the lower third, forming an empty negative space beneath her downward gaze.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 갇혀있는 개들 보면서 한숨 후~ 늦을 거 같다. 민숙에게 연락을 취해보지만, 받지 않는다.", "귀에 휴대전화를 댄 채 입술을 굳게 다물고 바닥을 응시하는 수리영의 상체." ], "visual_inferences": [ "fixed_element shelter_kennel_environment: cold, flat gray fluorescent lighting, damp shadows, rows of metal kennel cages", "scuffed, wet gray concrete floor occupying the lower frame as negative space" ], "creative_decisions": [ "At chest height, medium upper-body frame placing C01 slightly right of center to emphasize the receding cage line on the left", "Empty negative space on the floor to amplify the mood of exhaustion and waiting", "Cold flat lighting to flatten the skin and emphasize the damp, heavy atmosphere of the shelter" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "70392c6b382778f8", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "a9dcae94654b900b", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 7, "dependency_reason": "같은 보호소 내부 공간의 조명과 철창 연속성 유지", "_shot_index": 5, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 7, "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 chest height in the shelter aisle, the camera inherits the dolly-out path but settles into a held upper-body frame several feet from 수리영. She sits slightly right of center with the cage line receding along the left edge and a strip of stained floor in the lower third, her phone pressed to her ear as the negative space below her gaze feels heavier than the dogs behind her.", "lighting_mood": "Cold gray fluorescent light flattens the aisle while damp shadows under the cages add a coastal, exhausted heaviness.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "shelter_kennel_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of the animal shelter features cold, flat gray fluorescent lighting that casts deep, damp shadows. Rows of metal kennel cages and bars define the structural background of the aisle, giving the space a heavy and utilitarian atmosphere.", "applies_to_shots": [ 2, 5 ], "element_scope": "full", "source_facts": [ "Cold fluorescent shelter light flattens the skin while deep gray shadows...", "kennel bars fall into dark vertical blur at the edges", "Cold gray fluorescent light flattens the aisle while damp shadows under the cages...", "cage line receding along the left edge" ], "visual_inferences": [ "The metal cages run continuously along the aisle, creating consistent vertical and receding lines.", "The shelter environment feels slightly worn and inherently damp." ], "creative_decisions": [ "Maintaining a consistent cold gray color palette and flat lighting across the shots to emphasize the dreary, exhausted mood of the shelter aisle." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 7, "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": "6f55ad17832066db" }, { "scene_index": 8, "heading": "S#8. 번화가·대형 마트 / 해질 무렵", "beat_title": "마트 직원으로부터 엄마의 행방을 전해 듣다", "representative_moment": "A male employee holding a large paper box turns his head to look at Suri-young.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium framing from eye level mid-aisle, pale fluorescent wash with faint gold dusk and red neon bleed", "t2i_prompt": "Photorealistic cinematic still. Medium framing from eye level mid-aisle. A Korean adult man in a grey outdoor jacket stands at three-quarter angle from the right, his torso positioned center-left. He is bracing a large, worn cardboard box with both hands, the box prominent in the lower foreground and tilted upward toward his chest. His eyes are fixed on a young Korean woman in a casual daily outfit, captured over the shoulder in the right foreground, paused beside the shelf with her back to the camera, looking off-screen. The two figures share the same continuous supermarket aisle, framed by fully stocked shelves facing inward. Overhead fluorescents wash the scene pale, while faint gold dusk and red neon bleed from the outside neon sign and distant window catch the aisle edges.", "outfit_assignments": [], "source_facts": [ "Suri-young looks for her mother inside the supermarket.", "A male employee holding a cardboard box speaks to Suri-young.", "The setting is a supermarket interior at dusk with neon lights outside." ], "visual_inferences": [ "fixed_element suri_young_base_appearance: A young Korean woman in a casual daily outfit.", "fixed_element dusk_neon_lighting: Faint gold dusk and red neon bleed visible in the ambient light.", "The male employee holds the box at chest height, creating a physical barrier during the conversation." ], "creative_decisions": [ "Used a medium framing from eye level to capture both the male employee's reaction and Suri-young's over-the-shoulder presence in a single spatial anchor.", "Positioned the cardboard box prominently in the lower foreground to emphasize the interruption of mundane labor.", "Blended indoor pale fluorescent lighting with outdoor dusk and red neon bleed to match the mood of the commercial district." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "CCTV camera", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "automatic door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "awning", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bollard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "crate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "entrance", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hand truck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "neon sign", "usage_kind": "anchor", "source_phrase": "outside neon sign" }, { "owned_token": "road", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "shopping cart corral", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sidewalk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storefront", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "streetlight", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "taxi", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "vendor cart", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "distant window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "eb44601780ad4fad", "owned_hash": "34aec3f5d0e24a43", "camera_direction_hash": "e1c83963444f280d", "owned_usage_hash": "0c6067dc1d4dc84c", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "neon sign", "violating_phrase": "outside neon sign", "reason": "The prompt references red neon bleeding from the outside neon sign, acting as a lighting anchor without requesting to redraw the physical sign.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "distant window", "reason": "The prompt mentions a distant window only as a source of ambient light bleeding into the aisle, serving as a lighting anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct scene dependencies provided; internal shot continuity is maintained via fixed elements.", "_shot_index": 4, "visible_entities": [ "C07" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 8, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From eye level mid-aisle, the camera continues the pan and lands about 2.5 meters from 남자 직원, with the paper box prominent in the lower foreground but kept under half the frame. His torso sits center-left as he turns his head toward 수리영's side of the aisle, while stocked shelves on both sides keep their positions connected in one continuous supermarket corridor.", "lighting_mood": "Overhead supermarket fluorescents wash the scene pale, with faint gold dusk and red neon bleed catching the aisle edges.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C07" ], "allowed_outlook_pairs": [ { "character_id": "C07", "outlook_id": "O07" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": 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": "C07", "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": "L08B01", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "automatic door", "awning", "bollard", "crate", "curb", "entrance", "hand truck", "neon sign", "road", "shopping cart corral", "sidewalk", "sign", "storefront", "streetlight", "taxi", "vendor cart", "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": "suri_young_base_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman. Her overall physical appearance, including her facial features, hairstyle, and casual daily outfit, remains visually consistent as she transitions from the supermarket interior to the outdoor sidewalk.", "applies_to_shots": [ 4, 5 ], "element_scope": "full", "source_facts": [ "Suri-young appears consecutively in the supermarket interior (Shot 4) and on the street outside (Shot 5)." ], "visual_inferences": [ "Her clothing and general hairstyle remain unchanged between leaving the mart and standing outside." ], "creative_decisions": [ "Maintain identical character design and wardrobe to ensure clear visual continuity despite the change in location and posture." ], "confidence": "high" }, { "element_id": "dusk_neon_lighting", "element_type": "environment_state", "character_name": "", "description": "The environmental lighting establishes a dusk setting with noticeable red neon lights. This evening atmosphere and red neon bleed remain consistent, visibly affecting both the edges of the supermarket interior and the dark commercial street outside.", "applies_to_shots": [ 4, 5 ], "element_scope": "full", "source_facts": [ "The scene text explicitly mentions dusk ('해질 무렵') and neon signs ('네온사인') in the commercial district." ], "visual_inferences": [ "The red neon glow from the busy street permeates the surrounding areas, bridging the bright indoor lighting and the dark outdoor shadows." ], "creative_decisions": [ "Use the dusk ambiance and red neon light bleed as a unifying visual motif across both shots to emphasize the continuous evening setting." ], "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": "L08B01", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "551ca809075795da" }, { "scene_index": 8, "heading": "S#8. 번화가·대형 마트 / 해질 무렵", "beat_title": "엇갈린 만남과 초조한 기다림", "representative_moment": "어스름한 길거리에 선 수리영(C01)이 받지 않는 P01 화면을 내려다본다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level wide shot across the dark street with muted blue-gray dusk and sharp red neon bleed.", "t2i_prompt": "Photorealistic cinematic still. Eye-level wide framing from across the street, the shoes of C01O01 occupying the lower frame and her head in the upper frame. C01O01 in a gray windbreaker, a young Korean woman in her 20s with long straight black hair and a round face, stands at a three-quarter angle from the right on the dark sidewalk. Her posture is hunched over P01 facing her, her gaze looking down at the small cold rectangular glow of the phone illuminating her hands. The background preserves the exact muted blue-gray dusk environment from the prior shot. Behind her, the bright storefront and the glass door sit closed. Overhead, a red neon sign faces the camera, showing red and white lettering. The wet road and the curb beneath her feet reflect the fragmented luminous blocks of the red signage. A bicycle rack and a vendor stall stand faintly in the deep background shadow.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Suri-young steps out of the mart, holding her phone while trying to contact her mother.", "The background is a commercial street at dusk with neon signs and wet ground." ], "visual_inferences": [ "fixed_element suri_young_base_appearance: her gray windbreaker and long straight hair remain consistent with the prior interior shot.", "fixed_element dusk_neon_lighting: the muted dusk lighting and red neon bleeds extend consistently into this exterior shot." ], "creative_decisions": [ "Positioned the camera at eye-level to strictly capture the full body within the wide frame, following geometric consistency rules.", "Contrasted the cold blue-gray dusk shadows of the empty street against the sharp red neon reflections to emphasize her isolation and anxiety." ], "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": "bicycle glyph", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bicycle rack", "usage_kind": "anchor", "source_phrase": "A bicycle rack and a vendor stall stand faintly" }, { "owned_token": "curb", "usage_kind": "anchor", "source_phrase": "The wet road and the curb beneath her feet" }, { "owned_token": "display window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "electrical box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "glass door", "usage_kind": "anchor", "source_phrase": "the bright storefront and the glass door sit closed" }, { "owned_token": "lane marking", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pipe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "road", "usage_kind": "anchor", "source_phrase": "The wet road and the curb beneath her feet" }, { "owned_token": "sidewalk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sign", "usage_kind": "anchor", "source_phrase": "a red neon sign faces the camera" }, { "owned_token": "storefront", "usage_kind": "anchor", "source_phrase": "the bright storefront and the glass door sit closed" }, { "owned_token": "vendor stall", "usage_kind": "anchor", "source_phrase": "A bicycle rack and a vendor stall stand faintly" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wheel stand", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "7768f882ec5d464d", "owned_hash": "03387a3d0ef18fe5", "camera_direction_hash": "a6bf58bc97800581", "owned_usage_hash": "c8b380b60a571155", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "sidewalk", "violating_phrase": "on the dark sidewalk", "reason": "The prompt mentions the dark sidewalk as the surface the character stands on, which is an anchor reference, though the producer declared it absent.", "verdict": "anchor_reference" }, { "owned_object": "storefront", "violating_phrase": "the bright storefront and the glass door sit closed", "reason": "The storefront is referenced as sitting closed in the background, serving as an environmental anchor, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "glass door", "violating_phrase": "the bright storefront and the glass door sit closed", "reason": "The glass door is referenced as sitting closed in the background, serving as an environmental anchor, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "sign", "violating_phrase": "a red neon sign faces the camera, showing red and white lettering", "reason": "The sign is mentioned overhead to establish lighting and reflections, serving as an environmental anchor, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "road", "violating_phrase": "The wet road and the curb beneath her feet", "reason": "The wet road is referenced for reflection effects beneath her feet, acting as an anchor, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "curb", "violating_phrase": "The wet road and the curb beneath her feet", "reason": "The curb is referenced to describe reflections near her feet, acting as a spatial anchor, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "bicycle rack", "violating_phrase": "A bicycle rack and a vendor stall stand faintly", "reason": "The bicycle rack is mentioned as standing faintly in the background shadow, acting as a background anchor, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "vendor stall", "violating_phrase": "A bicycle rack and a vendor stall stand faintly", "reason": "The vendor stall is mentioned as standing faintly in the background shadow, acting as a background anchor, matching the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 8, "dependency_reason": "Shot 4의 배경 공간과 환경 조명 상태(dusk_neon_lighting)를 동일하게 유지하기 위해 이전 샷의 설정을 시각적으로 연속 반영함.", "_shot_index": 5, "visible_entities": [ "C01", "L08", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 8, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Static from across the street at waist height, the camera holds a wider full-body frame with 수리영 isolated near the lower center-left of a dark sidewalk. Her phone glow is a small cold rectangle at her hands, while the supermarket frontage and empty pavement take up more space than she does, turning the busy district into a hollow coastal night.", "lighting_mood": "Muted blue-gray dusk has fallen into deep sidewalk shadow, cut by red neon accents and the small icy glow of the phone.", "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", "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", "bicycle glyph", "bicycle rack", "curb", "display window", "electrical box", "glass door", "lane marking", "pipe", "rail", "road", "sidewalk", "sign", "storefront", "vendor stall", "wall", "wheel stand" ], "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_base_appearance", "element_type": "character_state", "character_name": "수리영", "description": "A young Korean woman. Her overall physical appearance, including her facial features, hairstyle, and casual daily outfit, remains visually consistent as she transitions from the supermarket interior to the outdoor sidewalk.", "applies_to_shots": [ 4, 5 ], "element_scope": "full", "source_facts": [ "Suri-young appears consecutively in the supermarket interior (Shot 4) and on the street outside (Shot 5)." ], "visual_inferences": [ "Her clothing and general hairstyle remain unchanged between leaving the mart and standing outside." ], "creative_decisions": [ "Maintain identical character design and wardrobe to ensure clear visual continuity despite the change in location and posture." ], "confidence": "high" }, { "element_id": "dusk_neon_lighting", "element_type": "environment_state", "character_name": "", "description": "The environmental lighting establishes a dusk setting with noticeable red neon lights. This evening atmosphere and red neon bleed remain consistent, visibly affecting both the edges of the supermarket interior and the dark commercial street outside.", "applies_to_shots": [ 4, 5 ], "element_scope": "full", "source_facts": [ "The scene text explicitly mentions dusk ('해질 무렵') and neon signs ('네온사인') in the commercial district." ], "visual_inferences": [ "The red neon glow from the busy street permeates the surrounding areas, bridging the bright indoor lighting and the dark outdoor shadows." ], "creative_decisions": [ "Use the dusk ambiance and red neon light bleed as a unifying visual motif across both shots to emphasize the continuous evening setting." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 8, "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": "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": "e1dec587eeda2c18" }, { "scene_index": 9, "heading": "S#9. 해안가 / N", "beat_title": "다급한 귀가", "representative_moment": "어둠이 깔린 해안 도로 위, 수리영이 안개 속을 뚫고 거친 숨을 내쉬며 자전거 페달을 밟고 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium side profile at waist-to-chest rider height, harsh headlight cutting through blue-black coastal fog", "t2i_prompt": "Photorealistic cinematic still. C01O01, a Korean woman in her 20s in a gray windbreaker, in the middle-center midground, moving toward screen right along the coastal road. Captured in left profile at waist-to-chest rider height, she sits on the saddle of the bicycle in mid-pedal, her upper torso pitched forward, hands gripping the handlebar, gazing into the distance with anxious dark eyes. Her long black hair streams backward into the dark air in the direction opposite to motion. The upper bicycle frame and front wheel are visible below her waist. The desolate environment is bathed in a weak blue-black coastal ambient night, sharply sliced by a harsh, narrow bicycle headlight beam projecting forward, casting a tight corridor of light over the damp asphalt and thick drifting sea fog.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#9 action: '해안도로를 자전거를 타고 바삐 집으로 달려가는 수리영.'", "shot description: observer at roadside, medium side profile of rider, upper torso pitched forward, hair streaming backward" ], "visual_inferences": [ "bicycle headlight provides the only practical light in the dark coastal ambient", "mid-pedal freeze frame preserves the forward motion direction while capturing the anxious rush", "visible bicycle frame anchors the rider to avoid floating appearance" ], "creative_decisions": [ "Chest-height camera angle to emphasize the upper body leaning forward in desperate haste", "Thick sea fog illuminated by the sharp headlight beam to create a lonely, ominous mood matching the deep blue-black coastal night" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "56bff905b6d4b6ec", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "94fbeb3f806d8d3d", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 6, "dependency_reason": "해안가 도로의 배경적 일관성 유지 (어둡고 안개 낀 밤바다 주변 환경)", "_shot_index": 1, "visible_entities": [ "C01", "L06" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 9, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Fixed at the roadside perpendicular to the bicycle path, the camera holds a rigid medium side profile at waist-to-chest rider height as 수리영 passes through the bicycle headlight spill. Her body cuts across the middle of frame, upper torso pitched forward over the handlebars, one pedal driven downward while her hair streams backward into the dark coastal air.", "lighting_mood": "Nearly all illumination comes from the bicycle headlight and a weak blue-black coastal ambient, with fog turning the light into a hard, anxious stripe.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "screen right along the coastal road", "label": "수리영 on bicycle", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L06" ], "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": "8cd350bba9bb3b94" }, { "scene_index": 10, "heading": "S#10. 옥탑방 / N", "beat_title": "유리창에 드리운 끔찍한 실루엣", "representative_moment": "불 꺼진 옥탑방의 반투명한 유리창 너머로 거대하고 짙은 C10의 실루엣이 불길하게 드리워진 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level voyeuristic medium framing outside, low-key lighting with cold gray spill against deep shadow.", "t2i_prompt": "Photorealistic cinematic still. An eye-level voyeuristic camera outside the rooftop room, staying close to the weathered exterior wall visible at the frame edge, aligned squarely with the frosted window. The translucent pane occupies most of the frame. Behind the glass, C10O10 in dark coat, an Asian man, looms looking away off-screen. C10O10 is consistently depicted as a heavily obscured dark mass or silhouette, showing the variant features carried from stable_traits — a black humanoid outline with faint edges and a dark mass forming a shadow-like figure. His clothing details and physical identity remain completely hidden in deep shadow, functioning purely as a menacing, formless presence with no facial features visible. The environment is enveloped in deep, swallowing blacks with extremely low ambient lighting, the window lit only by faint exterior night spill turning the frosted glass into a cold gray membrane.", "outfit_assignments": [ { "character_id": "C10", "outlook_id": "O10" } ], "source_facts": [ "창문에 비친 그림자", "얼굴은 보이지 않는다", "검은 실루엣", "불이 꺼진 옥탑방" ], "visual_inferences": [ "fixed_element unidentified_attacker_silhouette: C10O10 is consistently depicted as a heavily obscured dark mass or silhouette.", "fixed_element deep_interior_shadow: The environment is enveloped in deep, swallowing blacks with extremely low ambient lighting.", "The attacker's presence is purely shape-based against the frosted glass, emphasizing a voyeuristic angle from the outside." ], "creative_decisions": [ "Positioning the camera strictly outside to compress the exterior wall against the frame, heightening the tension.", "Keeping the subject completely abstract behind the frosted membrane, using faint cold gray rim light to contrast against the unlit interior." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "usage_kind": "anchor", "source_phrase": "outside the rooftop room" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "weathered exterior wall visible at the frame edge" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "frosted window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "7135d31f62aa8a6a", "owned_hash": "a46f894a96618611", "camera_direction_hash": "e3d7dae758820c14", "owned_usage_hash": "731802a286771d4f", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "rooftop room", "violating_phrase": "outside the rooftop room", "reason": "The rooftop room is referenced solely as a spatial locator for the camera, which is consistent with the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "weathered exterior wall visible at the frame edge", "reason": "The wall is used purely as a framing anchor at the edge of the frame, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "frosted window", "reason": "The window acts as a compositional framing element and a translucent membrane to obscure the silhouette, confirming its role as an anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 4, "dependency_reason": "동일한 옥탑방 공간의 연속성 및 정확한 배경 요소 유지(S4_Shot1 참조)", "_shot_index": 3, "visible_entities": [ "C10", "L04" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 10, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Outside the rooftop room, the camera stays eye level and close to the exterior wall, aligned squarely with the frosted window glass as the tracking move settles. The translucent pane occupies most of the frame as an architectural surface, with 한국인 남자 reduced to a broad, heavy central silhouette whose face never resolves through the dark glass.", "lighting_mood": "The window is lit only by faint exterior night spill, turning the frosted glass into a cold gray membrane against deep black interior shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C10", "L04" ], "allowed_outlook_pairs": [ { "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": "background_ref_attached", "bg_id": "L04B03", "reference_usage": "exact_background", "owned_objects": [ "door", "floor", "rooftop room", "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": "unidentified_attacker_silhouette", "element_type": "character_state", "character_name": "침입한 남자", "description": "The male attacker is consistently depicted as a heavily obscured dark mass or silhouette. His facial features, clothing details, and physical identity remain completely hidden in deep shadow, functioning purely as a menacing, formless presence.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "얼굴은 보이지 않는다", "창문에 비친 그림자", "검은 실루엣", "어두운 형상" ], "visual_inferences": [ "The figure's lack of detail is maintained across different camera angles to preserve anonymity and emphasize the terrifying nature of the attack." ], "creative_decisions": [ "The character is rendered entirely as a shadowy mass, avoiding any specific demographic or textural rendering to keep him abstract." ], "confidence": "high" }, { "element_id": "deep_interior_shadow", "element_type": "environment_state", "character_name": "", "description": "The environment is enveloped in deep, swallowing blacks with extremely low ambient lighting. The only illumination is a faint, cold gray rim or spill from the exterior window, creating a high-contrast, ominous atmosphere where details are lost to darkness.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "불이 꺼진 옥탑방", "deep black interior shadow", "faint exterior night spill", "swallowed interior blacks" ], "visual_inferences": [ "The primary light source is weak moonlight or ambient city light from outside, barely penetrating the interior." ], "creative_decisions": [ "Keeping the interior lighting minimal and swallowed by shadow to heighten the horror and focus entirely on the dark silhouettes." ], "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": "character_outlook", "id": "C10O10", "policy": "required" }, { "kind": "background", "id": "L04B03", "policy": "required" } ], "forbidden_refs": [], "readiness_policy": "block_if_missing", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "623eb5de3511ad0d" }, { "scene_index": 10, "heading": "S#10. 옥탑방 / N", "beat_title": "어둠 속의 습격", "representative_moment": "어둠 속에서 C10O10이 C03O03의 어깨를 강하게 덮쳐오고, C03O03가 고통으로 일그러진 표정을 짓는 끔찍한 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "tight handheld close-up at shoulder height, high-contrast deep interior shadow with cold gray rim light", "t2i_prompt": "Photorealistic cinematic still. A tight, unstable handheld close frame at shoulder height, placing the torso and head in frame while the lower body is implied beyond the frame edges. C03O03 in a knit sweater, an Asian woman in her 40s with black medium-length hair and soft facial contours, occupies the right half of the frame at a three-quarter angle from the left. She is recoiling under a firm grip, looking up with a visibly wounded, pained expression, her face contorted in agony. Pressing in from the left is C10O10 in a dark ragged coat, a dark male figure. C10O10 is consistently depicted as a heavily obscured dark mass or silhouette. His facial features, clothing details, and physical identity remain completely hidden in deep shadow, functioning purely as a menacing, formless presence. C10O10 is in right profile, looking down and hunched into C03O03's shoulder, exerting intense downward force. The physical action is implied through shadow and their tight proximity rather than graphic detail, both figures sharing the cramped space anchored against [: a dark, stained rooftop-room wall pressing diagonally behind them]. The environment is enveloped in deep, swallowing blacks with extremely low ambient lighting. The only illumination is a faint, cold gray rim or spill from the out-of-focus exterior window, creating a high-contrast, ominous atmosphere where details are lost to darkness, with a muted red symbolic glint faintly reflecting near the shoulder contact point.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O03" }, { "character_id": "C10", "outlook_id": "O10" } ], "source_facts": [ "민숙을 덮쳐 어깨를 물어뜯는 듯한 모습", "고통으로 잔뜩 일그러진 민숙의 얼굴", "얼굴은 보이지 않는다", "검은 실루엣" ], "visual_inferences": [ "C10O10 is a dark formless presence with no visible facial features, functioning as a shadowy mass", "C03O03 occupies the right half recoiling from the sudden attack", "deep interior shadow swallowed by blacks with a faint exterior rim light", "the attack is tightly framed to emphasize claustrophobia" ], "creative_decisions": [ "tight unstable handheld framing at shoulder height to anchor the claustrophobic horror and physical tension", "muted red symbolic glints reflecting on the shoulder contact point instead of graphic gore", "using [L00: rooftop-room wall] directly as a frame-edge anchor since the main background reference is skipped via close framing" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "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": "a1c3129f61f25616", "owned_hash": "a46f894a96618611", "camera_direction_hash": "0a153cb93088c6a6", "owned_usage_hash": "c20e86732e6cc314", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 4, "dependency_reason": "utilizing the established environmental lighting (deep interior shadow) and the character state of the dark silhouette from prior references", "_shot_index": 4, "visible_entities": [ "C03", "C10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 10, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "A hard cut places the camera inside the dark room at shoulder height, handheld and very close beside 민숙's face and shoulder. The frame is tight and unstable: 민숙's contorted face occupies the right half, while 검은 실루엣 presses in from the left as a dark mass at her shoulder, the action implied through shadow and proximity rather than graphic detail.", "lighting_mood": "A thin gray rim from the window and swallowed interior blacks create a tactile, ominous contrast, with any red kept as a muted symbolic glint rather than spectacle.", "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", "C10" ], "allowed_outlook_pairs": [ { "character_id": "C03", "outlook_id": "O03" }, { "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": "L04B03", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "unidentified_attacker_silhouette", "element_type": "character_state", "character_name": "침입한 남자", "description": "The male attacker is consistently depicted as a heavily obscured dark mass or silhouette. His facial features, clothing details, and physical identity remain completely hidden in deep shadow, functioning purely as a menacing, formless presence.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "얼굴은 보이지 않는다", "창문에 비친 그림자", "검은 실루엣", "어두운 형상" ], "visual_inferences": [ "The figure's lack of detail is maintained across different camera angles to preserve anonymity and emphasize the terrifying nature of the attack." ], "creative_decisions": [ "The character is rendered entirely as a shadowy mass, avoiding any specific demographic or textural rendering to keep him abstract." ], "confidence": "high" }, { "element_id": "deep_interior_shadow", "element_type": "environment_state", "character_name": "", "description": "The environment is enveloped in deep, swallowing blacks with extremely low ambient lighting. The only illumination is a faint, cold gray rim or spill from the exterior window, creating a high-contrast, ominous atmosphere where details are lost to darkness.", "applies_to_shots": [ 3, 4 ], "element_scope": "full", "source_facts": [ "불이 꺼진 옥탑방", "deep black interior shadow", "faint exterior night spill", "swallowed interior blacks" ], "visual_inferences": [ "The primary light source is weak moonlight or ambient city light from outside, barely penetrating the interior." ], "creative_decisions": [ "Keeping the interior lighting minimal and swallowed by shadow to heighten the horror and focus entirely on the dark silhouettes." ], "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": "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": "86bf5efb6185853b" }, { "scene_index": 11, "heading": "S#11. 옥탑 방 밖 / N", "beat_title": "역겨운 냄새와 마주한 멈칫", "representative_moment": "계단을 오르다 훅 끼치는 악취에 한 손으로 코를 틀어막고 찌푸리는 수리영의 멈칫한 상체.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "handheld chest-height observer perspective, cold blue-gray spill lighting", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, a Korean woman in her early 20s, viewed at a three-quarter angle from the right, her upper body filling the center of the frame. One hand is tightly clamped over her nose and mouth, her eyes looking up with a frowning expression. She is squeezed close against the wall, flattening her shoulder against the narrow passage. The stairs angle upward behind and below her, partially visible at the lower frame edge. The cold, dim night darkness is broken by a faint blue-gray ambient exterior spill softly illuminating the rough structural surfaces, giving her skin a cold gray cast while foggy shadows thicken in the upper darkness.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#11 action: '갑자기 훅- 느껴지는 역겨운 냄새.'", "shot description: 계단 중간에 멈춰 서서 한 손으로 코와 입을 강하게 틀어막은 채 눈살을 찌푸린 수리영의 상체.", "C01 direction: three_quarter_right, hand clamped over mouth, gaze=up, state=alive" ], "visual_inferences": [ "fixed_element night_exterior_ambience: The exterior environment is engulfed in a cold, dim night darkness.", "A faint blue-gray ambient spill softly illuminates the rough structural surfaces.", "Damp stained concrete wall and worn stair treads reflect the moisture and chilly atmosphere." ], "creative_decisions": [ "Upper body framing placed tightly against the wall to emphasize confinement and the overwhelming nature of the smell.", "Cold gray cast on the skin to highlight the suspenseful, ominous mood of the night stairwell." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "handrail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "passage", "usage_kind": "anchor", "source_phrase": "flattening her shoulder against the narrow passage" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "The stairs angle upward behind and below her" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "squeezed close against the wall" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "58d8dd4eaa6288e8", "owned_hash": "0fb905c6957715a7", "camera_direction_hash": "315498f7799d5bec", "owned_usage_hash": "19bb2a3c480120b7", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "passage", "violating_phrase": "flattening her shoulder against the narrow passage", "reason": "The passage is referenced solely to describe the character's tight posture and physical contact, acting as a spatial anchor rather than demanding a redraw.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "The stairs angle upward behind and below her", "reason": "The stairs are mentioned to define the compositional background and angle behind the character without requesting any new generation or modification, confirming the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "squeezed close against the wall", "reason": "The wall is used as a spatial reference for the character's position and physical contact, matching the producer's declaration as an anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 2, "visible_entities": [ "C01", "L10" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 11, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Midway up the exterior stairs, the handheld camera stays at chest height and arm's length, squeezed by the stair wall into a tight three-quarter frame of 수리영. Her upper body fills the center, one hand clamped over her nose and mouth while the stair edges blur at the margins, implying confinement without opening the space.", "lighting_mood": "Cold stairwell darkness is broken by weak exterior spill, giving her skin a gray cast while shadows thicken toward the rooftop above.", "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": [ "handrail", "landing", "parapet", "passage", "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": "night_exterior_ambience", "element_type": "environment_state", "character_name": "", "description": "The exterior environment is engulfed in a cold, dim night darkness. A faint blue-gray ambient spill softly illuminates the rough structural surfaces, creating a consistently chilly and ominous atmosphere across the location.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "Cold stairwell darkness is broken by weak exterior spill", "A weak blue-gray rooftop ambience grazes the rust" ], "visual_inferences": [ "The weak exterior spill on the stairs and the blue-gray ambience on the rooftop originate from the same night lighting condition, tying the two spaces together." ], "creative_decisions": [ "Utilizing a cohesive cold, blue-gray ambient tone to connect the spatial progression from the stairs to the rooftop while heightening the suspenseful mood." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [ { "kind": "character_outlook", "id": "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": "2118a0fc55b1ea7d" }, { "scene_index": 11, "heading": "S#11. 옥탑 방 밖 / N", "beat_title": "불길한 어둠을 품은 반쯤 열린 철문", "representative_moment": "A rusty, half-open iron door stands at the center, revealing an impenetrable pitch-black interior under dim blue-gray night lighting.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "eye-level close framing with a slight diagonal angle, cold blue-gray ambient contrasted by distant gold spill", "t2i_prompt": "Photorealistic cinematic still. A half-open rusty iron entrance door stands in the middle-center midground, revealing a dense dark room interior occupying the middle-center background. Captured from a slight diagonal at eye level, the close framing tightly isolates the corroded metal surface of the door and the widening black gap. The exterior environment is engulfed in a cold, dim night darkness, with a faint blue-gray ambient spill softly illuminating the rough structural surfaces and grazing the rust. The room beyond remains pure black, opaque and still, edged with an ominous gold spill from distant city light. The weathered gray concrete landing is visible along the lower frame edge, anchoring the chilly, damp night atmosphere.", "outfit_assignments": [], "source_facts": [ "옥상에 올라와 보니 반쯤 열린 현관 철문.", "Shot description: half-open, old, rusty iron entrance door revealing dense interior darkness. Suri-young is EXCLUDED." ], "visual_inferences": [ "fixed_element night_exterior_ambience: The exterior environment is engulfed in a cold, dim night darkness. A faint blue-gray ambient spill softly illuminates the rough structural surfaces.", "The interior darkness is completely unlit, acting as an ominous void that replaces character presence." ], "creative_decisions": [ "Using close framing to strictly isolate the threshold of the door and the black gap, keeping the focus entirely on the unseen threat.", "Highlighting the corroded rust with the cold blue-gray night spill to enhance the tactile grit of the coastal setting." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [], "owned_object_usage": [ { "owned_token": "handrail", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "passage", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e670de1b60588f3e", "owned_hash": "0fb905c6957715a7", "camera_direction_hash": "5252a4fd874ca27e", "owned_usage_hash": "4b9662f6814ede7a", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 11, "dependency_reason": "exact_background ref to maintain the same rooftop space and lighting conditions established in the prior shot", "_shot_index": 4, "visible_entities": [], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 11, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "On the rooftop, the camera faces the old iron entrance door from a slight diagonal at eye level and continues a slow dolly-in toward the half-open gap. 수리영 is excluded entirely; the rusty door sits center frame with the black interior widening in the middle third until the darkness feels like the subject looking back.", "lighting_mood": "A weak blue-gray rooftop ambience grazes the rust while the room beyond remains nearly pure black, edged with ominous gold spill from distant city light.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "dense dark room interior", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "half-open rusty iron entrance door", "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": 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": "L10B01", "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": "night_exterior_ambience", "element_type": "environment_state", "character_name": "", "description": "The exterior environment is engulfed in a cold, dim night darkness. A faint blue-gray ambient spill softly illuminates the rough structural surfaces, creating a consistently chilly and ominous atmosphere across the location.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "Cold stairwell darkness is broken by weak exterior spill", "A weak blue-gray rooftop ambience grazes the rust" ], "visual_inferences": [ "The weak exterior spill on the stairs and the blue-gray ambience on the rooftop originate from the same night lighting condition, tying the two spaces together." ], "creative_decisions": [ "Utilizing a cohesive cold, blue-gray ambient tone to connect the spatial progression from the stairs to the rooftop while heightening the suspenseful mood." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 11, "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": [], "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": "e37190b2f05ed81c" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / N", "beat_title": "Discovering the Mother's Body", "representative_moment": "C03 slumped lifelessly on the floor beside the drawn-back curtain, surrounded by dark blood traces.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide framing from low-to-mid height, muted gray shadows with sickly gold and faint TV flicker", "t2i_prompt": "Photorealistic cinematic still. Inside the dimly lit bedroom, captured from a low-to-mid height wide angle. C03O04 in a blood-stained knit sweater, a Korean woman in her early 40s with black mid-length hair and soft facial contours, seated lifelessly on the floor at a three-quarter angle from the right. Her head is bowed low, gazing down. Her motionless body is slumped slightly off-center beside the drawn-back curtain, its wrinkled fabric hanging along one side, turned partly toward the camera and catching a sickly gold edge. The rumpled foot of the bed occupies the near side of the room in the lower foreground. The upper floor surface faces the camera, showing smeared dark marks leading toward the corner. The room is enveloped in dark, muted gray shadows with a cold and sickly wash of faint light, illuminated by a weak practical glow and residual TV flicker spilling from behind the camera.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O04" } ], "source_facts": [ "S#12 action: 피묻은 발자국은 커튼으로 이어져 있다 호흡을 가다듬고 커튼을 젖히자 앉은 채로 고개를 떨군 채 죽어있는 엄마", "Shot 6 description: 젖혀진 커튼 옆으로, 바닥에 주저앉은 채 고개를 푹 숙이고 미동도 없는 민숙의 시신." ], "visual_inferences": [ "fixed_element: dim_gray_room_lighting - The room's environment is enveloped in dark, muted gray shadows with a cold and sickly wash of faint light.", "TV flicker casting residual light from outside the bedroom context." ], "creative_decisions": [ "Wide low-to-mid height framing to capture the dead body beside the curtain and the dark traces on the floor simultaneously without rushing the shock.", "Placing the foot of the bed in the foreground to anchor the domestic spatial scale and emphasize the perspective leading toward the body." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "The rumpled foot of the bed occupies the near side of the room in the lower foreground" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "beside the drawn-back curtain, its wrinkled fabric hanging along one side, turned partly toward the camera and catching a sickly gold edge" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "The upper floor surface faces the camera, showing smeared dark marks leading toward the corner" }, { "owned_token": "light fixture", "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": "20a18ae683851f72", "owned_hash": "bd551bcd5dc5a9d2", "camera_direction_hash": "b7c357db5b50fc48", "owned_usage_hash": "6095ff1bcc8e679f", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "The rumpled foot of the bed occupies the near side of the room in the lower foreground", "reason": "The prompt mentions the foot of the bed to establish the foreground composition without any redraw instructions, which matches the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "curtain", "violating_phrase": "beside the drawn-back curtain, its wrinkled fabric hanging along one side, turned partly toward the camera and catching a sickly gold edge", "reason": "The curtain is referenced as a spatial anchor for the character with lighting details, matching the producer's anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "seated lifelessly on the floor... The upper floor surface faces the camera, showing smeared dark marks leading toward the corner", "reason": "The prompt uses the floor as a physical anchor for the character and describes its surface texture details, which matches the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "Exact background location S5_Shot1 is referenced to maintain the established layout of the bedroom.", "_shot_index": 6, "visible_entities": [ "C03", "L05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 6 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Inside the bedroom near the foot of the bed, the camera holds low-to-mid height about six feet from the curtain corner, static after the reveal. 민숙's motionless seated body is placed slightly off-center beside the drawn-back curtain edge, while the bed foot and stained floor lead the eye toward her without rushing the shock.", "lighting_mood": "Dim practical light and residual TV flicker leave the room in muted gray shadow, with restrained red accents in the floor traces and a sickly gold edge on the curtain.", "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", "L05" ], "allowed_outlook_pairs": [ { "character_id": "C03", "outlook_id": "O04" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L05B05", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "curtain", "doll", "floor", "light fixture", "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": "dim_gray_room_lighting", "element_type": "environment_state", "character_name": "", "description": "The room's environment is enveloped in dark, muted gray shadows with a cold and sickly wash of faint light. The overall illumination remains extremely dim, keeping the surrounding areas steeped in deep, black shadows that emphasize a grim and monochromatic atmosphere across the space.", "applies_to_shots": [ 6, 11, 14 ], "element_scope": "full", "source_facts": [ "muted gray shadow (Shot 6)", "low, cold gray wash, surrounding shadows stay deep (Shot 11)", "Sickly interior shadow, gray-black room (Shot 14)" ], "visual_inferences": [ "The minimal light source creates a uniform cold gray tone that persists whether viewing the floor, the body, or the upper walls." ], "creative_decisions": [ "Applying a consistent sickly gray shadow across all angles to visually unify the isolated horror elements within the room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "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": "L05B05", "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": "76c3dad702184b42" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / 수리영 방", "beat_title": "시신의 손에 쥐어진 단서", "representative_moment": "A close-up insert of C03's lifeless hand tightly clutching P03 against the dirty floor.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low floor-level insert, dark muted gray shadows with a cold and sickly wash", "t2i_prompt": "Photorealistic cinematic still. Low beside the body and close to the floor, an insert framing the lifeless, rigid hand of C03, a Korean woman in her early 40s. The stiff fingers extending from a dark knit sleeve tightly clutch P03, a half-crumpled photograph, in the central lower half of the frame. The printed face of P03 is partly visible to the camera but bent and obscured by the fingers, revealing only fragments of its image. The hand rests on the dirty floor surface beneath, its upper surface facing the camera and showing small dark smears and pressure shadows around the knuckles. A narrow fold of a drawn-back curtain lies out of focus at the edge, turned diagonally toward the lens. The environment is enveloped in dark, muted gray shadows with a low, cold and sickly wash of faint light catching the knuckles and photo creases, while surrounding areas stay steeped in deep, black shadows.", "outfit_assignments": [ { "character_id": "C03", "outlook_id": "O04" } ], "source_facts": [ "shot description: 축 늘어진 민숙의 굳은 손 안에 반쯤 구겨진 사진 한 장이 쥐어져 있는 클로즈업", "scene text: 엄마 손안에 구겨진 사진 한 장이 보인다." ], "visual_inferences": [ "fixed_element dim_gray_room_lighting: muted gray shadow, cold sickly wash", "dark knit sleeve visible near the wrist due to the low floor-level angle", "pressure shadows and dirt on the floor anchor the rigid hand physically in the space" ], "creative_decisions": [ "Insert framing focused strictly on the hand, photo, and floor texture to comply with the base_id_required policy and DP instructions", "Placing the drawn-back curtain at the frame edge to provide depth and connect back to the earlier reveal" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "prop", "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "drawn-back curtain" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "the dirty floor surface beneath, its upper surface facing the camera" }, { "owned_token": "light fixture", "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": "479c794fe9a1ec49", "owned_hash": "bd551bcd5dc5a9d2", "camera_direction_hash": "2d57c759b918d193", "owned_usage_hash": "39013156aa16865e", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "curtain", "violating_phrase": "A narrow fold of a drawn-back curtain lies out of focus at the edge", "reason": "The prompt references the curtain as an out-of-focus background anchor element, which aligns with the producer's echo as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "The hand rests on the dirty floor surface beneath", "reason": "The prompt mentions the floor surface only as a physical plane supporting the character's hand, matching the producer's echo as an anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "S5_Shot1 is referenced as exact_background for maintaining the layout and floor elements of the room.", "_shot_index": 11, "visible_entities": [ "C03", "P03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 11 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Low beside 민숙's body and close to the floor, the dolly-in tightens into an insert of her rigid hand and the half-crumpled photograph. The photograph sits in the central lower half but does not cover the whole frame; stiff fingers, a sliver of sleeve, and dirty floor texture remain visible around it, keeping the object tied to the body.", "lighting_mood": "A low, cold gray wash catches the knuckles and photo creases while surrounding shadows stay deep, with the scene's red threat held just outside the frame.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C03", "P03" ], "allowed_outlook_pairs": [ { "character_id": "C03", "outlook_id": "O04" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C03", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The shot frames only 민숙's dead hand and the photograph, so a full outlook reference would overconstrain the image and risk forcing her face or body into frame." } ], "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": "L05B05", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "curtain", "doll", "floor", "light fixture", "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": "dim_gray_room_lighting", "element_type": "environment_state", "character_name": "", "description": "The room's environment is enveloped in dark, muted gray shadows with a cold and sickly wash of faint light. The overall illumination remains extremely dim, keeping the surrounding areas steeped in deep, black shadows that emphasize a grim and monochromatic atmosphere across the space.", "applies_to_shots": [ 6, 11, 14 ], "element_scope": "full", "source_facts": [ "muted gray shadow (Shot 6)", "low, cold gray wash, surrounding shadows stay deep (Shot 11)", "Sickly interior shadow, gray-black room (Shot 14)" ], "visual_inferences": [ "The minimal light source creates a uniform cold gray tone that persists whether viewing the floor, the body, or the upper walls." ], "creative_decisions": [ "Applying a consistent sickly gray shadow across all angles to visually unify the isolated horror elements within the room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "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", "id": "C03", "policy": "required" }, { "kind": "background", "id": "L05B05", "policy": "required" }, { "kind": "prop", "id": "P03", "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": [ 11 ], "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": 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": "b891c5295cd1d6ba" }, { "scene_index": 12, "heading": "S#12. 옥탑방 안 / 수리영 방", "beat_title": "벽에 생겨난 붉은 원 문양", "representative_moment": "A close view of the huge red circle mark (P02) painted in blood on the dark wall above the bed.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Slightly low-angled close framing from shoulder height, muted gray shadows with a cold sickly wash, highly saturated red focal point.", "t2i_prompt": "Photorealistic cinematic still. [: A bare wall surface above a bed, enveloped in dark, muted gray shadows with a cold and sickly wash of faint light, the surrounding air steeped in deep black shadows]. P02, a huge uneven red circle painted like wet blood, sits almost centered but slightly high on the camera-facing wall, its full shape clearly visible. The dim upper edge of the bed barely grazes the lower foreground edge as a scale reference. The perspective looks slightly upward from shoulder height. P02 catches the eye as the only saturated color in the grim, monochromatic gray-black space. No person is visible in the frame.", "outfit_assignments": [], "source_facts": [ "S12 action: A red circle appears on the wall above the bed.", "Shot description: close-up of the red circle mark, freshly painted like blood on the wall above the bed." ], "visual_inferences": [ "fixed_element dim_gray_room_lighting: muted gray shadow, cold sickly wash, grim and monochromatic atmosphere.", "The camera perspective from shoulder height slightly tilted up places the bed edge at the very bottom of the frame for scale." ], "creative_decisions": [ "Applied skipped_close_framing to omit background room context and tightly isolate the wall and the mark.", "Used stark color contrast by making P02 the only saturated element against the gray-black shadows to emphasize supernatural dread." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "prop" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "light fixture", "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": "dc2999f873511078", "owned_hash": "bd551bcd5dc5a9d2", "camera_direction_hash": "9f6d1934d734adb2", "owned_usage_hash": "61857106472c3117", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "previous_shot_refs indicates S5_Shot8 for location context, though background binding is skipped_close_framing for this specific shot", "_shot_index": 14, "visible_entities": [ "P02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 12, "shot_index": 14 }, "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, the camera completes the upward tilt and locks into a close, slightly low-angled composition of the wall above the bed. The huge red circle sits almost centered but a little high in frame, with the bed edge barely grazing the lower foreground as scale reference; no person is allowed into the image, making the mark feel like the room itself has spoken.", "lighting_mood": "Sickly interior shadow is cut by a faint cold spill, while the red circle catches the eye as the only saturated color in the gray-black room.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": 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": [ "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": "skipped_close_framing", "bg_id": "L05B05", "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": "dim_gray_room_lighting", "element_type": "environment_state", "character_name": "", "description": "The room's environment is enveloped in dark, muted gray shadows with a cold and sickly wash of faint light. The overall illumination remains extremely dim, keeping the surrounding areas steeped in deep, black shadows that emphasize a grim and monochromatic atmosphere across the space.", "applies_to_shots": [ 6, 11, 14 ], "element_scope": "full", "source_facts": [ "muted gray shadow (Shot 6)", "low, cold gray wash, surrounding shadows stay deep (Shot 11)", "Sickly interior shadow, gray-black room (Shot 14)" ], "visual_inferences": [ "The minimal light source creates a uniform cold gray tone that persists whether viewing the floor, the body, or the upper walls." ], "creative_decisions": [ "Applying a consistent sickly gray shadow across all angles to visually unify the isolated horror elements within the room." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "shot_index": 8, "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": "P02", "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": 12, "shot_indices": [ 14 ], "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": 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": "137edf11d46ed22c" }, { "scene_index": 13, "heading": "S#13. 옥탑방 외곽 / N", "beat_title": "넋이 나간 채 멈춰선 수리영", "representative_moment": "C01 stands frozen at the bottom of the stairs, her blank face washed by harsh red and blue siren flashes.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level frontal wide frame with high-contrast red and blue emergency lighting against cold dark night.", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, an Asian woman in her early 20s, with long straight black hair and dark eyes, centered in the midground facing the camera, frozen mid-breath. She stands gazing into the distance with blank unfocused eyes. Captured at eye level in a steady frontal wide frame, her shoes rest on the wet pavement in the lower frame. Behind her, the open dim mouth of the stair recedes upward into the deep shadows of the building. The cold night environment is starkly illuminated by pulsing red and blue emergency siren lights striking from the camera side, leaving her skin drained and shock-pale. The damp pavement under her feet reflects broken red and blue streaks, while the surrounding road and yard sink into the coastal pitch blackness.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "비명을 지르며 옥탑방을 빠져나오는 수리영, 허겁지겁 계단을 내려간다.", "동시에 사이렌이 울리고~.", "초점을 잃고 멍하니 서있는 수리영" ], "visual_inferences": [ "fixed_element: flashing_siren_lights creating harsh red-blue contrast against the night", "C01 is in a state of shock, frozen in place at the bottom of the outdoor stairs", "The damp roadside reflecting the intense emergency lights" ], "creative_decisions": [ "Used a steady frontal wide frame at eye level to anchor her frozen posture within the dark environment, showing both her full figure and the ominous stair exit.", "Accented the wet pavement to dynamically mirror the red and blue siren flashes, enhancing the eerie and dreadful atmosphere." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "building", "usage_kind": "anchor", "source_phrase": "into the deep shadows of the building" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pavement", "usage_kind": "anchor", "source_phrase": "shoes rest on the wet pavement in the lower frame" }, { "owned_token": "railing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "road", "usage_kind": "anchor", "source_phrase": "surrounding road and yard sink into the coastal pitch blackness" }, { "owned_token": "stair", "usage_kind": "anchor", "source_phrase": "open dim mouth of the stair recedes upward" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "yard", "usage_kind": "anchor", "source_phrase": "surrounding road and yard sink into the coastal pitch blackness" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "9e943bd0779abafe", "owned_hash": "04c70a555c10231f", "camera_direction_hash": "bc3f54cf90217bea", "owned_usage_hash": "3d16447ca7e73380", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "building", "violating_phrase": "into the deep shadows of the building", "reason": "The building is mentioned only to establish shadows as a spatial background anchor, matching the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "pavement", "violating_phrase": "shoes rest on the wet pavement in the lower frame", "reason": "The pavement serves as a ground surface anchor for the character and reflections, which is consistent with the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "road", "violating_phrase": "surrounding road and yard sink into the coastal pitch blackness", "reason": "The road is referenced only to describe the dark background environment, acting as an anchor, matching the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "stair", "violating_phrase": "open dim mouth of the stair recedes upward", "reason": "The stair is mentioned as a spatial backdrop anchor behind the character, aligning with the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "yard", "violating_phrase": "surrounding road and yard sink into the coastal pitch blackness", "reason": "The yard is referenced to describe the surrounding dark environment as a spatial anchor, 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's exact layout or camera angle is needed for this specific new moment and framing.", "_shot_index": 2, "visible_entities": [ "C01", "L11" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 13, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At eye level on the roadside, three meters in front of 수리영, the camera holds a steady frontal wide frame after the previous unstable descent. 수리영 is centered in the midground with the stair mouth and dark street behind her, while pulsing siren light from behind the camera washes across her blank face and makes the surrounding night flicker without moving the frame.", "lighting_mood": "Cold night gray is intermittently broken by hard red-blue siren pulses, leaving her skin drained and shock-pale between flashes.", "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": 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": "L11B01", "reference_usage": "exact_background", "owned_objects": [ "building", "curb", "gate", "landing", "pavement", "railing", "road", "stair", "wall", "yard" ], "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": "flashing_siren_lights", "element_type": "environment_state", "character_name": "", "description": "The dark, cold nighttime environment is starkly illuminated by pulsing red and blue emergency siren lights. The intense emergency flashes rhythmically break through the deep shadows of the coastal night, creating a harsh, dramatic contrast against the surrounding blackness.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "Cold night gray is intermittently broken by hard red-blue siren pulses", "makes the surrounding night flicker", "Blue-red emergency flashes", "the rest of the frame sinks into cold coastal black" ], "visual_inferences": [ "The alternating red and blue light casts dynamic, shifting colored highlights across any subject or surface in the area.", "The background beyond the immediate reach of the siren lights remains pitch black." ], "creative_decisions": [ "Maximizing the harsh color contrast of the red and blue emergency lights against the dark night to visually amplify the emotional shock and panic of 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": "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": "f9f8731ef122445e" }, { "scene_index": 13, "heading": "S#13. 옥탑방 외곽 / N", "beat_title": "충격에 빠져 넋이 나간 수리영과 붉은 경광등", "representative_moment": "C01, trembling with glassy eyes and parted lips, looks up toward a blurred shoulder in the upper frame edge, her face illuminated by harsh red and blue siren lights.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low angle extreme close-up, dramatic red-blue emergency lighting contrast, shallow depth of field", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, an East Asian young woman in her 20s, occupies the middle-center foreground, looking toward the blurred shoulder of a middle-aged woman above the upper frame edge. Framed from a low close angle just below her eyeline looking upward, C01O01's tearful face fills the frame at a three-quarter angle from the left. Her chin is lifted and trembling, lips slightly parted, glassy eyes wide with shock. The secondary figure is only suggested as an out-of-focus shoulder and jawline slice entering the top frame boundary, leaning toward the camera. The dark, cold nighttime environment is starkly illuminated by pulsing red and blue emergency siren lights coming from camera-side street level. The intense emergency flashes rhythmically break through the deep shadows of the coastal night, skimming her tears and casting harsh red and blue colored highlights across her pale face. [: The immediate background sinks into a soft, indistinct night-black void, marked only by sparse, blurred bokeh points of distant streetlights].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영(C01) looks up, tearful with parted lips and shocked expression", "Red and blue emergency siren lights pulse in the dark nighttime environment", "The flashing light comes from the camera-side street level and strikes the front of her face" ], "visual_inferences": [ "Low camera position captures C01's face occupying most of the frame", "Secondary character (혜수) is obscured as an out-of-focus shoulder/jawline slice at the upper edge due to close framing", "Background is blurred into a soft night-black void to isolate the emotional intensity on C01's face" ], "creative_decisions": [ "Utilized shallow depth of field to strictly adhere to close framing rules, rendering the background and the secondary character as soft blurs", "Harsh, rhythmic red and blue highlights on the skin to visually amplify the emotional shock and panic without exposing the surrounding area" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "building", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curb", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "landing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pavement", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "railing", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "road", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "stair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "yard", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d4e9e7c64c0e6df3", "owned_hash": "04c70a555c10231f", "camera_direction_hash": "5d28820db88b0f72", "owned_usage_hash": "ff2bdb9823927311", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 13, "dependency_reason": "Same location and timeline as the previous shot, but background reference is deliberately skipped due to close framing, isolating the emotional focus purely on the character.", "_shot_index": 4, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 13, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera finishes its dolly-in from a low close position just below 수리영's eyeline, angling upward so her tearful face occupies most of the frame as she looks toward 혜수. 혜수 is only suggested as a blurred shoulder and jawline at the upper edge of 수리영's gaze path, keeping the emotional pressure trapped on 수리영's parted lips and glassy eyes.", "lighting_mood": "Blue-red emergency flashes skim her tears while the rest of the frame sinks into cold coastal black.", "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": "foreground", "gesture_action": "looks_toward", "gesture_target_label": "Hyesu above frame edge", "label": "수리영", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": 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": "flashing_siren_lights", "element_type": "environment_state", "character_name": "", "description": "The dark, cold nighttime environment is starkly illuminated by pulsing red and blue emergency siren lights. The intense emergency flashes rhythmically break through the deep shadows of the coastal night, creating a harsh, dramatic contrast against the surrounding blackness.", "applies_to_shots": [ 2, 4 ], "element_scope": "full", "source_facts": [ "Cold night gray is intermittently broken by hard red-blue siren pulses", "makes the surrounding night flicker", "Blue-red emergency flashes", "the rest of the frame sinks into cold coastal black" ], "visual_inferences": [ "The alternating red and blue light casts dynamic, shifting colored highlights across any subject or surface in the area.", "The background beyond the immediate reach of the siren lights remains pitch black." ], "creative_decisions": [ "Maximizing the harsh color contrast of the red and blue emergency lights against the dark night to visually amplify the emotional shock and panic of the scene." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 13, "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": [ { "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": "4ed28b0aabd276b2" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "사라진 흔적", "representative_moment": "핏자국과 시신이 모두 사라진 채 지나칠 정도로 깨끗하게 정돈된 텅 빈 방 안.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "viewed from chest height just inside the room, locked frame aimed toward the bed and curtain wall, muted cold gray light", "t2i_prompt": "Photorealistic cinematic still. Viewed from chest height just inside the room, the frame is aimed toward the bed and curtain wall. The room is uncomfortably clean and stark. The empty curtain area occupies the right half of the frame, its parted inner folds facing the camera and opening toward a bare empty corner. The spotless, bare floor spreads across the lower third with absolutely no traces of blood or debris, its clean surface faintly reflecting the bed legs and curtain shadows instead of blood. The bed sits in the left-midground, neatly arranged and undisturbed, its side and top edge facing the camera from the left side of the frame. The wall above the bed is bare, pale, and clean, the camera-facing surface showing no red circle or stain. A backpack and a doll rest quietly in the space. The interior is illuminated by a muted, flat, and cold gray light that strips away warmth, creating a clinical and unnerving atmosphere focusing on the glaring absence of evidence.", "outfit_assignments": [], "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고, 침대 위에 붉은 표식은 물론 커튼 뒤의 시신도 감쪽 같이 사라졌다.", "수리영 방안에 놓여진 백팩, 인형.", "Flat cold room light strips away warmth", "spotless floor spreads across the lower third" ], "visual_inferences": [ "fixed_element cold_spotless_room_environment: The interior of the room is uncomfortably clean and stark.", "The physical absence of evidence is visually reinforced by the perfectly maintained surfaces of the room, leaving no visual noise." ], "creative_decisions": [ "Positioning the camera from just inside the room to frame the stark emptiness where the body and blood should have been.", "Using a flat, cold gray light to emphasize the unnerving cleanliness and clinical atmosphere." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "anchor", "source_phrase": "A backpack and a doll rest quietly" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "bed sits in the left-midground" }, { "owned_token": "ceiling light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "anchor", "source_phrase": "empty curtain area occupies the right half of the frame" }, { "owned_token": "doll", "usage_kind": "anchor", "source_phrase": "A backpack and a doll rest quietly" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "spotless, bare floor spreads across the lower third" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "wall above the bed is bare" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "81e0104c1bc2de1a", "owned_hash": "ed55a4552fb079e7", "camera_direction_hash": "ff7ceaaa785251f2", "owned_usage_hash": "c7c7b178bee50c08", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "backpack", "violating_phrase": "A backpack and a doll rest quietly in the space", "reason": "The prompt describes the backpack resting quietly as a static element in the room to establish atmosphere, which is a spatial reference matching the producer's anchor classification.", "verdict": "anchor_reference" }, { "owned_object": "bed", "violating_phrase": "The bed sits in the left-midground, neatly arranged and undisturbed", "reason": "The prompt specifies the bed's location and neat state to emphasize the clean, undisturbed environment, serving purely as a compositional anchor.", "verdict": "anchor_reference" }, { "owned_object": "curtain", "violating_phrase": "The empty curtain area occupies the right half of the frame", "reason": "The curtain is used as a spatial and compositional reference to define the frame's layout, which is an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "doll", "violating_phrase": "A backpack and a doll rest quietly in the space", "reason": "The doll's quiet presence is described to enhance the unnerving, clean atmosphere of the existing room, serving as an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "The spotless, bare floor spreads across the lower third", "reason": "The prompt describes the floor's empty, spotless surface to highlight the absence of debris, functioning as a key compositional anchor.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "The wall above the bed is bare, pale, and clean", "reason": "The wall's appearance is described to draw focus to the absence of markings, anchoring the narrative tone without suggesting a redraw.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 12, "dependency_reason": "같은 공간(수리영 옥탑방)에서 발생한 사건 직후의 모습으로, 이전 샷에서 구축된 방 구조와 시신이 있던 위치적 맥락의 일관성 유지가 필요함", "_shot_index": 4, "visible_entities": [ "L05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From just inside 수리영's room at chest height, the dolly-in settles into a wider, colder locked frame aimed toward the bed and curtain wall. The empty curtain area occupies the right half of the frame, the spotless floor spreads across the lower third, and the bed sits left-midground, all arranged to make the missing blood and body feel louder than any visible object.", "lighting_mood": "Flat cold room light strips away warmth, turning the immaculate order into something clinical and accusatory.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L05" ], "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": "L05B06", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "ceiling light", "curtain", "doll", "floor", "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": "cold_spotless_room_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of the room is uncomfortably clean and stark. A spotless, bare floor stretches across the area with absolutely no traces of blood, debris, or disturbance. The space is illuminated by a muted, flat, and cold gray light that strips away warmth, creating a clinical and unnerving atmosphere.", "applies_to_shots": [ 4, 8 ], "element_scope": "full", "source_facts": [ "spotless floor spreads across the lower third", "Flat cold room light strips away warmth", "patch of bare floor visible", "Muted gray interior light flattens the room" ], "visual_inferences": [ "The physical absence of evidence is visually reinforced by the perfectly maintained surfaces of the room, leaving no visual noise." ], "creative_decisions": [ "Consistent use of a cool, gray color palette across both wide and medium shots to amplify the chilling and isolated feeling of the environment." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 12, "shot_index": 14, "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": "L05B06", "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": "d99b98e22c0809e2" }, { "scene_index": 14, "heading": "S#14. 옥탑방 안 / N", "beat_title": "사라진 시신과 흔적 없는 방 앞에서 경악하는 수리영", "representative_moment": "수리영(C01)이 두 손으로 얼굴을 감싼 채 핏자국 하나 없이 깨끗한 방바닥을 내려다보며 충격에 빠져 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Medium framing from a low, seated chest-height observer, muted flat cold gray lighting", "t2i_prompt": "Photorealistic cinematic still. A low observer at seated chest height, torso and arms in frame, upper body occupying the slightly right-of-center area, a patch of bare floor visible at the lower frame edge. C01O01 in a casual gray windbreaker, a Korean young woman in her 20s with long straight black hair and dark eyes, positioned at three-quarter angle from the right, hands covering her face, looking down. The uncomfortably clean and stark interior surrounds her. The spotless, bare floor stretches across the lower area with absolutely no traces of blood, reflecting only the dull room light below her face. The space is illuminated by a muted, flat, and cold gray light that strips away warmth, casting a faint shadow under her hands and creating an unnerving atmosphere. Occupying the background, a backpack and a doll rest subtly out of place—the backpack's side angled toward the camera, the doll's front turned slightly away.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "지나칠 정도로 깨끗하게 정돈된 방안. 핏자국도 보이지 않고", "얼굴이 파리해지는 수리영... 수리영 방안에 놓여진 백팩, 인형. 약간씩 위치가 달라진", "Low at seated chest height and slightly off to one side... she is held slightly right of center" ], "visual_inferences": [ "fixed_element cold_spotless_room_environment: Muted gray interior light flattens the room", "C01 frozen in shock, hands covering face as a physical reaction to the impossible cleanliness of the floor", "backpack and doll positioned in the soft background to hint at the rearranged space" ], "creative_decisions": [ "Medium framing with a slight right-of-center composition emphasizes the empty floor she is staring at.", "Muted gray palette strips away warmth to reflect her chilling psychological realization." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "anchor", "source_phrase": "a backpack and a doll rest subtly out of place" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "ceiling light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "anchor", "source_phrase": "a backpack and a doll rest subtly out of place" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "a patch of bare floor visible at the lower frame edge" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "2f026c7a88a45c0b", "owned_hash": "ed55a4552fb079e7", "camera_direction_hash": "4cc094d8a5b31304", "owned_usage_hash": "c2b8bc3589278b94", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "floor", "violating_phrase": "a patch of bare floor visible at the lower frame edge", "reason": "The prompt describes the floor's visibility and spotless state at the bottom edge to anchor the composition, which aligns with the producer's declaration of an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "backpack", "violating_phrase": "a backpack and a doll rest subtly out of place", "reason": "The prompt references the backpack to specify its background position and angle relative to the camera as an anchor, which matches the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "doll", "violating_phrase": "a backpack and a doll rest subtly out of place", "reason": "The doll is mentioned to describe its placement and orientation in the background as a spatial anchor, which is consistent with the producer's anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "같은 공간의 직전 샷(Shot 4) 배경 연속성 유지, 정확한 환경(exact_background) 참조", "_shot_index": 8, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 14, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Low at seated chest height and slightly off to one side, the camera makes an almost imperceptible push toward 수리영's upper body. She is held slightly right of center with a patch of bare floor visible beneath her lowered gaze, so the empty space she cannot stop staring at becomes part of the composition.", "lighting_mood": "Muted gray interior light flattens the room, with faint shadow under her hands making her face feel sealed off from help.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L05B06", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "ceiling light", "curtain", "doll", "floor", "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": "cold_spotless_room_environment", "element_type": "environment_state", "character_name": "", "description": "The interior of the room is uncomfortably clean and stark. A spotless, bare floor stretches across the area with absolutely no traces of blood, debris, or disturbance. The space is illuminated by a muted, flat, and cold gray light that strips away warmth, creating a clinical and unnerving atmosphere.", "applies_to_shots": [ 4, 8 ], "element_scope": "full", "source_facts": [ "spotless floor spreads across the lower third", "Flat cold room light strips away warmth", "patch of bare floor visible", "Muted gray interior light flattens the room" ], "visual_inferences": [ "The physical absence of evidence is visually reinforced by the perfectly maintained surfaces of the room, leaving no visual noise." ], "creative_decisions": [ "Consistent use of a cool, gray color palette across both wide and medium shots to amplify the chilling and isolated feeling of the environment." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "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": "L05B06", "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": "3dad20752e03276d" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "멍하니 바다를 바라보는 수리영", "representative_moment": "짙은 어둠 속 해안가 벤치에 고개를 푹 숙이고 앉아 있는 수리영의 멍한 옆모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide framing from a low side angle, cool blue-black shadows contrasting with weak amber rim light", "t2i_prompt": "Photorealistic cinematic still. Viewed from a low side position at waist height a short distance away, C01O01 in a gray casual windbreaker, a Korean woman in her 20s with long straight black hair and a round face, sits in left profile, hunched left of center on a large, damp weather-resistant public bench. She is gazing down with her head sunk forward. The bench runs horizontally through the lower frame, anchoring her isolated figure. The environment is enveloped in cool, damp blue-black night shadows and pervasive gray coastal fog. A dim, sodium-tinted coastal lamp faces down toward the roadside, casting a weak amber rim light that traces the edge of her bowed profile against the dark surroundings. Behind her, the black, mist-heavy ocean is barely moving, opening like an empty witness under the desolate coastal atmosphere.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "바다를 바라보며 멍한 채로 앉아있는 수리영", "짙은 어둠이 깔린 해안가 벤치에 고개를 푹 숙인 채 멈춰 앉은 수리영의 멍한 옆모습" ], "visual_inferences": [ "fixed_element coastal_night_mist_and_lighting: blue-black night shadows, pervasive gray ocean mist, weak amber coastal lamp casting directional rim light", "fixed_element seaside_bench: stationary horizontal public bench serving as the physical anchor", "the vast, dark mist-heavy ocean emphasizes the character's psychological isolation and confusion" ], "creative_decisions": [ "low side waist-height camera angle to position the bench horizontally across the lower frame and isolate the subject against the dark ocean void", "amber sodium-tinted rim lighting strictly localized to trace her profile and hair, enhancing the gloomy blue-black ambient contrast" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "d4d8e206346baeb5", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "5f9892b9b4d8c7aa", "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", "L12" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From a low side position beside the seaside bench, waist height and about three meters away, the camera watches 수리영 in profile against the dark ocean. She sits left of center with her head sunk forward, the bench running horizontally through the lower frame and the black shoreline opening behind her like an empty witness.", "lighting_mood": "The scene is soaked in damp blue-black shadow, with only a weak amber coastal lamp tracing the edge of her bowed 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": [ "C01", "L12" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "coastal_night_mist_and_lighting", "element_type": "environment_state", "character_name": "", "description": "The environment is enveloped in cool, damp blue-black night shadows and pervasive gray ocean mist. Weak amber and gold-orange streetlamps provide a stark, warm contrast against the dark surroundings, casting directional rim light on subjects within the scene.", "applies_to_shots": [ 1, 3, 4 ], "element_scope": "full", "source_facts": [ "damp blue-black shadow", "weak amber coastal lamp", "ocean mist keeps the rest of the palette gray and cold", "thin gold-orange streetlamp", "cool coastal darkness", "gray mist", "sodium streaks" ], "visual_inferences": [ "The sea mist diffuses the streetlamp illumination, creating a hazy, heavy, and desolate coastal atmosphere." ], "creative_decisions": [ "Maintaining a consistent color palette of warm amber lighting cutting through the cool blue-black ocean darkness to emphasize the chilling and isolated mood." ], "confidence": "high" }, { "element_id": "seaside_bench", "element_type": "persistent_prop", "character_name": "", "description": "A stationary seaside bus stop bench situated against the backdrop of the dark ocean. It is positioned horizontally in the frame, serving as the physical structure the young woman initially sits on and subsequently rises from.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "bench running horizontally through the lower frame", "rising from the bench" ], "visual_inferences": [ "A standard public outdoor bench made of weather-resistant materials, large enough to seat multiple people." ], "creative_decisions": [ "Placing the bench as an anchoring physical prop in the foreground to ground the character's movement and highlight her departure." ], "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": "5ad9e5e873d472dc" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "굳어진 결심과 급작스러운 귀가", "representative_moment": "어두운 버스정류장 벤치에서 결연한 표정으로 일어서며 혜수를 향해 시선을 고정하는 수리영.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "medium framing at chest height focusing on the rising subject, illuminated by warm streetlamp rim light against cool ocean mist", "t2i_prompt": "Photorealistic cinematic still. At chest height, the composition centers on a damp seaside bench cutting across the lower frame. C01O01 in a gray windbreaker, an East Asian young woman with long black hair, at three-quarter angle from the right, fills the center-right of the frame, torso and head visible in mid-rise from the stationary seaside bus stop bench with a slight nod, her eyes fixed on the other figure. On the left margin as a partial presence, a Korean middle-aged woman in right profile stands near the shared bench, her arm lowering a small drink bottle whose side faces the camera briefly. Their cheeks catch a thin gold-orange streetlamp rim light, while pervasive gray ocean mist keeps the rest of the palette gray and cold against the dark blue-black night shoreline in the background.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "shot description: 수리영이 벤치에서 일어선 자세로 혜수를 향해 고개를 끄덕인 채 굳은 결연한 표정", "S#15 action: 혜수, 사온 음료수를 수리영에게 건네며", "S#15 action: 훅 일어나며 집에 가봐야 겠어" ], "visual_inferences": [ "fixed_element persistent_prop: mid-rise from the stationary seaside bus stop bench", "chest height medium framing places the bench edge in the lower frame, cutting across", "혜수 (unregistered) rendered as a partial presence on the left margin using a generic demographic descriptor" ], "creative_decisions": [ "warm gold-orange rim light contrasting with the blue-black night shadows to emphasize the chilling and isolated mood", "keeping the older woman marginalized in the frame to concentrate the visual weight on C01O01's hardened resolve", "drink bottle serves as a small color contrast near the left edge without overtaking the frame" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "1d530efe62ee34fd", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "9a3602376b50917b", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 15, "dependency_reason": "동일 씬 내 동일 장소(버스정류장 및 해안가) 배경 연속성 유지", "_shot_index": 3, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "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 chest height just off 수리영's front side, the camera pushes into a tighter medium frame as she rises from the bench and nods toward 혜수 near the edge of her eyeline. 수리영 fills the center-right of the frame, while 혜수 is kept near the left margin as a partial presence, letting the resolve harden primarily in 수리영's face rather than becoming an equal two-shot.", "lighting_mood": "A thin gold-orange streetlamp rim catches their cheeks while the ocean mist keeps the rest of the palette gray and cold.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "coastal_night_mist_and_lighting", "element_type": "environment_state", "character_name": "", "description": "The environment is enveloped in cool, damp blue-black night shadows and pervasive gray ocean mist. Weak amber and gold-orange streetlamps provide a stark, warm contrast against the dark surroundings, casting directional rim light on subjects within the scene.", "applies_to_shots": [ 1, 3, 4 ], "element_scope": "full", "source_facts": [ "damp blue-black shadow", "weak amber coastal lamp", "ocean mist keeps the rest of the palette gray and cold", "thin gold-orange streetlamp", "cool coastal darkness", "gray mist", "sodium streaks" ], "visual_inferences": [ "The sea mist diffuses the streetlamp illumination, creating a hazy, heavy, and desolate coastal atmosphere." ], "creative_decisions": [ "Maintaining a consistent color palette of warm amber lighting cutting through the cool blue-black ocean darkness to emphasize the chilling and isolated mood." ], "confidence": "high" }, { "element_id": "seaside_bench", "element_type": "persistent_prop", "character_name": "", "description": "A stationary seaside bus stop bench situated against the backdrop of the dark ocean. It is positioned horizontally in the frame, serving as the physical structure the young woman initially sits on and subsequently rises from.", "applies_to_shots": [ 1, 3 ], "element_scope": "full", "source_facts": [ "bench running horizontally through the lower frame", "rising from the bench" ], "visual_inferences": [ "A standard public outdoor bench made of weather-resistant materials, large enough to seat multiple people." ], "creative_decisions": [ "Placing the bench as an anchoring physical prop in the foreground to ground the character's movement and highlight her departure." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 15, "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": "1c73840f0c1076d8" }, { "scene_index": 15, "heading": "S#15. 해안가. 버스정류장 부근 / N", "beat_title": "불길함을 떨치고 집으로 달려가는 수리영", "representative_moment": "어두운 도로를 향해 전력으로 뛰어가는 C01의 측면을 따라가는 시선.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "torso-height tracking wide shot, cool coastal darkness with motion-blurred background", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, a Korean young woman in her early 20s, appears as a running figure in the middle-center midground, moving toward the dark road to the right, with the dark roadside direction stretching into the middle-right background. Observed from a torso-height side-tracking perspective, her full body is framed in right profile, caught mid-stride with one foot just lifted off the ground, her gaze fixed off-screen. Her dark long straight hair and the jacket flaps trail backward in the direction opposite to her forward motion. The surrounding environment is enveloped in cool, damp blue-black night shadows and thick gray ocean mist. Weak amber and gold-orange streetlamps cast a sharp, warm directional rim light over her moving silhouette, while the dim asphalt and occasional sodium streaks blur horizontally behind her to emphasize a voyeuristic sense of being watched from a parallel distance.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#15 action: '수리영 (달려간다)'", "shot description: 'one foot lifted and her clothing trailing backward'", "lighting_mood: 'Cool coastal darkness swallows the road, with occasional sodium streaks and gray mist'" ], "visual_inferences": [ "fixed_element coastal_night_mist_and_lighting: blue-black night shadows and weak amber/gold-orange streetlamps casting directional rim light", "mid-stride action freeze with hair and clothing preserving motion direction", "voyeuristic camera perspective maintained by horizontal motion streaks in the background" ], "creative_decisions": [ "side-tracking perspective framed at torso height to capture the full running silhouette within the wide environment", "warm rim light contrasting against the cool mist to separate the primary subject from the dark background" ], "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": "a7d088d7952add16", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "5e94e6d8d937aaa5", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 15, "dependency_reason": "Maintains exact environmental consistency (coastal_night_mist_and_lighting) with the prior shot in the same location.", "_shot_index": 4, "visible_entities": [ "C01", "L12" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 15, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera pivots into a parallel side track along the dark roadside at torso height, three to four meters from 수리영. She is caught mid-stride in the middle of the frame with one foot lifted and her clothing trailing backward, while the road streaks horizontally behind her to preserve the sense of being watched from a distance.", "lighting_mood": "Cool coastal darkness swallows the road, with occasional sodium streaks and gray mist cutting across her running silhouette.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "dark roadside direction", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "dark road to the right", "label": "수리영 running", "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", "L12" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "not_applicable", "bg_id": null, "reference_usage": "none", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": false, "constraints": [ "no chain_bg is bound to this shot — describe the location freely without referencing a background image" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "coastal_night_mist_and_lighting", "element_type": "environment_state", "character_name": "", "description": "The environment is enveloped in cool, damp blue-black night shadows and pervasive gray ocean mist. Weak amber and gold-orange streetlamps provide a stark, warm contrast against the dark surroundings, casting directional rim light on subjects within the scene.", "applies_to_shots": [ 1, 3, 4 ], "element_scope": "full", "source_facts": [ "damp blue-black shadow", "weak amber coastal lamp", "ocean mist keeps the rest of the palette gray and cold", "thin gold-orange streetlamp", "cool coastal darkness", "gray mist", "sodium streaks" ], "visual_inferences": [ "The sea mist diffuses the streetlamp illumination, creating a hazy, heavy, and desolate coastal atmosphere." ], "creative_decisions": [ "Maintaining a consistent color palette of warm amber lighting cutting through the cool blue-black ocean darkness to emphasize the chilling and isolated mood." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 15, "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": "eda43faa641b2d8b" }, { "scene_index": 16, "heading": "S#16. 경찰서 사무실 / N", "beat_title": "김형사에게 민숙의 번호를 내미는 혜수", "representative_moment": "혜수가 김형사의 책상 위로 휴대전화(P01) 액정을 내민 채, 단호한 눈빛으로 그를 내려다본다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "over-the-shoulder medium dolly-in shot with harsh greenish fluorescent lighting", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a navy safari jacket, a Korean woman in her early 40s with neat black hair, stands in the middle-center midground, reaching for the detective across the desk, while the phone screen of P01 is held prominently in the lower-center foreground. She faces the camera, aggressively leaning her upper body forward over the workstation surface, her eyes fixed sharply downward on the seated detective. A Korean male detective is seated behind the desk, his head and shoulder appearing at the near edge as a compressed, blurred silhouette from an over-the-shoulder perspective. The lit front face of P01 is angled toward the camera, casting a small hard glow onto C02O02's face. The desk between them acts as a physical boundary, scattered with paperwork. The room is bathed in harsh, greenish fluorescent light from above that carves dark shadows, with a worn bulletin board visible on the background wall behind her, anchoring the cold and metallic late-night atmosphere.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "김형사의 책상 위로 휴대전화 액정을 내민 채 단호한 눈빛으로 내려다보는 혜수의 정면", "사무실 안에는 김형사 혼자 당직 근무 중이다" ], "visual_inferences": [ "fixed_element hye_su_desk_stance: aggressively leaning her upper body forward over the surface", "fixed_element detective_kim_seated_posture: head and shoulder appear in the near foreground as a compressed, blurred silhouette", "fixed_element police_office_desk_environment: harsh, greenish fluorescent overhead lighting that casts hard shadows" ], "creative_decisions": [ "Placing P01 prominently in the lower-center foreground to establish the narrative focal point of the evidence.", "Using the blurred silhouette of the detective's shoulder to create a deep, over-the-shoulder spatial tension across the physical barrier of the desk." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bulletin board", "usage_kind": "anchor", "source_phrase": "worn bulletin board visible on the background wall" }, { "owned_token": "computer monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "anchor", "source_phrase": "reaching for the detective across the desk" }, { "owned_token": "file cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "anchor", "source_phrase": "greenish fluorescent light from above" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "office phone", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "paperwork", "usage_kind": "anchor", "source_phrase": "scattered with paperwork" }, { "owned_token": "partition", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "task chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "anchor", "source_phrase": "over the workstation surface" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "c56b8de7553aacfb", "owned_hash": "9a6567776e0fc0ca", "camera_direction_hash": "6c4755fc8c664f69", "owned_usage_hash": "e5e3a217ab5316d4", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bulletin board", "violating_phrase": "worn bulletin board visible on the background wall behind her", "reason": "The bulletin board is mentioned only to anchor the background setting and is not instructed to be newly generated, which aligns with the producer's 'anchor' classification.", "verdict": "anchor_reference" }, { "owned_object": "desk", "violating_phrase": "reaching for the detective across the desk", "reason": "The desk is referenced solely as a spatial boundary and anchor point between characters, verifying the producer's 'anchor' declaration.", "verdict": "anchor_reference" }, { "owned_object": "fluorescent light", "violating_phrase": "harsh, greenish fluorescent light from above", "reason": "The fluorescent light is described to set the overhead atmospheric illumination and shadows, making it a valid anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "paperwork", "violating_phrase": "scattered with paperwork", "reason": "The paperwork is referenced as static background details cluttered on the desk to enhance the scene's atmosphere, consistent with 'anchor'.", "verdict": "anchor_reference" }, { "owned_object": "workstation", "violating_phrase": "over the workstation surface", "reason": "The workstation is used purely as a positional surface reference for the character's body leaning forward, matching the producer's 'anchor' declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "N/A", "_shot_index": 3, "visible_entities": [ "C02", "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": "From 김형사's side of the desk, just above tabletop height and three feet back, the camera dollies in frontally toward 혜수 across the cluttered work surface. Her phone screen is held prominently in the lower foreground, occupying less than a third of the frame, while her sharp eyes remain in focus above it and 김형사 is present as a blurred seated shoulder at the near edge.", "lighting_mood": "Greenish fluorescent office light drains the color from the room, while the phone screen adds a small hard glow under 혜수's urgent eyes.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "reaches_for", "gesture_target_label": "detective across desk", "label": "혜수", "screen_zone": "middle_center", "target_id": "C02", "target_kind": "character" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "phone screen foreground", "screen_zone": "lower_center", "target_id": "P01", "target_kind": "prop" } ] }, "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", "P01" ], "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": "background_ref_attached", "bg_id": "L13B02", "reference_usage": "exact_background", "owned_objects": [ "bulletin board", "computer monitor", "desk", "file cabinet", "fluorescent light", "keyboard", "office phone", "paperwork", "partition", "task chair", "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": "hye_su_desk_stance", "element_type": "character_state", "character_name": "혜수", "description": "A Korean woman stands at the opposite side of a police office desk, aggressively leaning her upper body forward over the surface. She maintains an intense downward gaze toward the seated person across from her, exerting a strong physical presence over the workstation.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "김형사의 책상 위로... 내려다보는 혜수의 정면", "상체를 바짝 숙인 혜수의 강경한 자세", "facing_camera" ], "visual_inferences": [ "Her overall upper body maintains a tilted, forward-leaning angle throughout the dialogue to emphasize her urgency." ], "creative_decisions": [ "Keeping her body constantly braced over the desk establishes the desk as a physical boundary she is trying to cross." ], "confidence": "high" }, { "element_id": "detective_kim_seated_posture", "element_type": "character_state", "character_name": "김형사", "description": "A Korean male detective is seated behind his office desk. His head and shoulder appear in the near foreground as a compressed, blurred silhouette, framed from an over-the-shoulder perspective while he looks back up at the standing woman.", "applies_to_shots": [ 3, 5 ], "element_scope": "close", "source_facts": [ "사무실 안에는 김형사 혼자 당직 근무 중이다", "over_shoulder, seated behind desk", "blurred seated shoulder at the near edge" ], "visual_inferences": [ "He remains firmly seated in his duty chair, acting as a passive recipient to the woman's aggressive approach." ], "creative_decisions": [ "Using his blurred shoulder as a persistent foreground element to heighten the confrontation and depth of the shots." ], "confidence": "high" }, { "element_id": "police_office_desk_environment", "element_type": "environment_state", "character_name": "", "description": "A cluttered police station desk sits as a physical barrier in the middle of the frame. The room is bathed in harsh, greenish fluorescent overhead lighting that casts hard shadows, creating a cold and metallic late-night office atmosphere.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Greenish fluorescent office light drains the color from the room", "cluttered work surface", "Harsh fluorescent light from above carves hard shadows", "turning the desk confrontation into a cold, metallic standoff" ], "visual_inferences": [ "The desk is scattered with standard police paperwork and station equipment typical of a night shift." ], "creative_decisions": [ "Enhancing the greenish tint of the lighting to reflect the exhaustion and grim mood of the late-night police station." ], "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": "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": "20997d06708e982b" }, { "scene_index": 16, "heading": "S#16. 경찰서 사무실 / N", "beat_title": "위치 추적을 강력히 요구하는 혜수", "representative_moment": "혜수가 두 주먹을 책상에 짚고 김형사를 향해 상체를 바짝 숙이며 강압적으로 내려다보는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low tabletop observer viewpoint with harsh greenish fluorescent lighting", "t2i_prompt": "Photorealistic cinematic still. A low observer viewpoint at tabletop height. C02O02 in a navy safari jacket, a Korean woman in her 40s with neat black hair and sharp eyes, facing the camera, upper body leaning aggressively forward over a shared police office desk. Her clenched fists planted on the scuffed desktop in the lower foreground among disturbed paperwork, her face bearing down from the upper-center frame, eyes fixed downward on the seated person. At the near foreground edge, a Korean male detective's blurred head and shoulder appearing as a compressed silhouette in an over-the-shoulder perspective. The cluttered desk positioned between them as a physical boundary. [: A flat partition face behind her forming a static, overlit backdrop showing dull office papers]. Harsh greenish fluorescent overhead light carving hard shadows under her fists and brow, the scene steeped in a cold metallic late-night standoff atmosphere.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "두 주먹을 책상에 짚고 김형사를 향해 상체를 바짝 숙인 혜수의 강경한 자세", "사무실 안에는 김형사 혼자 당직 근무 중이다", "Harsh fluorescent light from above carves hard shadows under her fists and brow" ], "visual_inferences": [ "detective Kim framed as a compressed, blurred silhouette at the near foreground edge acting as a passive recipient", "low camera at tabletop height emphasizing Hye-su's fists and her imposing downward gaze" ], "creative_decisions": [ "greenish tint to reflect the exhaustion and cold, metallic mood of the late-night police station", "using the desk as a shared spatial anchor and physical pressure plate to heighten the confrontation" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "bulletin board", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "computer monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "file cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "office phone", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "paperwork", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "partition", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "task chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "56b9e1fee95151a5", "owned_hash": "9a6567776e0fc0ca", "camera_direction_hash": "f9a89d31e0631571", "owned_usage_hash": "fb9d5a6e60182047", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 16, "dependency_reason": "같은 공간(경찰서 사무실)의 직전 샷(Shot 3) 배경 분위기와 연속성 유지", "_shot_index": 5, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 16, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Still on 김형사's side of the desk, the lens drops to very low tabletop height and creeps closer, letting 혜수's clenched fists lead the frame before the eye rises to her leaning upper body. Her hands sit in the lower foreground, her face bears down from the center, and 김형사 remains a compressed foreground edge, making the desk feel like a pressure plate between them.", "lighting_mood": "Harsh fluorescent light from above carves hard shadows under her fists and brow, turning the desk confrontation into a cold, metallic standoff.", "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": "L13B02", "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": "hye_su_desk_stance", "element_type": "character_state", "character_name": "혜수", "description": "A Korean woman stands at the opposite side of a police office desk, aggressively leaning her upper body forward over the surface. She maintains an intense downward gaze toward the seated person across from her, exerting a strong physical presence over the workstation.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "김형사의 책상 위로... 내려다보는 혜수의 정면", "상체를 바짝 숙인 혜수의 강경한 자세", "facing_camera" ], "visual_inferences": [ "Her overall upper body maintains a tilted, forward-leaning angle throughout the dialogue to emphasize her urgency." ], "creative_decisions": [ "Keeping her body constantly braced over the desk establishes the desk as a physical boundary she is trying to cross." ], "confidence": "high" }, { "element_id": "detective_kim_seated_posture", "element_type": "character_state", "character_name": "김형사", "description": "A Korean male detective is seated behind his office desk. His head and shoulder appear in the near foreground as a compressed, blurred silhouette, framed from an over-the-shoulder perspective while he looks back up at the standing woman.", "applies_to_shots": [ 3, 5 ], "element_scope": "close", "source_facts": [ "사무실 안에는 김형사 혼자 당직 근무 중이다", "over_shoulder, seated behind desk", "blurred seated shoulder at the near edge" ], "visual_inferences": [ "He remains firmly seated in his duty chair, acting as a passive recipient to the woman's aggressive approach." ], "creative_decisions": [ "Using his blurred shoulder as a persistent foreground element to heighten the confrontation and depth of the shots." ], "confidence": "high" }, { "element_id": "police_office_desk_environment", "element_type": "environment_state", "character_name": "", "description": "A cluttered police station desk sits as a physical barrier in the middle of the frame. The room is bathed in harsh, greenish fluorescent overhead lighting that casts hard shadows, creating a cold and metallic late-night office atmosphere.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Greenish fluorescent office light drains the color from the room", "cluttered work surface", "Harsh fluorescent light from above carves hard shadows", "turning the desk confrontation into a cold, metallic standoff" ], "visual_inferences": [ "The desk is scattered with standard police paperwork and station equipment typical of a night shift." ], "creative_decisions": [ "Enhancing the greenish tint of the lighting to reflect the exhaustion and grim mood of the late-night police station." ], "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": "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": "9a501f11e3616238" }, { "scene_index": 17, "heading": "S#17. 옥탑방 / 동틀무렵", "beat_title": "불 켜진 옥탑방을 응시하는 고양이의 눈", "representative_moment": "A close-up of a black cat's eye reflecting the brightly lit rooftop window in the cold dawn.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Extreme close-up macro photography with shallow depth of field, cool blue-gray dawn color grading contrasting with a tiny warm reflection.", "t2i_prompt": "Photorealistic cinematic still. In an extreme close-up at eye level from the front-left, a black cat with short black fur, a sleek face, fine whiskers, and small triangular ears is seen at three-quarter angle from the left, head turned still and looking off-screen. The bright eye occupies most of the right-center frame, with only a thin edge of the dark muzzle visible for scale. Inside the glossy pupil, a small, hard warm reflection of a brightly lit window is caught. The cold blue-gray dawn ambient light leaves the fur sinking into ominous shadow, against [: a soft gray blur of a damp, cold dawn parapet wall running diagonally behind the cat's face, veiled in thin drifting mist].", "outfit_assignments": [], "source_facts": [ "S#17 action: '검은 고양이가 불이 켜져 있는 옥탑 방을 슬쩍 보더니'", "shot description: '환하게 불이 켜진 낡은 옥탑방 창문을 향해 고개를 돌린 채 멈춰 있는 검은 고양이의 눈동자'" ], "visual_inferences": [ "Macro framing isolates the eye, making the window reflection the focal narrative point", "The background wall is completely out of focus due to the extreme close-up scale" ], "creative_decisions": [ "Used a small, hard warm reflection inside the glossy pupil to indicate the lit window without rendering the full background room", "Contrasted the cold blue-gray ambient dawn light with the warm window reflection to enhance the ominous atmosphere" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [], "owned_object_usage": [ { "owned_token": "drying platform", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "laundry line", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "metal door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "parapet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pipe", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "rooftop room", "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": "a2d060fbe54bebe4", "owned_hash": "d76ce15da6b4f174", "camera_direction_hash": "9c1a6fefbde5b58a", "owned_usage_hash": "ec665c37ebb8c53d", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 4, "dependency_reason": "동일한 옥탑방 외부 공간 및 새벽 분위기 (단, close framing으로 배경 이미지는 직접 참조하지 않음)", "_shot_index": 2, "visible_entities": [ "C12" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 17, "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 cat-eye height from the cat’s front-left side, the camera tightens into an extreme close-up so the black cat’s eye occupies most of the right-center frame, with only a thin edge of fur and muzzle visible for scale. The bright old rooftop room window is not shown directly; it appears as a small, hard reflection caught inside the glossy pupil while the dawn parapet falls into soft gray blur behind it.", "lighting_mood": "Cold blue-gray dawn is pierced by a tiny warm window reflection, making the cat’s black fur sink into ominous shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C12" ], "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": "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": [], "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": [], "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": "0c55ee9f44152c11" }, { "scene_index": 18, "heading": "S#18. 안방 안 / D", "beat_title": "밤샘의 피로와 선잠", "representative_moment": "침대 모서리에 머리를 기댄 채 지친 얼굴로 눈을 감고 웅크려 있는 C01.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Close bedside profile shot looking slightly downward from chest height, sickly gray-beige morning light", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray casual windbreaker, a young Korean woman, is positioned in the middle-left midground in right profile, her curled upper body filling the frame as her cheek nearly touches the bed edge. Her long straight black hair is slightly unkempt from restless sleep, falling over her visibly exhausted face with eyes tightly closed. The camera, positioned at chest height looking slightly downward, compresses the mattress edge that angles toward the lens into the lower frame, its side seam and creased, pale bedding fabric physically supporting her head. [L05: A plain, dim bedroom wall recedes into soft focus behind her, leaving only a narrow vertical strip of stale room space on the far right]. Weak morning light mixes with heavy interior shadows, casting a sickly gray-beige tone across her pale skin and the rumpled sheets.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "침대에 머리를 대고 앉아있는 수리영 (Suriyoung sitting with her head on the bed)", "밤새 엄마를 기다리다 깜박 졸았다 (dozed off waiting for her mom all night)", "눈을 질끈 감은 채 (eyes tightly closed)", "profile_right", "gaze=closed_eyes" ], "visual_inferences": [ "She wears her casual daytime windbreaker as she fell asleep waiting without changing", "Her hair is messy from sleeping uncomfortably against the bed edge", "The flat daylight inside the apartment is gloomy and lacks warmth, reflecting her anxiety" ], "creative_decisions": [ "Tight close framing on her curled profile to isolate her physical and emotional exhaustion", "Low camera angle looking slightly downward from chest height to emphasize her collapsed posture", "Sickly gray-beige color grading to mute the morning light, bridging the mundane apartment with the unfolding dread" ], "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": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bedding", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mattress", "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": "49a6b8354c41b22e", "owned_hash": "792fd7b55fc7cece", "camera_direction_hash": "ef4df90e2a9b3b21", "owned_usage_hash": "d5de17d516306cf1", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "동일한 옥탑방 내부의 우울하고 회갈색 빛이 도는 조명/톤 일관성 유지 (background reference 자체는 close framing으로 생략되나, 조명 분위기 계승)", "_shot_index": 1, "visible_entities": [ "C01", "L05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the low bedside profile position, the frame is tightened so 수리영’s curled side profile and closed eyes fill the left and center of the image, her cheek nearly touching the bed edge. The camera looks slightly downward from chest height, compressing the bed corner behind her and leaving only a narrow strip of dim room space on the far right.", "lighting_mood": "Weak morning light mixes with stale interior shadow, keeping the palette gray-beige and exhausted rather than warm.", "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": "수리영 curled at bed edge", "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", "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": "skipped_close_framing", "bg_id": "L05B03", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "suriyoung_closed_eyes_face", "element_type": "character_state", "character_name": "Suriyoung", "description": "A young Korean woman wearing casual indoor clothing, as she has just woken up. Her hair is slightly unkempt from sleeping and washing her face. Her eyes are tightly closed, and her face clearly carries a heavy expression of emotional distress and physical exhaustion.", "applies_to_shots": [ 1, 15 ], "element_scope": "close", "source_facts": [ "밤새 엄마를 기다리다 깜박 졸았다", "눈을 질끈 감은 채", "gaze=closed_eyes" ], "visual_inferences": [ "She remains in casual indoor sleepwear since she has not changed clothes after waking up", "Slightly messy hair resulting from her restless sleep and splashing water on her face" ], "creative_decisions": [ "Using her tightly closed eyes as a recurring focal point to emphasize her attempt to cope with psychological horror and exhaustion" ], "confidence": "medium" }, { "element_id": "gloomy_apartment_daylight", "element_type": "environment_state", "character_name": "", "description": "The interior of the apartment is illuminated by a weak, flat daytime light that is heavily mixed with stale interior shadows. The overall lighting palette remains sickly and gray-beige, preventing the environment from feeling warm or inviting.", "applies_to_shots": [ 1, 13, 15 ], "element_scope": "full", "source_facts": [ "안방 안 / D", "Weak morning light mixes with stale interior shadow", "Neutral apartment daylight turns sickly and gray", "Flat daylight is darkened by interior shadows" ], "visual_inferences": [ "The ambient daylight entering through the windows is subdued, casting deep and dull shadows across different rooms (bedroom, dining area)" ], "creative_decisions": [ "Maintaining a consistent flat and gray lighting tone across all shots to visually bridge the mundane apartment setting with the unfolding supernatural horror" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 8, "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": "56fa0478874b6e01" }, { "scene_index": 18, "heading": "S#18. 안방 안 외 / D", "beat_title": "사진 속 기괴하게 일그러지는 엄마의 얼굴", "representative_moment": "식탁 위 P06 액자 속 환하게 웃고 있던 중년 여성의 얼굴이 기괴하게 아래로 뭉개져 흘러내린 채 고정된 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Close perpendicular framing at table height, flattening the image with sickly gray-beige daytime lighting", "t2i_prompt": "Photorealistic cinematic still. A close framing squarely at table height, positioned almost perpendicularly over P06 lying flat on a clean, faintly reflective dining tabletop. Inside the ordinary, rigid rectangular borders of P06, the printed photograph of an Asian middle-aged woman facing the camera is grotesquely distorted, her brightly smiling facial features flattened and smeared downward in a fixed, wet-looking shadow tone. The surrounding lighting consists of a weak, flat daylight that turns into a sickly gray-beige palette immediately around the photograph. The intact physical frame edge and [L05B04: the smooth tabletop surface extending to the frame borders] remain perfectly still and realistic, grounding the hallucinated distortion.", "outfit_assignments": [], "source_facts": [ "수리영이 식탁에 놓인 사진을 보는데", "환하게 웃는 엄마 얼굴이 뭉개져 흘러내린다", "Weak morning light mixes with stale interior shadow" ], "visual_inferences": [ "The camera acts as a direct observer positioned squarely above the photo", "The distortion is physically constrained within the boundaries of the photo frame", "The surrounding tabletop is rendered as an unaffected, stable surface to contrast with the hallucination" ], "creative_decisions": [ "Framed perpendicularly to emphasize the flat, evidence-like presentation of the hallucination", "Applied a sickly gray-beige color tone to bridge the mundane apartment setting with the unfolding supernatural horror", "Used a wet-looking shadow tone specifically on the smeared face to enhance the grotesque, downward-dragged texture" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "prop" ], "owned_object_usage": [ { "owned_token": "TV", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "photo frame", "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": "2d5380de56fb7d5d", "owned_hash": "92ac09612a8b6430", "camera_direction_hash": "940e74f75fc19d38", "owned_usage_hash": "ecb3831551d1aa03", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 5, "dependency_reason": "previous_shot_refs indicates a locational connection to Scene 5 Shot 1, establishing the base apartment environment tone before applying the skipped_close_framing background logic.", "_shot_index": 13, "visible_entities": [ "P06" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 18, "shot_index": 13 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera abandons the previous over-shoulder position and sits squarely at table height, almost perpendicular to the framed photo surface, flattening the image like evidence pinned under glass. 민숙’s smiling face is fixed in the center of the frame as the hallucinated features smear downward, while the frame edge and tabletop remain rigid and real around the distortion.", "lighting_mood": "Neutral apartment daylight turns sickly and gray around the photo, with the distorted face carrying a subtly dark, wet-looking shadow 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": 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": [ "P06" ], "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": "L05B04", "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": "gloomy_apartment_daylight", "element_type": "environment_state", "character_name": "", "description": "The interior of the apartment is illuminated by a weak, flat daytime light that is heavily mixed with stale interior shadows. The overall lighting palette remains sickly and gray-beige, preventing the environment from feeling warm or inviting.", "applies_to_shots": [ 1, 13, 15 ], "element_scope": "full", "source_facts": [ "안방 안 / D", "Weak morning light mixes with stale interior shadow", "Neutral apartment daylight turns sickly and gray", "Flat daylight is darkened by interior shadows" ], "visual_inferences": [ "The ambient daylight entering through the windows is subdued, casting deep and dull shadows across different rooms (bedroom, dining area)" ], "creative_decisions": [ "Maintaining a consistent flat and gray lighting tone across all shots to visually bridge the mundane apartment setting with the unfolding supernatural horror" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 5, "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": "prop", "id": "P06", "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": [ 13 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P06", "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": "1ec9c3087a355621" }, { "scene_index": 18, "heading": "S#18. 안방 안 / D", "beat_title": "환영에 시달리며 자신의 뺨을 치는 수리영", "representative_moment": "C01이 눈을 질끈 감고 자신의 뺨을 강하게 때리는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Tight eye-level close-up, flat interior daylight", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, a young Korean woman, occupies the middle-center foreground, facing the camera with eyes tightly closed. Her palm strikes one cheek from the right side of the frame in an exact mid-impact moment, her long unkempt black hair shifting slightly. Her face shows a heavy expression of emotional distress and physical exhaustion, with a hard red flush forming at the sharp skin contact point beneath her pressing hand. [L05B03: a heavily blurred flat interior wall and a soft out-of-focus dining table edge behind her shoulder, illuminated by a sickly gray-beige flat daylight darkened by stale interior shadows].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "shot description: her palm striking one cheek from frame right, exact mid-impact moment", "action: slapping her cheek to wake up from hallucination", "gaze=closed_eyes, state=alive" ], "visual_inferences": [ "fixed_element suriyoung_closed_eyes_face: unkempt hair from restless sleep, heavy emotional distress", "fixed_element gloomy_apartment_daylight: sickly gray-beige palette, stale interior shadows" ], "creative_decisions": [ "Tight center framing isolates the subject's face to maximize the psychological tension and physical impact", "The dining table and wall are heavily blurred to serve as a soft spatial reminder without distracting from the sharp skin contact point" ], "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": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bedding", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "dining table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor lamp", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "mattress", "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": "d5b97e71d4d21961", "owned_hash": "792fd7b55fc7cece", "camera_direction_hash": "414b09520a700349", "owned_usage_hash": "d5de17d516306cf1", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "같은 아파트 내부의 약하고 흐릿한 실내 채광 톤(gloomy_apartment_daylight) 유지", "_shot_index": 15, "visible_entities": [ "C01" ], "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": "Locked at tight eye level directly in front of 수리영, the frame holds still at the exact mid-impact moment: her face fills the center, her palm striking one cheek from frame right, with the dining area crushed into blur behind her. The composition refuses shake or release, making the skin contact point the sharpest visual weight in the shot.", "lighting_mood": "Flat daylight is darkened by interior shadows, leaving a hard red flush on the cheek as the only sharp 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": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영 face at slap impact", "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": "L05B03", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "suriyoung_closed_eyes_face", "element_type": "character_state", "character_name": "Suriyoung", "description": "A young Korean woman wearing casual indoor clothing, as she has just woken up. Her hair is slightly unkempt from sleeping and washing her face. Her eyes are tightly closed, and her face clearly carries a heavy expression of emotional distress and physical exhaustion.", "applies_to_shots": [ 1, 15 ], "element_scope": "close", "source_facts": [ "밤새 엄마를 기다리다 깜박 졸았다", "눈을 질끈 감은 채", "gaze=closed_eyes" ], "visual_inferences": [ "She remains in casual indoor sleepwear since she has not changed clothes after waking up", "Slightly messy hair resulting from her restless sleep and splashing water on her face" ], "creative_decisions": [ "Using her tightly closed eyes as a recurring focal point to emphasize her attempt to cope with psychological horror and exhaustion" ], "confidence": "medium" }, { "element_id": "gloomy_apartment_daylight", "element_type": "environment_state", "character_name": "", "description": "The interior of the apartment is illuminated by a weak, flat daytime light that is heavily mixed with stale interior shadows. The overall lighting palette remains sickly and gray-beige, preventing the environment from feeling warm or inviting.", "applies_to_shots": [ 1, 13, 15 ], "element_scope": "full", "source_facts": [ "안방 안 / D", "Weak morning light mixes with stale interior shadow", "Neutral apartment daylight turns sickly and gray", "Flat daylight is darkened by interior shadows" ], "visual_inferences": [ "The ambient daylight entering through the windows is subdued, casting deep and dull shadows across different rooms (bedroom, dining area)" ], "creative_decisions": [ "Maintaining a consistent flat and gray lighting tone across all shots to visually bridge the mundane apartment setting with the unfolding supernatural horror" ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "shot_index": 8, "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": "0d9a01bd066e081c" }, { "scene_index": 19, "heading": "S#19. 마트 / D", "beat_title": "마트 외부 CCTV 확인", "representative_moment": "마트 밖, 건물 외벽에 매달린 낡은 CCTV 카메라를 손가락으로 가리킨 채 멈춘 수리영의 측면.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Observer POV, shoulder-height medium framing with cool gray early morning light and a subtle coastal haze", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, an East Asian woman in her 20s with long straight black hair, stands in the middle-left foreground in right profile, her frozen finger pointing upward toward the old wall-mounted CCTV camera in the upper-right background. Her dark eyes are fixed off-screen. Pale morning light washes the storefront in muted gray, softened by a subtle coastal haze. The exterior wall rises steeply behind her, providing a quiet background, while the old CCTV camera casts a small shadow under its bracket, sitting visibly attached to the surface.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#19 action: '엄마와 남자 직원CCTV 위치를 확인하다.'", "shot description: 마트 밖, 건물 외벽에 매달린 낡은 CCTV 카메라를 손가락으로 가리킨 채 멈춘 수리영의 측면.", "DP instruction: 'Pale morning light washes the storefront in muted gray'" ], "visual_inferences": [ "fixed_element morning_supermarket_exterior: pale, cool gray daylight softened by coastal haze", "C01's pointing gesture directs the visual flow diagonally upward from the lower-left to the upper-right CCTV camera" ], "creative_decisions": [ "Placed C01 strictly in the middle-left foreground with right profile to satisfy the frame spatial contract", "Kept the color palette muted and cool to emphasize the serious investigative mood of the early morning" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "CCTV camera", "usage_kind": "anchor", "source_phrase": "the old wall-mounted CCTV camera" }, { "owned_token": "bracket", "usage_kind": "anchor", "source_phrase": "under its bracket" }, { "owned_token": "conduit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "display window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "glass sliding door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sidewalk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storefront", "usage_kind": "anchor", "source_phrase": "washes the storefront in muted gray" }, { "owned_token": "utility box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "The exterior wall rises steeply behind her" }, { "owned_token": "window decal", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "16fd8a7fa1cd898f", "owned_hash": "767b682db43e5ead", "camera_direction_hash": "28fa2412dbb814a9", "owned_usage_hash": "55536d20476f61a5", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "CCTV camera", "violating_phrase": "pointing upward toward the old wall-mounted CCTV camera", "reason": "The CCTV camera is used as a spatial anchor for the character's gesture and shadow description, with no instruction to redraw or alter it.", "verdict": "anchor_reference" }, { "owned_object": "bracket", "violating_phrase": "under its bracket", "reason": "The bracket is mentioned purely to specify the shadow placement of the existing CCTV camera, serving as an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "storefront", "violating_phrase": "washes the storefront in muted gray", "reason": "The storefront is mentioned as receiving ambient morning light, acting as a background reference with no redraw instructions.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "The exterior wall rises steeply behind her", "reason": "The exterior wall is described as rising behind the character to provide a quiet background, acting strictly as a spatial anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 8, "dependency_reason": "마트 외부의 배경적 연속성 및 동일한 이른 아침 시간대 무드(cool gray daylight) 유지", "_shot_index": 4, "visible_entities": [ "C01", "L08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 19, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Outside the supermarket, the camera settles after the upward tilt at shoulder height beside 수리영, keeping her in clean side profile along the lower-left frame while her frozen pointing finger leads diagonally toward the old CCTV camera high on the exterior wall. The wall rises steeply behind her, and the CCTV sits in the upper-right background as a small but readable object rather than an oversized insert.", "lighting_mood": "Pale morning light washes the storefront in muted gray, while the old CCTV casts a small, suspicious shadow under its bracket.", "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": "old wall-mounted CCTV camera", "screen_zone": "upper_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "points_to", "gesture_target_label": "old wall-mounted CCTV camera", "label": "수리영 pointing beside wall", "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", "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": "background_ref_attached", "bg_id": "L08B03", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "bracket", "conduit", "display window", "glass sliding door", "sidewalk", "storefront", "utility box", "wall", "window decal" ], "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": "morning_supermarket_exterior", "element_type": "environment_state", "character_name": "", "description": "The exterior of the supermarket during the early morning. The area is bathed in a pale, cool gray daylight that is softened by a subtle coastal haze. The exterior walls and storefront provide a muted, quiet background without any overly bright or distracting elements.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "Pale morning light washes the storefront in muted gray", "exterior wall", "Cool daylight is softened by coastal haze", "supermarket exterior softens behind her" ], "visual_inferences": [ "The building's exterior is plain, typical of a local mart, and slightly weathered by the coastal environment" ], "creative_decisions": [ "Maintain a desaturated, cool color palette for the environment to emphasize the character's serious mood and the tense early morning atmosphere" ], "confidence": "high" } ], "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": "ab134e716f9c69d7" }, { "scene_index": 19, "heading": "S#19. 마트 / D", "beat_title": "증거 확보를 위한 전화", "representative_moment": "마트 밖에 서서 심각한 표정으로 어딘가 전화를 거는 수리영.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "tight medium shot from in front, upper body dominant, cool daylight softened by coastal haze", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray casual windbreaker, an Asian woman in her early 20s with long straight black hair and clear dark eyes, stands at three-quarter angle from the left, her upper body dominating the right half of the frame. She is pressing P01 tight against her ear, gazing into the distance across a strip of empty outdoor space on the left side of the frame with a tense expression. The scene is bathed in cool pale daylight softened by coastal haze, leaving her face shadowed with only faint warmth catching on her phone-side cheek. The muted supermarket storefront and exterior wall recede behind her shoulder, the out-of-focus display window providing a quiet background.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 혜수에게 전화 걸어 상황을 설명한다", "휴대전화를 귀에 바싹 댄 채 심각한 표정으로 허공을 응시하는 수리영의 상체" ], "visual_inferences": [ "fixed_element morning_supermarket_exterior: The exterior of the supermarket during the early morning. The area is bathed in a pale, cool gray daylight that is softened by a subtle coastal haze" ], "creative_decisions": [ "placed the character on the right side of the frame to leave negative space on the left, emphasizing her distant gaze and anxious mood", "kept the background surfaces out of focus to draw attention to her tense facial expression", "applied faint warmth on one cheek to contrast with the overall cool gray atmospheric tone" ], "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": "bracket", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "conduit", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "display window", "usage_kind": "anchor", "source_phrase": "the out-of-focus display window providing a quiet background" }, { "owned_token": "glass sliding door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "sidewalk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storefront", "usage_kind": "anchor", "source_phrase": "The muted supermarket storefront" }, { "owned_token": "utility box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "anchor", "source_phrase": "and exterior wall recede behind her shoulder" }, { "owned_token": "window decal", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "e9bb0e2411d93aa8", "owned_hash": "767b682db43e5ead", "camera_direction_hash": "433e1a7c23ff24e3", "owned_usage_hash": "b5ab0daf1313f505", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "display window", "violating_phrase": "the out-of-focus display window providing a quiet background", "reason": "The display window is described as an out-of-focus background element, aligning with the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "storefront", "violating_phrase": "The muted supermarket storefront", "reason": "The storefront is used as a muted background element receding behind the character's shoulder, serving as a spatial anchor.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "exterior wall recede behind her shoulder", "reason": "The exterior wall is mentioned as receding behind her shoulder to establish depth, which represents an anchor usage.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 8, "dependency_reason": "exact_background preservation from the prior scene's supermarket exterior", "_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": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera has lowered and moved in to a tight medium shot from in front of 수리영, keeping her upper body dominant while leaving a strip of empty outdoor space in the direction of her fixed stare. Her phone presses hard against her ear near frame right, and the supermarket exterior softens behind her without competing with her serious expression.", "lighting_mood": "Cool daylight is softened by coastal haze, keeping her face gray and tense with only faint warmth on the phone-side cheek.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "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": "background_ref_attached", "bg_id": "L08B03", "reference_usage": "exact_background", "owned_objects": [ "CCTV camera", "bracket", "conduit", "display window", "glass sliding door", "sidewalk", "storefront", "utility box", "wall", "window decal" ], "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": "morning_supermarket_exterior", "element_type": "environment_state", "character_name": "", "description": "The exterior of the supermarket during the early morning. The area is bathed in a pale, cool gray daylight that is softened by a subtle coastal haze. The exterior walls and storefront provide a muted, quiet background without any overly bright or distracting elements.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "Pale morning light washes the storefront in muted gray", "exterior wall", "Cool daylight is softened by coastal haze", "supermarket exterior softens behind her" ], "visual_inferences": [ "The building's exterior is plain, typical of a local mart, and slightly weathered by the coastal environment" ], "creative_decisions": [ "Maintain a desaturated, cool color palette for the environment to emphasize the character's serious mood and the tense early morning atmosphere" ], "confidence": "high" } ], "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": "7f4013a230069541" }, { "scene_index": 20, "heading": "S#20. 마트 총무과 / 실내", "beat_title": "식별 불가능한 화면 앞의 고조되는 불안감", "representative_moment": "C02가 미간을 깊게 찌푸린 채 모니터의 차가운 푸른 불빛을 받으며 오프스크린 화면을 노려보는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level tight close-up, stark contrast with cold blue monitor spill and deep oily shadows", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a navy safari jacket, an East Asian woman in her 40s with neat black hair and distinct features, her concentrated face occupying the middle-center foreground. She is positioned at a three-quarter angle from the right, leaning toward the unseen screen, looking off-screen with a deeply furrowed brow. The off-screen monitor casts a harsh, cold blue spill light from the front, sharply catching her brow and eye sockets while the rest of her face sinks into deep oily shadow. The immediate [: cramped security office] behind her is enveloped in heavy gray shadows, creating indistinct dark blocks that isolate her brightly lit features.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "Shot 4: 혜수의 얼굴 클로즈업, 화면을 보며 미간을 잔뜩 찌푸린 채 멈춤", "Cold monitor spill carves her brow and eye sockets while the room stays in deep gray shadow" ], "visual_inferences": [ "fixed_element security_room_lighting: deep heavy shadows, cold blue monitor spill acting as the primary illumination", "The background is underlit and soft, isolating her face without revealing specific architectural details due to close framing" ], "creative_decisions": [ "Used tight close-up framing to push the monitor completely off-screen, focusing purely on the psychological tension in her expression", "Avoided physical cutting verbs for lighting to prevent T2I misinterpretation, using 'catching her brow' instead", "Replaced forbidden L01 reference with L15 as the valid background frame-edge surface tag" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "USB hub", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "computer", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "document tray", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "file box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "storage shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "a51326e4d4aeb9e5", "owned_hash": "b49d5184eb0320e6", "camera_direction_hash": "2d7e76f8acd0faeb", "owned_usage_hash": "cc6cd93c283bf159", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 4, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 20, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From beside the monitor line at eye level, the dolly-in ends tighter than expected until the monitor disappears entirely, leaving only 혜수’s face and furrowed brow filling the center-left frame. She is angled toward the unseen screen just off camera, so her concentration reads as a profile-leaning three-quarter close-up rather than a direct address.", "lighting_mood": "Cold monitor spill carves her brow and eye sockets while the room stays in deep gray shadow, emphasizing frustrated scrutiny.", "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": "혜수 concentrated face", "screen_zone": "middle_center", "target_id": "C02", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02" ], "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": "L15B01", "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": "security_room_lighting", "element_type": "environment_state", "character_name": "", "description": "The surrounding room is enveloped in deep, heavy shadows, creating a dark and tense atmosphere. A harsh, cold blue light spilling from an off-screen computer monitor acts as the primary illumination, sharply catching the edges of subjects and objects in the immediate foreground against the dark background.", "applies_to_shots": [ 4, 7 ], "element_scope": "full", "source_facts": [ "Cold monitor spill carves her brow and eye sockets while the room stays in deep gray shadow (Shot 4)", "A cold monitor-blue edge light catches the USB and fingertips, while the rest of the desk sinks into oily shadow (Shot 7)" ], "visual_inferences": [ "The main overhead lights in the room are either turned off or extremely dim, isolating the bright, cool light coming from the CCTV monitor." ], "creative_decisions": [ "Use the stark contrast between the oily, deep shadows and the cold blue monitor spill to emphasize the character's focus and psychological 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": "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": "5824db462121ffc7" }, { "scene_index": 20, "heading": "S#20. 마트 총무과 / 실내", "beat_title": "USB 메모리를 꽉 쥐는 혜수", "representative_moment": "혜수의 손가락이 USB 메모리를 강하게 쥐고 있는 클로즈업 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Macro-like focus from a low desk height, intense cold monitor-blue edge light against oily deep shadows", "t2i_prompt": "Photorealistic cinematic still. C02 in a navy safari jacket, a Korean woman in her 40s, her hand occupying the lower-center frame, no faces visible. Her fingers clamp a small USB memory tightly, physical pressure showing in her knuckles, the USB's narrow side and metal connector edge angled toward the camera between her fingers. The surrounding room is enveloped in deep, heavy shadows, creating a dark and tense atmosphere. A harsh, cold monitor-blue edge light spills from an off-screen source, sharply catching the USB and fingertips, while the scuffed table surface sinks into oily shadow. In the rear plane, a blurred monitor and computer equipment sit out of focus as soft dark rectangular forms against the dark background.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "Shot description: USB 메모리를 손가락 사이에 꽉 쥔 혜수의 손 클로즈업.", "DP instruction: The USB remains small and realistically scaled against her hand and the desk surface, while the pressure in her knuckles carries the tension." ], "visual_inferences": [ "fixed_element security_room_lighting: The surrounding room is enveloped in deep, heavy shadows with a harsh, cold blue light catching the edges of foreground subjects.", "To maintain the 'insert' framing scale requested, the primary focus is entirely on the hand and the USB, keeping the face completely out of the frame." ], "creative_decisions": [ "Position the hand in the lower-center frame to ground the tension.", "Render the background computer equipment as soft, blurred dark rectangular forms to anchor the spatial depth without distracting from the USB." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "USB hub", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "computer", "usage_kind": "anchor", "source_phrase": "a blurred monitor and computer equipment sit out of focus" }, { "owned_token": "document tray", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "file box", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "fluorescent light", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "monitor", "usage_kind": "anchor", "source_phrase": "a blurred monitor and computer equipment sit out of focus" }, { "owned_token": "storage shelf", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "table", "usage_kind": "anchor", "source_phrase": "the scuffed table surface sinks into oily shadow" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "workstation", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "f2d4ad56a3f51940", "owned_hash": "b49d5184eb0320e6", "camera_direction_hash": "d6534765675d80e2", "owned_usage_hash": "f27b5bff4f14f112", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "computer", "violating_phrase": "a blurred monitor and computer equipment sit out of focus", "reason": "The computer is described as blurred and out of focus in the background to set the atmosphere, which correctly serves as a background anchor matching the producer's echo.", "verdict": "anchor_reference" }, { "owned_object": "monitor", "violating_phrase": "a blurred monitor and computer equipment sit out of focus", "reason": "The monitor is mentioned as an out-of-focus background element, which serves as a positional anchor and matches the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "table", "violating_phrase": "the scuffed table surface sinks into oily shadow", "reason": "The table surface is mentioned in the context of lighting and shadow description, acting as a spatial anchor rather than demanding a redesign of the table.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 7, "visible_entities": [ "C02" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 20, "shot_index": 7 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "At low desk height within inches of 혜수’s hand, the shot becomes a macro-like insert: her fingers clamp the USB memory in the lower-center frame, with no faces visible. The USB remains small and realistically scaled against her hand and the desk surface, while the pressure in her knuckles carries the tension.", "lighting_mood": "A cold monitor-blue edge light catches the USB and fingertips, while the rest of the desk sinks into oily shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C02" ], "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": [ { "subject_id": "C02", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "Only 혜수’s hand is visible in a macro insert; a full outlook reference would over-specify an identity that the frame cannot show." } ], "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": [ "USB hub", "chair", "computer", "document tray", "door", "file box", "fluorescent light", "keyboard", "monitor", "storage shelf", "table", "wall", "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": "security_room_lighting", "element_type": "environment_state", "character_name": "", "description": "The surrounding room is enveloped in deep, heavy shadows, creating a dark and tense atmosphere. A harsh, cold blue light spilling from an off-screen computer monitor acts as the primary illumination, sharply catching the edges of subjects and objects in the immediate foreground against the dark background.", "applies_to_shots": [ 4, 7 ], "element_scope": "full", "source_facts": [ "Cold monitor spill carves her brow and eye sockets while the room stays in deep gray shadow (Shot 4)", "A cold monitor-blue edge light catches the USB and fingertips, while the rest of the desk sinks into oily shadow (Shot 7)" ], "visual_inferences": [ "The main overhead lights in the room are either turned off or extremely dim, isolating the bright, cool light coming from the CCTV monitor." ], "creative_decisions": [ "Use the stark contrast between the oily, deep shadows and the cold blue monitor spill to emphasize the character's focus and psychological tension." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 20, "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", "id": "C02", "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": "de371b9aa247fb55" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "사진 속 참혹한 진실을 마주한 경악", "representative_moment": "사건 현장 사진을 보고 충격에 빠져 두 눈을 크게 뜬 C01의 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "seated-eye height composition emphasizing psychological shock with dirty greenish lighting", "t2i_prompt": "Photorealistic cinematic still. From a seated-eye height perspective near the desk corner, C01O01 in a gray windbreaker, a Korean woman in her early 20s with long straight black hair and a round face, appears at a three-quarter angle from the left, leaning over the surface. Her face is illuminated by harsh greenish fluorescent overhead lights that flatten her skin, looking down with widened dark eyes dominating the upper half of the frame. The desk corner cuts diagonally across the lower foreground, falling into cold, dirty gray shadows. On the desk, the printed front faces of enlarged photographs lie upward, pushed completely out of focus as indistinct pale and dark blurred shapes at the bottom edge. [: Soft, vacant office partitions recede behind her shoulders as dim, featureless blocks, maintaining a bleak atmosphere.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "사진 속 살점이 뜯겨나간 형상을 보고 두 눈을 크게 뜬 수리영의 얼굴 클로즈업", "Greenish office fluorescents flatten her skin while the desk foreground falls into dirty gray shadow", "their printed front faces lie upward on the desk but are pushed out of focus below frame" ], "visual_inferences": [ "fixed_element police_office_environment: harsh, greenish fluorescent overhead lights", "The lighting is positioned above, casting stark highlights and desaturating the colors of the desk area", "The disturbing photographs remain only as blurred pale shapes at the bottom edge to keep focus on the character's reaction" ], "creative_decisions": [ "Placed the face in the upper half of the frame and blurred the foreground elements to simulate a tight framing effect without using forbidden camera terminology", "Used L01 generic block for the frame-edge background as the background reference (L13) is omitted due to close framing constraints" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "computer monitor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "desk", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "office phone", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "paperwork", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "partition", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "task chair", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "6f485742353182b9", "owned_hash": "6afee18c21e3650a", "camera_direction_hash": "4704ae88c8d7cbe5", "owned_usage_hash": "5a43f5e8df51a6f6", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 16, "dependency_reason": "경찰서 사무실의 환경 조명(police_office_environment) 및 전체적인 톤을 앞선 경찰서 씬(S16)과 연속적으로 유지 (단, 이 샷은 close framing으로 인해 배경 디테일을 skip함)", "_shot_index": 5, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 5 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From seated-eye height near 혜수’s desk corner, the push-in completes as a close-up on 수리영’s face, her widened eyes dominating the upper half of the frame. The disturbing photographs remain only as blurred pale shapes at the bottom edge, keeping the shock on her reaction rather than on explicit detail.", "lighting_mood": "Greenish office fluorescents flatten her skin while the desk foreground falls into dirty gray shadow, making her eyes feel exposed and trapped.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "skipped_close_framing", "bg_id": "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": "police_office_environment", "element_type": "environment_state", "character_name": "", "description": "The environment around the police office desk is illuminated by harsh, greenish fluorescent overhead lights. The desk surface and surrounding areas fall into cold, dirty gray shadows, with generic, blurred office clutter visible in the background. This lighting setup maintains a bleak and tense atmosphere across the setting.", "applies_to_shots": [ 5, 8 ], "element_scope": "full", "source_facts": [ "Greenish office fluorescents flatten her skin while the desk foreground falls into dirty gray shadow", "Harsh fluorescent light glints on the pen tip and whitens the paper, while the surrounding desk stays gray and tense", "blurred desk clutter" ], "visual_inferences": [ "The lighting is positioned above, casting stark highlights and desaturating the colors of the desk area.", "The desk surface has a dull finish that absorbs the ambient gray shadows." ], "creative_decisions": [ "Maintain a consistent greenish-gray color palette in the environment to emphasize the gritty and uncomfortable mood of the police station." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 16, "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": "8bae258eee9fbe6e" }, { "scene_index": 21, "heading": "S#21. 경찰서 안 / D", "beat_title": "좌표 메모 기록", "representative_moment": "C01이 P07에 P08을 꽉 쥐고 해상 좌표를 황급히 메모하는 순간의 손과 종이 표면.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Very low at tabletop height, focusing tightly on the hand and paper, blurred background", "t2i_prompt": "Photorealistic cinematic still. C01 in a gray windbreaker sleeve, a young Korean woman, her wrist and fingers occupying the foreground, gripping P08 tightly. The barrel of P08 angles diagonally from the upper left, its tip pressed firmly down into the center of P07. P07 sits flat beneath the fingers on the desk, its top face visible to the camera with handwritten numbers forming under the pressure. The hard desk surface immediately below falls into cold, dirty gray shadows, illuminated by harsh greenish fluorescent overhead lights that glint on the pen tip and whiten the paper. Beyond the hand, the tense environment is reduced to soft, distant blur, showing out-of-focus paperwork and the faint glow of a computer monitor screen.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#21 action: '필기도구를 찾아 좌표를 메모 한다.'", "shot description: 한 손으로 볼펜을 꽉 쥔 채 작은 종이에 숫자를 쓰고 있는 순간, 펜촉이 종이를 누르고 있는 상태의 손 클로즈업." ], "visual_inferences": [ "fixed_element police_office_environment: The desk surface falls into cold, dirty gray shadows with generic, blurred office clutter visible in the background.", "The computer monitor with the location result is faintly glowing but kept completely out of focus to match the depth of field on the hand." ], "creative_decisions": [ "Extreme low tabletop angle to emphasize the physical pressure of the pen on the paper, anchoring the urgent action.", "Harsh fluorescent greenish highlights on the pen tip and whitened paper contrasting against dirty gray shadows for a tense, gritty mood." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "computer monitor", "usage_kind": "anchor", "source_phrase": "faint glow of a computer monitor screen" }, { "owned_token": "desk", "usage_kind": "anchor", "source_phrase": "The hard desk surface immediately below falls into cold, dirty gray shadows" }, { "owned_token": "keyboard", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "office phone", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "paperwork", "usage_kind": "anchor", "source_phrase": "showing out-of-focus paperwork" }, { "owned_token": "partition", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "task chair", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "7046cdc38fe62631", "owned_hash": "6afee18c21e3650a", "camera_direction_hash": "12caa4f6f077380e", "owned_usage_hash": "a0f549134b3b9342", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "computer monitor", "violating_phrase": "faint glow of a computer monitor screen", "reason": "The computer monitor is mentioned in the blurred background to establish the environment glow, serving as an anchor reference with no redraw instruction, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "desk", "violating_phrase": "The hard desk surface immediately below falls into cold, dirty gray shadows", "reason": "The desk surface is referenced as a physical anchor for placing the paper and casting shadows under light, with no redraw instructions, matching the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "paperwork", "violating_phrase": "showing out-of-focus paperwork", "reason": "The paperwork is mentioned as being part of the soft, distant background blur to anchor the scene, matching the producer's declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 21, "dependency_reason": "S21_Shot5 exact_background 유지", "_shot_index": 8, "visible_entities": [ "C01", "P07", "P08" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 21, "shot_index": 8 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "insert", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Very low at tabletop height within eight inches of the paper, the camera fixes on 수리영’s writing hand as the ballpoint pen tip presses into a small note in the center of the frame. Her wrist and fingers occupy the foreground, the note sits flat beneath them, and the police-office background is reduced to blurred desk clutter beyond the hand.", "lighting_mood": "Harsh fluorescent light glints on the pen tip and whitens the paper, while the surrounding desk stays gray and tense.", "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", "P07", "P08" ], "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 frame isolates 수리영’s hand and writing action; the full character outlook is not visible and should not be forced." } ], "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": [ "chair", "computer monitor", "desk", "keyboard", "office phone", "paperwork", "partition", "task chair" ], "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": "police_office_environment", "element_type": "environment_state", "character_name": "", "description": "The environment around the police office desk is illuminated by harsh, greenish fluorescent overhead lights. The desk surface and surrounding areas fall into cold, dirty gray shadows, with generic, blurred office clutter visible in the background. This lighting setup maintains a bleak and tense atmosphere across the setting.", "applies_to_shots": [ 5, 8 ], "element_scope": "full", "source_facts": [ "Greenish office fluorescents flatten her skin while the desk foreground falls into dirty gray shadow", "Harsh fluorescent light glints on the pen tip and whitens the paper, while the surrounding desk stays gray and tense", "blurred desk clutter" ], "visual_inferences": [ "The lighting is positioned above, casting stark highlights and desaturating the colors of the desk area.", "The desk surface has a dull finish that absorbs the ambient gray shadows." ], "creative_decisions": [ "Maintain a consistent greenish-gray color palette in the environment to emphasize the gritty and uncomfortable mood of the police station." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 21, "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": "C01", "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": "1b9a13bd4103798e" }, { "scene_index": 22, "heading": "S#22. 경찰서 밖 / D", "beat_title": "좌표를 확인하고 급하게 자전거로 출발하는 수리영", "representative_moment": "수리영(C01)이 자전거 핸들을 꽉 쥐고 상체를 숙인 채 도로를 향해 힘차게 페달을 밟으며 나아간다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Low side-tracking wide shot with flat daytime gray and hard side contrast", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker jacket, a Korean young woman in her early 20s, is positioned in the middle-center midground mounted on the existing bicycle, moving toward screen right along the street. Viewed from a low side-tracking observer position at hip height, her full body fits within the wide frame as she interacts with the bicycle. She is in right profile, leaning her upper body forward, gripping the handlebar of the bicycle tightly, and gazing off-screen. Frozen in mid-pedal, one knee is bent downward for a powerful forward stroke, her long straight black hair flowing back in the direction opposite to her motion. Her dynamic movement is defined against the background of the dull gray pavement. Flat daytime gray ambient light is cut by hard side contrast. Behind her, the muted gray facade of the L14B01 police station exterior slips by as a soft background, with the ordinary sidewalk and curb line receding parallel to the lower edge to emphasize the surge of acceleration.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Suri-young action: riding the parked bicycle and running quickly.", "Shot 2 description: upper body leaning forward, gripping handles tightly, one knee bent, forceful pedal stroke, hair blowing back in the wind.", "C01 angles: profile_right, leaning into pedal, gaze=off_screen, state=alive." ], "visual_inferences": [ "Low side-tracking camera captures wide frame for speed impression, preserving the full body and bicycle geometry.", "Hair flowing backward acts as a visual vector opposite to the bicycle's forward motion, capturing the mid-action freeze." ], "creative_decisions": [ "Flat daytime lighting is used to set the gritty mood, focusing sharp side contrast directly on the bicycle frame.", "Police station facade and curb line moving in the background emphasize mid-action acceleration." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bicycle", "usage_kind": "anchor", "source_phrase": "mounted on the existing bicycle, moving toward screen right along the street. Viewed from a low side-tracking observer position at hip height, her full body fits within the wide frame as she interacts with the bicycle. She is in right profile, leaning her upper body forward, gripping the handlebar of the bicycle tightly" }, { "owned_token": "bicycle rack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curb", "usage_kind": "anchor", "source_phrase": "and curb line receding parallel" }, { "owned_token": "driveway", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "facade", "usage_kind": "anchor", "source_phrase": "the muted gray facade of the L14B01 police station exterior" }, { "owned_token": "glass door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "pavement", "usage_kind": "anchor", "source_phrase": "against the background of the dull gray pavement" }, { "owned_token": "police station", "usage_kind": "anchor", "source_phrase": "the L14B01 police station exterior" }, { "owned_token": "sidewalk", "usage_kind": "anchor", "source_phrase": "with the ordinary sidewalk and curb line" }, { "owned_token": "stair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "street", "usage_kind": "anchor", "source_phrase": "moving toward screen right along the street" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "047ab9d901fc55fe", "owned_hash": "4d3543c50265e9a8", "camera_direction_hash": "4970177e2ca7af1c", "owned_usage_hash": "c2d8f1245210af6c", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bicycle", "violating_phrase": "mounted on the existing bicycle, moving toward screen right along the street. Viewed from a low side-tracking observer position at hip height, her full body fits within the wide frame as she interacts with the bicycle. She is in right profile, leaning her upper body forward, gripping the handlebar of the bicycle tightly", "reason": "The prompt describes the character interacting with and mounted on the existing background bicycle, serving as a physical spatial anchor, which aligns with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "curb", "violating_phrase": "and curb line receding parallel", "reason": "The curb line is mentioned as a compositional background guide receding in perspective, confirming it is used as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "facade", "violating_phrase": "the muted gray facade of the L14B01 police station exterior", "reason": "The facade is referenced as a soft background element that slips by, acting purely as an environmental anchor.", "verdict": "anchor_reference" }, { "owned_object": "pavement", "violating_phrase": "against the background of the dull gray pavement", "reason": "The pavement is used to define the background backdrop of the action, behaving as an environmental anchor.", "verdict": "anchor_reference" }, { "owned_object": "police station", "violating_phrase": "the muted gray facade of the L14B01 police station exterior", "reason": "The police station exterior is mentioned as a background element to anchor the scene's location.", "verdict": "anchor_reference" }, { "owned_object": "sidewalk", "violating_phrase": "with the ordinary sidewalk and curb line receding parallel", "reason": "The sidewalk is mentioned to establish perspective and background context, acting as an anchor.", "verdict": "anchor_reference" }, { "owned_object": "street", "violating_phrase": "moving toward screen right along the street", "reason": "The street is referenced as a path along which the character moves, serving as a spatial anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 2, "visible_entities": [ "C01", "L14" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 22, "shot_index": 2 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the flow's low side-tracking position, the camera runs parallel to 수리영 at hip height, held slightly wider than default so her full body, the entire bicycle, the bent knee, and the forward pedal stroke all sit inside the frame. She travels left-to-right across the midground, with the police-station exterior slipping by in the background and the front wheel kept just inside the right edge to preserve the feeling of acceleration.", "lighting_mood": "Flat daytime gray is cut by hard side contrast on the bicycle frame, keeping the realism gritty while her motion provides the visual surge.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "movement_direction", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "midground", "gesture_action": "moves_toward", "gesture_target_label": "screen right along the street", "label": "수리영 on bicycle", "screen_zone": "middle_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L14" ], "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": "L14B01", "reference_usage": "exact_background", "owned_objects": [ "bicycle", "bicycle rack", "curb", "driveway", "facade", "glass door", "pavement", "police station", "sidewalk", "stair", "street", "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": "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": "2f57c3489cce2f73" }, { "scene_index": 23, "heading": "S#23. 어느 PC방 / 실내", "beat_title": "좌표의 시각화", "representative_moment": "모니터 화면 가득 띄워진 차가운 해상 지도 한가운데 붉은 핀이 꽂혀 있는 핏빛 정지 화면.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Tight flat insert shot of a digital screen with pixel-glow texture and heavy crushed black vignette", "t2i_prompt": "Photorealistic cinematic still. A digital maritime map graphic fills the middle-center midground, displaying a bright red location pin graphic firmly embedded in the middle-center midground. The composition is a tight, flat graphic insert flush with the screen plane, featuring muted coastlines and grid-like sea markings spanning the interface. The visual texture carries a slight pixel-glow under harsh, sterile bluish digital light. Deep, heavy crushed black vignetting at the edges implies the dark, cold blue-black ambient atmosphere of the unseen PC room beyond the screen boundaries. The unmoving red pin strongly contrasts with the cool-toned map, glowing with intense saturation like an occult wound isolated against the stark digital space.", "outfit_assignments": [], "source_facts": [ "monitor screen displaying a maritime map with coastlines and grid", "red location pin held near the visual center", "static image plane without camera drift" ], "visual_inferences": [ "fixed_element pc_room_cold_lighting: dark, cold, blue-black ambient shadows with crushed black surroundings", "the bezel and observer are excluded to focus entirely on the digital interface", "slight pixel-glow texture reflects the 'through_device' perception mode" ], "creative_decisions": [ "used heavy crushed black vignetting at the frame edges to blend the digital map into the surrounding cold PC room darkness", "maximized the color contrast between the sterile bluish monitor light and the bright red pin to emphasize the occult wound feeling" ], "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": "2f4dfc0218dfb0ca", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "3ed5a8e0315f68f8", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 3, "visible_entities": [], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 23, "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 camera inherits the tight frontal monitor position and becomes almost flush with the screen plane, excluding 수리영 and most of the bezel so the maritime map reads as a flat graphic insert. The red location pin is held near the visual center, surrounded by muted coastlines and grid-like sea markings, with no camera drift.", "lighting_mood": "Cold monitor light and crushed surrounding blacks make the red pin feel like an occult wound on an otherwise sterile digital map.", "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": "maritime map graphic", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "red location pin", "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": "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_room_cold_lighting", "element_type": "environment_state", "character_name": "", "description": "The indoor space is steeped in a dark, cold atmosphere characterized by blue-black ambient shadows and crushed black surroundings. The primary illumination emanates from the harsh, cold glow of a computer monitor, casting a sterile bluish light that strongly contrasts with the deep, heavy shadows of the room.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Cold monitor light and crushed surrounding blacks (Shot 3)", "Blue-black PC-room ambience wraps her torso (Shot 5)" ], "visual_inferences": [ "The room is generally dark, making the monitor the dominant light source.", "The overall color temperature is distinctly cool and blueish." ], "creative_decisions": [ "Emphasizing the heavy, contrasty shadows to evoke a sterile and isolating mood within the PC room setting." ], "confidence": "high" } ], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [], "forbidden_refs": [], "readiness_policy": "not_applicable", "constraints": [ "do not imply or describe a reference image that is not listed in required_refs (no phantom references)", "if a required character_outlook ref is missing, scene image generation must block before the image stage" ] }, "render_contracts": [], "_card_metadata": { "lift_status": { "rule_a_lifted": true, "rule_c_lifted": true, "rule_e_lifted": false, "rule_h_lifted": true, "id_policy_composite_lifted": true, "id_policy_close_framing_face_lifted": true, "id_policy_reproduction_surface_lifted": true, "continuity_cross_shot_id_substitution_lifted": true, "continuity_ref_usage_constraints_lifted": true, "continuity_view_consistency_lifted": true, "continuity_constraints_extended": true, "spatial_camera_frame_lifted": true, "spatial_fg_bg_shared_anchor_lifted": true, "spatial_primary_framing_lifted": true, "spatial_constraints_extended": true }, "rule_source": { "camera_rule": "A", "owned_rule": "C", "close_skip_rule": "E", "demographic_rule": "H", "composite_id_rule": "C## 사용 규칙 — 참조 이미지 연동 (v17:32-64)", "close_framing_face_rule": "극단 클로즈업 표현 (v17:66-85)", "reproduction_surface_rule": "사진·포스터·화면·거울 속 인물 규칙 (v17:133-146)", "cross_shot_id_substitution_rule": "v20/scene_detail/system.md#continuity-section", "ref_usage_rule": "v20/scene_detail/system.md#continuity-section", "view_consistency_rule": "v20/scene_detail/system.md#continuity-section", "camera_frame_rule": "v20/scene_detail/system.md#spatial-consistency", "fg_bg_shared_anchor_rule": "v20/scene_detail/system.md#spatial-consistency", "primary_framing_rule": "v20/scene_detail/system.md#spatial-consistency" } } }, "render_prompt_card_hash": "39f27dbff1b67e9b" }, { "scene_index": 23, "heading": "S#23. 어느 PC방 / 실내", "beat_title": "좌표 확인 및 지도 확보", "representative_moment": "수리영이 푸른 모니터 빛 아래서 종이 지도를 반으로 접어 자신의 백팩 입구에 밀어 넣는다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "waist-to-chest height observer framing with cold blue-black ambient shadows", "t2i_prompt": "Photorealistic cinematic still. P05, a folded paper map halfway inserted, sits in the middle-center foreground, pointing into the open backpack P04 mouth located in the lower-right foreground. From a waist-to-chest height observer position beside the desk, the camera looks diagonally across the cropped torso of C01O01 in a gray casual windbreaker, a Korean young woman at a three-quarter angle from the right. She is gazing down, her face mostly extending beyond the upper frame edge. The visible folded face of P05 shows faint map lines and creases, catching a dull gray highlight, while its inner half disappears into the unzipped mouth of P04, which faces diagonally up toward the camera revealing a dark, heavily shadowed interior. Blue-black PC-room ambience wraps her torso against crushed black surroundings. A cluttered computer desk edge cuts diagonally through the midground. A nearby monitor screen in the background is out of focus, its blurred side portion displaying indistinct pale interface light, providing a sterile, harsh bluish glow over the scene.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#23 action: '여러 장의 지도를 접어 백팩에 넣는다.'", "shot description: folded map halfway inserted into open bag", "Cold monitor light and crushed surrounding blacks (Shot 3)", "Blue-black PC-room ambience wraps her torso (Shot 5)" ], "visual_inferences": [ "The room is generally dark, making the monitor the dominant light source.", "The overall color temperature is distinctly cool and blueish.", "Desk cuts diagonally across the midground to anchor the workstation space." ], "creative_decisions": [ "Waist-to-chest medium framing emphasizes the prop interaction in the foreground while placing the face mostly out of frame to concentrate on the physical action.", "Heavy shadows inside the backpack contrast with the dull gray highlight on the map, enhancing the gritty tactile mood." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "character", "prop" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "5c25010f4b6f2822", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "d9c9cbdd90d5fa3a", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 23, "dependency_reason": "exact_background ref from Shot 3 of the same scene ensures consistent PC room lighting and spatial layout", "_shot_index": 5, "visible_entities": [ "C01", "P04", "P05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 23, "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 the side of the PC-room desk at waist-to-chest height, the camera holds the end of the tracking move about three feet from 수리영, looking diagonally across her cropped torso toward the open bag at the lower edge. The folded paper map is frozen halfway into the bag mouth, while her face remains secondary and mostly outside the upper frame.", "lighting_mood": "Blue-black PC-room ambience wraps her torso, with the paper map catching a dull gray highlight and the bag interior falling into heavy shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "folded map halfway inserted", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "open backpack mouth", "screen_zone": "lower_right", "target_id": "", "target_kind": "background" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "P04", "P05" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": 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_room_cold_lighting", "element_type": "environment_state", "character_name": "", "description": "The indoor space is steeped in a dark, cold atmosphere characterized by blue-black ambient shadows and crushed black surroundings. The primary illumination emanates from the harsh, cold glow of a computer monitor, casting a sterile bluish light that strongly contrasts with the deep, heavy shadows of the room.", "applies_to_shots": [ 3, 5 ], "element_scope": "full", "source_facts": [ "Cold monitor light and crushed surrounding blacks (Shot 3)", "Blue-black PC-room ambience wraps her torso (Shot 5)" ], "visual_inferences": [ "The room is generally dark, making the monitor the dominant light source.", "The overall color temperature is distinctly cool and blueish." ], "creative_decisions": [ "Emphasizing the heavy, contrasty shadows to evoke a sterile and isolating mood within the PC room setting." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 23, "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": "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": [ 5 ], "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": "83a59445ea5d9a73" }, { "scene_index": 24, "heading": "S#24. 해안가 / 소나기", "beat_title": "거친 폭우 속 필사적인 질주", "representative_moment": "쏟아지는 폭우를 뚫고 결연한 표정으로 자전거 페달을 밟는 C01의 젖은 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level frontal close-up traveling backward, shallow depth of field, cold metallic storm light", "t2i_prompt": "Photorealistic cinematic still. C01O01 in a gray windbreaker, a Korean woman in her early 20s with a round face and dark eyes, her rain-soaked face occupying the middle-center foreground. She is facing the camera, looking at the camera with ominous determination, her wet face filling the center in sharp focus. Her long black hair flows back in the direction opposite to her forward motion. Cold, metallic storm light casts deep shadows under her wet bangs, with sudden pale highlights catching on the heavy raindrops slicing across the upper foreground. The top bar and grips of the bicycle frame are visible as blurred fragments at the lower frame edge. [: Heavy, wind-driven rain streaks veil the frame, backed by a gray-black, slick atmospheric blur of the stormy coastal road.]", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#24 action: 자전거를 타고 빗속을 달리는 수리영", "shot description: 빗물이 맺힌 앞머리 사이로 매섭게 앞을 노려보는 수리영의 젖은 얼굴 클로즈업", "DP camera framing: The camera follows the flow into an eye-level frontal position less than a meter ahead of the moving bicycle" ], "visual_inferences": [ "cold metallic storm light emphasizing determination without red tones", "hair flowing backward to preserve the mid-action motion direction", "bicycle handlebars positioned at the lower edge to indicate motion without overtaking the face focus" ], "creative_decisions": [ "tight frontal close-up to isolate the intense facial expression against the harsh weather", "shallow depth of field to blur the road and focus sharply on the face and raindrops", "L02 block used instead of L01 for frame-edge surface description due to close framing skipping the background reference" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "914f2d6a054f70a9", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "88bcdbc4482b7c17", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 9, "dependency_reason": "동일한 해안가 공간의 배경 톤 참조 (클로즈업 샷 정책에 따라 직접적인 레이아웃 합성 대신 분위기와 조명 톤만 유지)", "_shot_index": 3, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 24, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera follows the flow into an eye-level frontal position less than a meter ahead of the moving bicycle, traveling backward at her speed but compressed tighter than a normal close-up. 수리영's wet face fills the center, her bangs and raindrops slicing across the upper foreground while the road and handlebars are only blurred fragments at the lower edge.", "lighting_mood": "Storm light is cold and metallic, with sudden pale highlights on raindrops and deep shadow under her bangs sharpening the redless, ominous determination.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영 rain-soaked face", "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": 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": "9504c391e288591a" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "물리적 공간에 남겨진 기괴한 흔적과 직면", "representative_moment": "거칠게 그려진 붉은 원 P02 한가운데 피 묻은 낡은 흑백 사진 P03이 벽에 꽂혀 있는 섬뜩한 정물.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close framing, chest-height observer POV, stark high contrast dusk lighting", "t2i_prompt": "Photorealistic cinematic still. In the middle-center background, P03, a blood-stained old photograph pinned at center, sits perfectly within P02, a rough red circle on the bedroom wall. The camera holds rigidly square to the surface, framing only the [L05: scuffed muted gray plaster] and the central tableau. The front face of P03 is stained with blood, its image content partially obscured by a small metal pin head pressing directly into the plaster. A low, golden dusk light barely leaks in and grazes the rough textures of the wall, letting P02 flare intensely against the deep, thick blue and gray-brown shadows that heavily engulf the frame edges.", "outfit_assignments": [], "source_facts": [ "wall with a red circle drawn on it", "blood-stained photo pinned in the center", "shot description: the camera holds rigidly at chest height three to four feet from the wall, square to the rough red circle and the pinned photograph" ], "visual_inferences": [ "fixed_element dusk_lighting_atmosphere: low golden dusk light grazes the surfaces, contrasting sharply with deep thick blue and gray-brown shadows", "skipped_close_framing: omitting the wider room layout to focus entirely on the wall surface texture and the props" ], "creative_decisions": [ "high contrast grazing light to emphasize the tactile surface of the plaster and the stark redness of the circle", "strictly centered composition to create an inescapable, documenting-a-crime-scene dread" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002" ], "reference_phrase_kinds": [ "prop" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "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": "4476f11ea90b5919", "owned_hash": "b9a131aea2139fe0", "camera_direction_hash": "0b5bcd69ecef17dd", "owned_usage_hash": "a90230c91f24a0b9", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "동일한 옥탑방 배경(단, 본 샷은 close framing으로 벽면 질감에만 한정하여 이전 샷의 조명 분위기를 유지)", "_shot_index": 4, "visible_entities": [ "L05", "P02", "P03" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Inside the bedroom, the camera holds rigidly at chest height three to four feet from the wall, square to the rough red circle and the pinned photograph as if documenting a crime-scene discovery. The red circle occupies the mid-to-upper wall area, while the blood-stained old black-and-white photograph sits exactly at its center, small but inescapable.", "lighting_mood": "Gold dusk leaking from elsewhere barely grazes the wall, letting the red circle flare against muted gray plaster and deep bedroom shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "blood-stained old photograph pinned at center", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "rough red circle on bedroom wall", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "L05", "P02", "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": "L05B02", "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": "dusk_lighting_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The interior of the room is bathed in a low, golden dusk light that barely leaks in and grazes the surfaces. This warm illumination contrasts sharply with deep, thick blue and gray-brown shadows that heavily engulf the rest of the space.", "applies_to_shots": [ 4, 7, 9 ], "element_scope": "full", "source_facts": [ "Gold dusk leaking from elsewhere barely grazes the wall", "deep bedroom shadow", "Low gold dusk mixes with blue interior shadow", "narrow dusk highlight catches the knuckles", "room fall into thick, tactile shadow" ], "visual_inferences": [ "The lighting remains consistent across different camera angles, defining the somber and heavy mood of the late afternoon scene." ], "creative_decisions": [ "Using the stark high contrast between the warm gold streak and the cold, oppressive shadows to physically manifest the character's psychological tension." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 14, "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": "prop", "id": "P02", "policy": "required" }, { "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": 25, "shot_indices": [ 4 ], "duration": "single_shot" }, "targets": [ { "entity_id": "P02", "role": "visual_target" } ], "requirements": [ { "dimension": "visual_identity", "operation": "preserve", "strength": "required", "reference_policy": "use_entity_reference" } ] }, { "contract_id": "rc_002", "scope": { "scene_index": 25, "shot_indices": [ 4 ], "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": "ea0d7747c94bdb3d" }, { "scene_index": 25, "heading": "옥탑방 안 / 해질 무렵", "beat_title": "충격과 무력감 속 새로운 단서에 이끌리는 수리영", "representative_moment": "다리에 힘이 풀린 듯 침대 모서리에 걸터앉은 자세로 굳어버린 수리영의 전신.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Wide framing from a waist-height angle aiming at the bed corner, integrating high-contrast dusk and interior shadows.", "t2i_prompt": "Photorealistic cinematic still. Waist-height wide camera aimed at the bed corner, full body in frame. C01O01 in a gray windbreaker, a Korean woman in her 20s, is perched on the edge of the bed mattress covered in muted fabric, positioned at a three-quarter angle from the left, gazing into the distance with her weight collapsed downward. The interior of the room is bathed in a low, golden dusk light that barely leaks in and grazes the surfaces. This warm illumination contrasts sharply with deep, thick blue and gray-brown shadows that heavily engulf the rest of the space. The bed acts as a stationary structural anchor within the room, partially submerged and fading into the surrounding deep interior shadows. Behind her, the open window reveals dim exterior dusk, and muddy footprints on the floor carry through the room disorder, with the existing chairs cast in deep shadow.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영 three_quarter_left, perched on bed edge, gaze=distant, state=alive", "From near the foot of the bed at waist height, the camera tracks into a medium-wide angle aimed at the bed corner, widened enough to contain 수리영's full body and the violated room around her.", "overturned furniture, and scattered traces remain visible behind her as psychological pressure rather than mere set dressing." ], "visual_inferences": [ "fixed_element dusk_lighting_atmosphere: The interior of the room is bathed in a low, golden dusk light that barely leaks in and grazes the surfaces. This warm illumination contrasts sharply with deep, thick blue and gray-brown shadows that heavily engulf the rest of the space.", "fixed_element bedroom_bed_edge: The edge and corner of a bed mattress covered in muted fabric. It acts as a stationary structural anchor within the room, partially submerged and fading into the surrounding deep interior shadows." ], "creative_decisions": [ "Using wide framing to capture both the subject's collapse on the bed corner and the psychological pressure of the chaotic room behind her.", "Applying the specified high contrast between the warm dusk light and thick interior shadows to emphasize the heavy mood." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "on the edge of the bed mattress covered in muted fabric" }, { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "anchor", "source_phrase": "the existing chairs cast in deep shadow" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "usage_kind": "anchor", "source_phrase": "muddy footprints on the floor" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "the open window reveals dim exterior dusk" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "b6a30313804abe59", "owned_hash": "b9a131aea2139fe0", "camera_direction_hash": "5043094115b48cb2", "owned_usage_hash": "50bab24c7e220386", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "on the edge of the bed mattress covered in muted fabric", "reason": "The bed and its mattress are used as physical anchors for the character's pose, aligning with the producer's echo as an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "chair", "violating_phrase": "the existing chairs cast in deep shadow", "reason": "The prompt references existing chairs in the shadows as background context rather than requesting new models to be generated, confirming it is an anchor.", "verdict": "anchor_reference" }, { "owned_object": "floor", "violating_phrase": "muddy footprints on the floor", "reason": "The floor is used as a spatial surface reference for adding environmental footprint details, which acts as an anchor reference.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "the open window reveals dim exterior dusk", "reason": "The open window serves as a background anchor to frame the exterior dusk lighting, without requesting a redrawing of the window structure.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 18, "dependency_reason": "previous_shot_refs indicates S18_Shot1 exact_background usage for L05B02", "_shot_index": 7, "visible_entities": [ "C01", "L05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 25, "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 near the foot of the bed at waist height, the camera tracks into a medium-wide angle aimed at the bed corner, widened enough to contain 수리영's full body and the violated room around her. She is fixed on the edge of the mattress with her weight collapsed downward, while the open window, overturned furniture, and scattered traces remain visible behind her as psychological pressure rather than mere set dressing.", "lighting_mood": "Low gold dusk mixes with blue interior shadow, leaving her face drained while red echoes from the wall remain implied off-frame.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "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": 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": "L05B02", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "cabinet", "chair", "curtain", "doll", "door", "floor", "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": "dusk_lighting_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The interior of the room is bathed in a low, golden dusk light that barely leaks in and grazes the surfaces. This warm illumination contrasts sharply with deep, thick blue and gray-brown shadows that heavily engulf the rest of the space.", "applies_to_shots": [ 4, 7, 9 ], "element_scope": "full", "source_facts": [ "Gold dusk leaking from elsewhere barely grazes the wall", "deep bedroom shadow", "Low gold dusk mixes with blue interior shadow", "narrow dusk highlight catches the knuckles", "room fall into thick, tactile shadow" ], "visual_inferences": [ "The lighting remains consistent across different camera angles, defining the somber and heavy mood of the late afternoon scene." ], "creative_decisions": [ "Using the stark high contrast between the warm gold streak and the cold, oppressive shadows to physically manifest the character's psychological tension." ], "confidence": "high" }, { "element_id": "bedroom_bed_edge", "element_type": "persistent_prop", "character_name": "", "description": "The edge and corner of a bed mattress covered in muted fabric. It acts as a stationary structural anchor within the room, partially submerged and fading into the surrounding deep interior shadows.", "applies_to_shots": [ 7, 9 ], "element_scope": "full", "source_facts": [ "edge of the mattress", "beside the bed edge", "bed fabric" ], "visual_inferences": [ "The bed is a basic, low-profile structure typical of a modest rooftop room setup, heavily obscured by the lack of light." ], "creative_decisions": [ "Keeping the bed's fabric texture soft and unobtrusive so it blends seamlessly into the dark background, keeping focus on the lighting contrast." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 18, "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": "background", "id": "L05B02", "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": "1525f0cb4d793edf" }, { "scene_index": 25, "heading": "S#25. 옥탑방 안 / 해질 무렵", "beat_title": "피 묻은 사진을 향한 무력감과 결단", "representative_moment": "수리영이 구겨진 옛날 사진을 주먹에 꽉 쥔 채 멈춰 있는 손의 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "A macro hold at hand level with extremely shallow depth of field, illuminated by a narrow dusk highlight.", "t2i_prompt": "Photorealistic cinematic still. C01 in a gray windbreaker sleeve, a young Korean woman in her 20s, her clenched fist with the crumpled photograph, P03, occupies the lower-center foreground. Captured in a close macro hold at hand level, the fist clenches the paper tightly. P03's creased front portions face the camera, revealing fragments of a black-and-white image and dark blood stains visibly pinched between the fingers. A narrow, low golden dusk highlight barely leaks in, catching the knuckles and the photo creases, contrasting sharply with the deep, thick blue and gray-brown shadows. Behind the hand, the edge of the bed mattress covered in muted fabric runs horizontally, acting as a stationary structural anchor that fades into the surrounding dark interior shadows. The rest of the room is reduced to a soft, out-of-focus void.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "Shot description: Close-up of Suri-young's fist clutching the crumpled old photograph.", "DP instruction: The fist and crumpled photograph occupy the lower-center foreground in a near-static macro hold.", "DP instruction: The creased front portions of the photograph face the camera, revealing fragments of a black-and-white image and dark blood stains." ], "visual_inferences": [ "fixed_element dusk_lighting_atmosphere: A narrow, low golden dusk highlight catches the knuckles and photo creases, while deep blue and gray-brown shadows engulf the rest of the space.", "fixed_element bedroom_bed_edge: The edge of the bed mattress covered in muted fabric acts as a stationary structural anchor behind the hand, fading into the shadow.", "The shallow depth of field isolates the tense grip on the photo from the chaotic background, emphasizing the psychological weight of the moment." ], "creative_decisions": [ "Applying a macro insert scale to keep the focus entirely on the physical interaction between the hand and the symbolic prop (P03).", "Using the sharp lighting contrast between the warm golden dusk and cold ambient shadows to sculpt the knuckles and the creases of the photo." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001" ], "reference_phrase_kinds": [ "character", "prop", "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "anchor", "source_phrase": "the edge of the bed mattress covered in muted fabric runs horizontally" }, { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "floor", "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": "664bf74f506d75ee", "owned_hash": "b9a131aea2139fe0", "camera_direction_hash": "c5bde2ed0aab51ea", "owned_usage_hash": "725228b6325838b4", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "bed", "violating_phrase": "the edge of the bed mattress covered in muted fabric runs horizontally", "reason": "The prompt mentions the bed mattress edge as a horizontal structural anchor that fades into shadows, which aligns with the producer's declaration of an anchor reference rather than a redraw request.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 18, "dependency_reason": "Same room (rooftop interior); exact background continuity applies, though rendered out of focus in this insert shot.", "_shot_index": 9, "visible_entities": [ "C01", "P03" ], "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": "The camera settles at the end of the low dolly-in beside the bed edge, hand level and within inches of 수리영's clenched fist. The fist and crumpled photograph occupy the lower-center foreground in a near-static macro hold, with the rest of the room reduced to soft, gray-brown shadow beyond the bed fabric.", "lighting_mood": "A narrow dusk highlight catches the knuckles and photo creases while the surrounding bed and room fall into thick, tactile shadow.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "primary_subject_isolation", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영 clenched fist with crumpled photograph", "screen_zone": "lower_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "P03" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C01", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "The shot isolates only 수리영's hand and clenched fist; full outlook reference would not meaningfully align with the visible body fragment." } ], "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": "L05B02", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "cabinet", "chair", "curtain", "doll", "door", "floor", "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": "dusk_lighting_atmosphere", "element_type": "environment_state", "character_name": "", "description": "The interior of the room is bathed in a low, golden dusk light that barely leaks in and grazes the surfaces. This warm illumination contrasts sharply with deep, thick blue and gray-brown shadows that heavily engulf the rest of the space.", "applies_to_shots": [ 4, 7, 9 ], "element_scope": "full", "source_facts": [ "Gold dusk leaking from elsewhere barely grazes the wall", "deep bedroom shadow", "Low gold dusk mixes with blue interior shadow", "narrow dusk highlight catches the knuckles", "room fall into thick, tactile shadow" ], "visual_inferences": [ "The lighting remains consistent across different camera angles, defining the somber and heavy mood of the late afternoon scene." ], "creative_decisions": [ "Using the stark high contrast between the warm gold streak and the cold, oppressive shadows to physically manifest the character's psychological tension." ], "confidence": "high" }, { "element_id": "bedroom_bed_edge", "element_type": "persistent_prop", "character_name": "", "description": "The edge and corner of a bed mattress covered in muted fabric. It acts as a stationary structural anchor within the room, partially submerged and fading into the surrounding deep interior shadows.", "applies_to_shots": [ 7, 9 ], "element_scope": "full", "source_facts": [ "edge of the mattress", "beside the bed edge", "bed fabric" ], "visual_inferences": [ "The bed is a basic, low-profile structure typical of a modest rooftop room setup, heavily obscured by the lack of light." ], "creative_decisions": [ "Keeping the bed's fabric texture soft and unobtrusive so it blends seamlessly into the dark background, keeping focus on the lighting contrast." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 18, "shot_index": 15, "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": "C01", "policy": "required" }, { "kind": "background", "id": "L05B02", "policy": "required" }, { "kind": "prop", "id": "P03", "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": "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": 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": "635e2cc839056b09" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "안개 속 고립", "representative_moment": "짙은 안개가 깔린 텅 빈 포구, 갯바위 위에 멍하니 앉아 있는 수리영의 뒷모습.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "distant low-to-medium wide framing emphasizing the vast isolating fog", "t2i_prompt": "Photorealistic cinematic still. C01 in a gray windbreaker, a young Korean woman, appears as a small seated back in the lower-center midground, while the empty foggy harbor and unseen sea dominate the middle-center background. Seen from a distant low-to-medium observer perspective, the camera looks past her figure positioned low in the frame, framed by dark, slick, rain-soaked coastal rocks in the foreground. With her back to camera, she is seated on the wet rocks, gazing into the distance. The calm, cold, partly hidden harbor water reflects the thick, unmoving, opaque gray fog and a faint dull gold dusk residue rather than clear objects. In the distance, damp, desaturated wooden pier structures are barely visible as horizontal and vertical silhouettes receding into the mist. The after-rain ambient lighting bathes the environment in a dim silver-gray, making her dark silhouette feel entirely absorbed by the heavy atmospheric field.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#26 action: '비는 그치고 안개가 자욱하다. 텅 빈 포구... 갯바위에 앉아있는 수리영.'", "shot description: 짙은 안개가 깔린 텅 빈 포구, 갯바위 위에 멍하니 앉아 있는 수리영의 뒷모습." ], "visual_inferences": [ "fixed_element foggy_harbor_environment applied for dense fog, wet rocks, and silver-gray dusk lighting", "distant observer angle with the character appearing small against the overwhelming atmosphere" ], "creative_decisions": [ "wide framing to emphasize isolation and the vastness of the thick fog", "dull gold residue contrasting with dim silver-gray to create a melancholic, chilling dusk mood" ], "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": "518c683fb29291d7", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "4eb9014a39ea2f94", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No dependent scene indicated", "_shot_index": 1, "visible_entities": [ "C01", "L18" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "From the flow's distant low-to-medium position behind the wet rocks, the camera holds static about twelve meters behind 수리영, looking past her small seated back toward the fog-covered harbor. She remains low in the lower frame while the blank water, pier shapes, and dense mist swallow most of the composition.", "lighting_mood": "After-rain dusk turns the harbor silver-gray with a thin gold residue, making 수리영's small dark shape feel absorbed by the mist.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": { "reason": "required_background_position", "constraints": [ { "constraint_id": "fsc_001", "depth_plane": "background", "gesture_action": "none", "gesture_target_label": "", "label": "empty foggy harbor and unseen sea", "screen_zone": "middle_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "수리영 small seated back", "screen_zone": "lower_center", "target_id": "C01", "target_kind": "character" } ] }, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01", "L18" ], "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": "수리영 is seen as a small back-facing figure at a great distance; base identity is enough while detailed outlook reference would be visually underdetermined." } ], "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_environment", "element_type": "environment_state", "character_name": "", "description": "The harbor is enveloped in thick, muted gray fog following a recent heavy rain. The physical environment, including the coastal rocks and wooden pier structures, appears completely wet. The ambient lighting is a dim silver-gray typical of dusk, featuring a faint dull gold residue that lightly illuminates the dense mist and wet surfaces.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "소나기가 한차례 퍼 붓는가 싶더니, 비는 그치고 안개가 자욱하다.", "After-rain dusk turns the harbor silver-gray with a thin gold residue", "Muted gray fog wraps the dock while a dull gold dusk rim" ], "visual_inferences": [ "All exposed ground surfaces at the harbor are visibly wet and slightly reflective due to the preceding downpour.", "The thick fog severely limits visibility, obscuring distant elements and emphasizing the immediate surroundings like the wet pier and rocks." ], "creative_decisions": [ "Utilizing the contrast between the cool silver-gray fog and the subtle dull gold dusk light to create a melancholic, isolated atmosphere." ], "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" } ], "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": "74c95b3684e8f503" }, { "scene_index": 26, "heading": "S#26. 포구 / 해질 무렵", "beat_title": "안개를 뚫고 온 인우와 배", "representative_moment": "C05 stands on the foggy pier, one foot planted on the wet wood, looking down as he secures the taut mooring line to his arrived boat.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "wide low-angle full-body shot holding below eye level, muted gray fog with dull gold dusk rim lighting", "t2i_prompt": "Photorealistic cinematic still. C05O05 in a khaki waterproof jumper, a Korean man in his mid-20s with short black hair, tanned skin, and thick eyes, holding the mooring line in the middle-center midground, anchors the frame, while a diagonal mooring rope stretches across the lower-center foreground, and the arrived boat hull rests in the middle-right background. Viewed from a low observer height, his shoes and lower legs occupy the lower frame, his upper body extending upward. He stands at a three-quarter angle from the left, one foot planted firmly on the dark wet dock planks, looking down. Muted gray fog envelops the harbor, a dim silver-gray ambient dusk with a faint dull gold residue grazing the taut damp rope and the wet side of the boat hull, emphasizing the slippery wood under his split stance against the dense mist.", "outfit_assignments": [ { "character_id": "C05", "outlook_id": "O05" } ], "source_facts": [ "소나기가 한차례 퍼 붓는가 싶더니, 비는 그치고 안개가 자욱하다.", "안개를 뚫고 배 한척이 다가오는 게 보인다.", "배에서 내린 인우(24세. 남)는 닷줄을 고정시키고 있다." ], "visual_inferences": [ "fixed_element foggy_harbor_environment: dim silver-gray dusk, dull gold residue, wet coastal rocks and pier structures", "low camera angle visually anchors the dock planks to the bottom frame while pushing the upper body into the misty background", "the taut diagonal rope connects the character to the boat, showing physical tension without depicting mid-action movement" ], "creative_decisions": [ "wide framing below eye-level to capture the full body stance, the pier, and the docked boat in a single composition", "dull gold rim light traces the rope and wet boat hull, isolating the practical work action against the muted silver-gray fog", "integrating the frame spatial contract firmly into the first sentence to establish scene depth and relationships" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [ "fsc_001", "fsc_002", "fsc_003" ], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "01772e0e956afaae", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "d818368ed6eb6bed", "owned_usage_hash": "4f53cda18c2baa0c", "validator": "scene_detail_owned_objects.v1", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": 26, "dependency_reason": "exact_background rule linking to S26_Shot1, establishing the thick gray fog and wet harbor environment after the rain.", "_shot_index": 4, "visible_entities": [ "C05", "L18" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 26, "shot_index": 4 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "wide", "moment_lock": "single still moment from the selected shot only", "camera_direction": "On the dock, the camera holds slightly below eye level about four meters from 인우 in a three-quarter frontal full-body frame, preserving the boat, pier, and mooring action from the flow. The rope cuts diagonally from his hand toward the boat, and his one-foot-on-pier stance makes the vessel's recent arrival readable without changing the static reveal.", "lighting_mood": "Muted gray fog wraps the dock while a dull gold dusk rim grazes the rope and wet boat side, keeping 인우 practical yet ominously isolated.", "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": "arrived boat hull", "screen_zone": "middle_right", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_002", "depth_plane": "foreground", "gesture_action": "none", "gesture_target_label": "", "label": "diagonal mooring rope", "screen_zone": "lower_center", "target_id": "", "target_kind": "background" }, { "constraint_id": "fsc_003", "depth_plane": "midground", "gesture_action": "none", "gesture_target_label": "", "label": "인우 holding mooring line", "screen_zone": "middle_center", "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": [ "C05", "L18" ], "allowed_outlook_pairs": [ { "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": 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_environment", "element_type": "environment_state", "character_name": "", "description": "The harbor is enveloped in thick, muted gray fog following a recent heavy rain. The physical environment, including the coastal rocks and wooden pier structures, appears completely wet. The ambient lighting is a dim silver-gray typical of dusk, featuring a faint dull gold residue that lightly illuminates the dense mist and wet surfaces.", "applies_to_shots": [ 1, 4 ], "element_scope": "full", "source_facts": [ "소나기가 한차례 퍼 붓는가 싶더니, 비는 그치고 안개가 자욱하다.", "After-rain dusk turns the harbor silver-gray with a thin gold residue", "Muted gray fog wraps the dock while a dull gold dusk rim" ], "visual_inferences": [ "All exposed ground surfaces at the harbor are visibly wet and slightly reflective due to the preceding downpour.", "The thick fog severely limits visibility, obscuring distant elements and emphasizing the immediate surroundings like the wet pier and rocks." ], "creative_decisions": [ "Utilizing the contrast between the cool silver-gray fog and the subtle dull gold dusk light to create a melancholic, isolated atmosphere." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 26, "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": "C05O05", "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": "3780448d1dd873c2" }, { "scene_index": 27, "heading": "S#27. 옥탑방 안 / 해질 무렵", "beat_title": "절망적 상황을 깨달은 혜수의 경악", "representative_moment": "난장판이 된 옥탑방 한가운데서 혜수가 휴대전화를 귀에 댄 채 충격으로 굳어 있다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "eye-level medium framing with pale highlight against dirty interior shadow", "t2i_prompt": "Photorealistic cinematic still. C02O02 in a navy safari jacket, an Asian woman in her 40s, waist-up visible in the center of the L05B02 rooftop living room at three-quarter angle from the right, P01 smartphone pressed to her ear, eyes fixed off-screen. Her rigid face catches a pale highlight from the fading dusk on the phone side of her head. Dirty interior shadow flattens the space into gray-brown dread. In the background, the pre-existing chair, the window, and the door are visible against the wall as anchors.", "outfit_assignments": [ { "character_id": "C02", "outlook_id": "O02" } ], "source_facts": [ "S#27 action: 혜수가 수리영의 녹음메시지를 확인하며 전화통화 상태", "shot description: 혜수가 휴대전화를 귀에 댄 채 눈을 크게 뜬 굳은 얼굴" ], "visual_inferences": [ "C06(김형사) is not listed in allowed IDs, so he is treated as off-screen for this single frame", "the sunset hour and dirty room create fading dusk and dirty interior shadow" ], "creative_decisions": [ "medium framing to keep the rigid facial expression prominent while allowing the chaotic room to remain legible", "pale highlight on the face against gray-brown dread tone to emphasize the psychological shock" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "backpack", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "bed", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cabinet", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "chair", "usage_kind": "anchor", "source_phrase": "the pre-existing chair" }, { "owned_token": "curtain", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doll", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "door", "usage_kind": "anchor", "source_phrase": "the door" }, { "owned_token": "floor", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "anchor", "source_phrase": "the window" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "10eb0357cc5d9a33", "owned_hash": "b9a131aea2139fe0", "camera_direction_hash": "cef62f82699e2a9d", "owned_usage_hash": "859209f08170d7bd", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "chair", "violating_phrase": "the pre-existing chair", "reason": "The prompt mentions the chair as a background anchor, which matches the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "door", "violating_phrase": "the door", "reason": "The prompt references the door as a background anchor, which is consistent with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "window", "violating_phrase": "the window", "reason": "The prompt references the window as a background anchor, which is consistent with the producer's declaration.", "verdict": "anchor_reference" }, { "owned_object": "wall", "violating_phrase": "against the wall", "reason": "The prompt mentions the wall as an anchor point, though the producer declared it as absent. It is used as an spatial anchor.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 14, "dependency_reason": "동일한 옥탑방 배경(L05B02)의 구조 및 가구 배치 등 연속성 유지 (exact_background)", "_shot_index": 1, "visible_entities": [ "C02", "L05", "P01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 27, "shot_index": 1 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "medium", "moment_lock": "single still moment from the selected shot only", "camera_direction": "The camera holds the flow's eye-level medium-close position in the center of the ruined rooftop living room, slightly off-center in front of 혜수 so the surrounding mess remains visible around her rigid face. Her phone is pressed to her ear in the foreground side of her head, while the background chaos stays legible but softened, ready for the next move past her shoulder.", "lighting_mood": "Fading dusk and dirty interior shadow flatten the room into gray-brown dread, with the phone side of 혜수's face catching a pale, anxious highlight.", "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", "L05", "P01" ], "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": "L05B02", "reference_usage": "exact_background", "owned_objects": [ "backpack", "bed", "cabinet", "chair", "curtain", "doll", "door", "floor", "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": [], "previous_shot_refs": [ { "scene_index": 14, "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": "C02O02", "policy": "required" }, { "kind": "background", "id": "L05B02", "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": "3c983a5289d3f6aa" }, { "scene_index": 28, "heading": "S#28. 바다 + 갑판 / 해질 무렵", "beat_title": "사진의 정체를 깨닫고 경악하는 인우", "representative_moment": "수리영이 내민 사진을 보고 두 눈을 크게 뜬 채 심하게 당황하며 얼어붙은 인우의 얼굴 클로즈업.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Eye-level tight framing with shallow depth of field, contrasting warm gold and cool blue-gray lighting", "t2i_prompt": "Photorealistic cinematic still. C05O05 in a khaki waterproof jumper, a Korean man in his mid-20s with short black hair, tanned skin, and bold eye contours, his head and shoulders at eye level dominating the frame at a three-quarter angle from the left. His jaw is frozen slack and lips parted in sudden shock, looking down with widened eyes fixed on a blurred old photograph edge jutting low into the foreground. The image-bearing face of the photograph is turned toward him and partly toward the camera, its contents completely out of focus and unreadable. The lighting creates a stark contrast, with dying warm golden dusk light skimming one side of his face while the other side sinks into encroaching cool blue-gray maritime shadows. The shallow depth of field softens the immediate frame-edge surfaces, revealing only a dim wheelhouse-side sea haze and a weathered, damp boat railing running diagonally away from the camera, all enveloped in a damp sea mist under a fading dusk sky.", "outfit_assignments": [ { "character_id": "C05", "outlook_id": "O05" } ], "source_facts": [ "Shot 4 description: 사진을 본 순간 두 눈을 크게 뜨고 입술을 뗀 채 심하게 당황하여 얼어붙은 인우의 얼굴 클로즈업.", "dusk_maritime_environment: The environment on the small boat is enveloped in a damp sea mist under a fading dusk sky." ], "visual_inferences": [ "fixed_element body_full_pose: frozen jaw slack, looking down at the photograph with widened eyes.", "The shallow depth of field isolates the character's panic by blurring the photograph and the frame-edge background surfaces." ], "creative_decisions": [ "Described the framing purely through spatial placement ('head and shoulders at eye level dominating the frame') to avoid forbidden shot type terminology.", "Emphasized the stark contrast between the gold dusk light and blue-gray shadows to heighten the psychological tension." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "boat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cleat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "crate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gunwale", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hatch", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "a135d13069be0ef8", "owned_hash": "5a74d82b0ae8b8d1", "camera_direction_hash": "012783cb845338d2", "owned_usage_hash": "3fa982e805dfaef3", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 4, "visible_entities": [ "C05" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 28, "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 eye level and only about two feet from 인우, the camera holds the dolly-in at its closest endpoint, just off the line where 수리영’s hand would be offering the photograph. His face occupies most of the frame in a tight, direct close-up, angled slightly three-quarter so the widened eyes and parted lips carry the shock while the photograph edge remains a blurred pressure point low in the foreground.", "lighting_mood": "Gold dusk light skims one side of his face while the other falls into damp gray shadow, with the sea mist muting everything except the panic in his eyes.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C05" ], "allowed_outlook_pairs": [ { "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": [], "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": "L19B02", "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": "dusk_maritime_environment", "element_type": "environment_state", "character_name": "", "description": "The environment on the small boat is enveloped in a damp sea mist under a fading dusk sky. The ambient lighting features a stark contrast between the dying warm golden light and the encroaching cool blue-gray maritime shadows.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "해질 무렵 (dusk)", "바다위에 떠있는 소형어선 (small fishing boat floating on the sea)", "Gold dusk light skims one side... sea mist muting everything (Shot 4)", "narrow rim of dying gold... wheelhouse behind her sinks into blue-gray shadow (Shot 6)" ], "visual_inferences": [ "The overall atmosphere feels humid and increasingly dark, with the golden hour light rapidly transitioning into twilight on the ocean." ], "creative_decisions": [ "Emphasizing the contrast between the warm fading gold and the cold blue-gray shadows to heighten the psychological tension and isolation of being at sea." ], "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": "C05O05", "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": "aedaac6de560b809" }, { "scene_index": 28, "heading": "S#28. 바다 + 갑판 / 해질 무렵", "beat_title": "조타실 입구를 막아선 수리영", "representative_moment": "C01 stands at the wheelhouse entrance, her arms firmly braced across the doorway to block any entry.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "chest-height observer angle, near-lockoff framing, contrasting warm dusk rim light against cool interior shadows", "t2i_prompt": "Photorealistic cinematic still. At chest height squaring perfectly to the open wheelhouse doorway, C01O01 in a gray windbreaker, a young Korean woman in her 20s, stands centered like a human barricade facing the camera. Her arms are braced across the open threshold from the left doorframe to the right doorframe, her eyes fixed directly on the camera. Her torso and arms occupy the center of the frame, with her legs implied beyond the lower edge. Behind her back, the unlit helm console sits inside the deep blue-gray shadow of the wheelhouse interior. Smudged, salt-streaked glass panels flank the entrance. Enveloped in a damp sea mist under a fading dusk sky, a narrow rim of dying warm golden light traces the edges of her braced arms, contrasting sharply with the cool maritime darkness pooling inside the cabin. The damp outer deck extending toward the camera glows faintly with the remaining dusk.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "수리영, 조타실로 따라들어가 막으면서", "damp sea mist under a fading dusk sky, contrasting warm golden light and cool blue-gray maritime shadows (from S28 Shot 4)", "chest height, square to the doorway, arms braced from jamb to jamb", "damp outer deck, unlit helm console, smudged glass panels" ], "visual_inferences": [ "fixed_element dusk_maritime_environment", "torso and arms in frame, legs beyond lower edge to geometrically respect the chest-height camera position" ], "creative_decisions": [ "using the doorway as a rigid architectural frame around her body to emphasize the blockade", "applying a narrow rim of golden light to define her silhouette against the deep blue-gray shadow of the wheelhouse" ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background" ], "owned_object_usage": [ { "owned_token": "bulkhead", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "cleat", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "crate", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "deck", "usage_kind": "anchor", "source_phrase": "damp outer deck" }, { "owned_token": "door", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "doorframe", "usage_kind": "anchor", "source_phrase": "left doorframe to the right doorframe" }, { "owned_token": "glass panel", "usage_kind": "anchor", "source_phrase": "Smudged, salt-streaked glass panels" }, { "owned_token": "gunwale", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "hatch", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "threshold", "usage_kind": "anchor", "source_phrase": "open threshold" }, { "owned_token": "wall", "usage_kind": "absent", "source_phrase": "" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "ef185712eef8ca01", "owned_hash": "b31edb4d93991c59", "camera_direction_hash": "b9c28cea74628d46", "owned_usage_hash": "8a8ce7543c996b7d", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "deck", "violating_phrase": "damp outer deck", "reason": "The deck is mentioned to anchor the lighting and atmosphere (glowing with dusk) without requesting redraw or structural modification, aligning with the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "doorframe", "violating_phrase": "left doorframe to the right doorframe", "reason": "The doorframes are used strictly as spatial reference points for the character's bracing stance, matching the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "glass panel", "violating_phrase": "Smudged, salt-streaked glass panels", "reason": "The glass panels are described to set the texture and atmosphere of the existing environment rather than being newly drawn, aligning with the anchor declaration.", "verdict": "anchor_reference" }, { "owned_object": "threshold", "violating_phrase": "open threshold", "reason": "The threshold is used as a spatial anchoring reference for the character's arm placement, which matches the anchor declaration.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": 28, "dependency_reason": "Continuity with exact background and environmental lighting from S28 Shot 4.", "_shot_index": 6, "visible_entities": [ "C01" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 28, "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 outside the wheelhouse entrance at chest height, the camera squares itself perfectly to the doorway and settles into a near-lockoff about five feet from 수리영. She is centered like a human barricade, arms braced across the frame from jamb to jamb, with the dark wheelhouse interior behind her and the open deck compressed in front of her approach line.", "lighting_mood": "A narrow rim of dying gold outlines her arms while the wheelhouse behind her sinks into blue-gray shadow, turning the doorway into a black mouth she refuses to let him enter.", "perception_mode": "direct", "spatial_consistency": { "camera_frame_rule": { "applies_when": "camera_direction expresses a vertical camera height or viewpoint position and the scene description includes frame-edge surfaces", "consistency_check_summary": "elements described inside the frame must be at vertical/depth positions consistent with the camera's height — low cameras place feet/lower-body in the lower frame, overhead cameras place head/shoulders in upper frame, hip-height cameras place torso in frame with head/feet implied beyond edges", "core_principles": [ "low cameras at ground/quay/floor level place feet and lower legs in the lower frame; chest-height frame-edge surfaces behind the subject are geometrically forbidden", "overhead high angles from above place the top of head and shoulders in the upper-center frame; floor surfaces visible below feet at the lower frame edge are forbidden", "hip-height side cameras place torso and arms in frame; head and feet are implied beyond frame edges, not both fully visible together", "full body shots require either eye-level or high-angle framing where head and feet both fit; low cameras + full silhouette in one frame is forbidden" ], "forbidden_combinations": [ "low camera at ground/quay/floor level + frame-edge surface described at chest/waist height behind the subject", "overhead high angle + feet visible at the lower frame edge with floor visible below them", "hip-height side camera + full silhouette including head and feet visible in same frame", "low angle looking up + ceiling and floor both fully visible in same frame" ], "recommended_phrasings": [ "low observer at ground level, shoes and lower legs occupying the lower frame, upper body extending toward the upper edge", "overhead high angle from above, the top of head and shoulders fill the upper-center frame, the floor visible around them", "at hip height beside the subject, torso and arms in frame, head and lower legs implied beyond frame edges", "low angle looking up at the figure, ceiling lamps in the upper frame, the wall behind extending downward" ], "self_check_steps": [ "identify the camera position phrasing (height + angle)", "list elements described inside the frame (foreground prop, body parts, background surfaces)", "verify each element's vertical/depth placement is geometrically consistent with that camera position", "if mismatch found, adjust frame-edge description to match camera position" ], "rationale_summary": "T2I 모델은 카메라 위치와 frame 안 요소 vertical 좌표 모순 시 절충 합성 — 인물이 가구에 박히거나 신체 일부가 잘리거나 prop 이 떠 있는 부자연스러운 결과" }, "fg_bg_shared_anchor_rule": { "applies_when": "two characters share an interaction (dialogue / handing object / facing / co-watching) AND are described as foreground/background separated AND a shared surface or spatial anchor exists between them (bench / table / door / room)", "shared_anchor_required": "describe the shared surface or spatial anchor explicitly so both figures occupy the same physical space", "forbidden_phrasings": [ "fg/bg separation without naming the shared surface/space", "isolated foreground hand reaching with another figure 'in the rear background' (no shared anchor named)", "one figure 'in the foreground' and another 'in the back' without naming the room/surface they share" ], "recommended_keywords": [ "shared bench / shared table / shared surface", "the same X they both occupy (X = bench, sofa, table, doorway, room)", "across the X from each other", " crosses the air between them (interaction prop visually links the two)", "the sharing a single line of sight from the camera" ], "self_check_steps": [ "do the two characters share a single interaction (dialogue, handing, facing, co-watching)?", "is one foreground and the other background-separated?", "is the shared surface/space anchor explicitly named?", "if not, add one of the recommended keywords" ], "rationale_summary": "fg/bg 분리만 묘사하면 모델이 두 인물을 다른 공간에 배치하거나 인터랙션이 끊긴 어색한 합성 — 공유 anchor 명시가 같은 물리적 공간 강제" }, "primary_framing_rule": { "applies_when": "framing_scale ∈ {close, medium, wide, insert} from `shot_staging.framing_scale` enum (render_strategy.framing_scale, production SOT). LLM consumes the enum only; keyword classification is forbidden.", "close_framing_rules": { "primary_subject_only_fully_visible": "the named primary subject (hand / face / object) is the only fully visible entity in the frame", "other_entities_appearance_options": [ "partial visibility — shoulder slice / arm slice / sleeve / fingertips entering the frame edge", "soft background — out-of-focus silhouette via shallow depth of field", "absent — not rendered at all even if listed in visible_entities" ], "forbidden_combinations": [ "same character's full body and body-part close-up in the same frame (view mixing)", "two characters both rendered as sharp close-up faces in the same frame" ], "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "wide_medium_rules": { "primary_and_secondary_both_visible": "primary subject + secondary entities all renderable", "body_part_close_up_forbidden": "body-part close-up of any character is forbidden in wide/medium framing — close-up belongs in a separate shot", "fg_bg_separation_requires_shared_anchor": "if two characters are rendered fg/bg separated, also apply fg_bg_shared_anchor_rule", "subject_reference_policy_cross_ref": "see id_policy.subject_reference_policy (subject_reference_policy SOT — policy enum 에 따라 ID/outlook 사용)" }, "self_check_steps": [ "read framing_scale enum from render_strategy.framing_scale (do NOT infer from camera_direction)", "if close: verify only primary subject is fully visible; other entities are partial / soft bg / absent", "verify same character's full body + body-part close-up are not combined", "if close: verify two faces are not both rendered sharp" ], "rationale_summary": "한 shot 안 framing scale 가 일관되어야 모델이 단일 카메라 셔터로 해석. close/wide/medium 혼합은 multi-camera 합성으로 인식되어 인물 분열 / 부위 분리 / 시점 혼합 합성" }, "rationale_summary": "Rule F/G/J 의 spatial consistency 룰 통합 — camera 위치 vs frame edge 모순, fg/bg 분리 시 공유 anchor 누락, framing scale 혼합. 3 sub-rule 모두 single t2i_prompt = single camera shutter 원칙의 spatial 측면." }, "frame_spatial_contract": null, "constraints": [ "one prompt captures one still moment", "do not combine full-body framing and body-part close-up in one frame", "camera position and frame-edge elements must be geometrically consistent — see render_strategy.spatial_consistency.camera_frame_rule", "two characters in fg/bg separation must share a named surface/space anchor — see render_strategy.spatial_consistency.fg_bg_shared_anchor_rule", "framing_scale binds entity rendering — close = primary subject only fully visible, wide/medium = no body-part close-up — see render_strategy.spatial_consistency.primary_framing_rule" ] }, "id_policy": { "allowed_base_entity_ids": [ "C01" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": false, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [], "constraints": [ "ID format for visible subjects (C## bare, C##O## composite IDs, or common-noun descriptor) is determined per-subject by the id_policy per-subject policy array — see constraints[2] below", "do not use C##O## for a reproduced face inside any element flagged by id_policy.reproduction_surface_rule.applies (declared reproduction surface) — use a generic descriptor with demographic instead. NOTE: the physical photo/poster/document/picture-frame/map/key prop itself MUST use its P## ID — the generic-descriptor rule applies only to a face reproduction, not to the prop entity.", "per-subject identity reference policy is provided in id_policy.subject_reference_policy[] — apply each subject's policy when emitting references: (a) id_and_outlook_required ⇒ subject must use C##O## composite ID; (b) base_id_required ⇒ subject must use bare C## only (no outlook); (c) generic_descriptor_allowed ⇒ subject must NOT use C## or C##O##, render with a common/demographic descriptor only. omitted subjects default to id_and_outlook_required", "when an ID is required for a subject (policy = id_and_outlook_required or base_id_required) and a C##O## (or C##) is introduced for the first time in this t2i_prompt, append a 1-2 token demographic descriptor (ethnicity + age band, or ethnicity + role hint) consistent with visual_world_rules.region — never use scenario-specific proper nouns. does not apply to generic_descriptor_allowed subjects (no ID emitted)", "when an ID is required (policy = id_and_outlook_required or base_id_required), use the format `C##O## in <옷 1-3 단어>, ` or `C##O##, , <자세 표현>` — main characters with face refs may include this descriptor too (image inject takes precedence). does not apply to generic_descriptor_allowed subjects — those use a common/demographic descriptor with no C##/C##O##" ] }, "background_binding": { "mode": "background_ref_attached", "bg_id": "L19B01", "reference_usage": "exact_background", "owned_objects": [ "bulkhead", "cleat", "crate", "deck", "door", "doorframe", "glass panel", "gunwale", "hatch", "threshold", "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": "dusk_maritime_environment", "element_type": "environment_state", "character_name": "", "description": "The environment on the small boat is enveloped in a damp sea mist under a fading dusk sky. The ambient lighting features a stark contrast between the dying warm golden light and the encroaching cool blue-gray maritime shadows.", "applies_to_shots": [ 4, 6 ], "element_scope": "full", "source_facts": [ "해질 무렵 (dusk)", "바다위에 떠있는 소형어선 (small fishing boat floating on the sea)", "Gold dusk light skims one side... sea mist muting everything (Shot 4)", "narrow rim of dying gold... wheelhouse behind her sinks into blue-gray shadow (Shot 6)" ], "visual_inferences": [ "The overall atmosphere feels humid and increasingly dark, with the golden hour light rapidly transitioning into twilight on the ocean." ], "creative_decisions": [ "Emphasizing the contrast between the warm fading gold and the cold blue-gray shadows to heighten the psychological tension and isolation of being at sea." ], "confidence": "high" } ], "previous_shot_refs": [ { "scene_index": 28, "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": "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": "17e9dcbbd0f012d4" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "사진 속 소녀를 가리키며 진실을 밝히는 수리영", "representative_moment": "C01 presses her fingertip hard against the printed face of a young girl in the old P03 photograph on the work surface.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Extreme insert shot from a top-down perspective, with high-contrast amber dusk and cold gray shadows emphasizing the tactile texture of the photo and the pressing finger.", "t2i_prompt": "Photorealistic cinematic still. A dirty, scratched work surface inside the wheelhouse completely fills the frame from a top-down perspective. C01 in a gray windbreaker, a young Korean woman, is visible only as an isolated hand entering from the frame edge, her fingertip pressed hard against the surface. The skin is blanched white at the contact point. The fingertip pins down P03, a worn, slightly curled old black-and-white photograph. Beneath the pressing finger, the reproduced printed face of a young Asian girl in the photograph faces directly upward toward the camera. Low amber dusk light leaks obliquely across the matte paper, casting a slight warm glow on the skin, while heavy cold gray ambient shadows engulf the outer edges of the surface.", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" } ], "source_facts": [ "S#29 action: 수리영이 사진 속의 한 아이를 가리키며 '이 사진속의 얘가 우리 엄마예요'라고 말함.", "Shot description: 낡은 흑백 사진 속 어린 소녀의 얼굴을 손가락으로 꾹 짚은 채 정지한 수리영의 손 클로즈업.", "Fixed element: Low amber dusk leaks across the paper, wheelhouse shadows turn whites gray." ], "visual_inferences": [ "Since the framing scale is an extreme insert, C01's face and full body are out of frame, leaving only the hand visible.", "The photograph lies flat on a wheelhouse work surface to allow a firm downward press.", "The printed face inside the photograph is upward toward the camera to show the subject beneath the fingertip." ], "creative_decisions": [ "Used a top-down framing without camera terminology in the prompt, focusing purely on the spatial arrangement of the work surface, photograph, and hand.", "Enhanced tactile tension by describing the blanched skin at the contact point of the fingertip.", "Applied the base_id_required policy for C01, using a bare ID with demographic and clothing descriptors, and applied generic descriptors for the reproduced face in the photograph." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character", "background", "prop" ], "owned_object_usage": [ { "owned_token": "control console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "helm wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "instrument gauge", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "navigation display", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "radio", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "window", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "work surface", "usage_kind": "anchor", "source_phrase": "A dirty, scratched work surface inside the wheelhouse completely fills the frame" } ], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "5d9766d25891fe02", "owned_hash": "c7a7cb8dd15ed0f0", "camera_direction_hash": "43f6a1a07b9571cf", "owned_usage_hash": "6fcd38f406f860a5", "validator": "scene_detail_owned_objects.v1", "violations": [ { "owned_object": "work surface", "violating_phrase": "A dirty, scratched work surface inside the wheelhouse completely fills the frame", "reason": "The prompt references the work surface as the spatial background and contact plane for the hand and photograph, aligning with the producer's declaration of an anchor reference.", "verdict": "anchor_reference" } ] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "No direct prior visual dependency is required for this isolated insert moment.", "_shot_index": 3, "visible_entities": [ "C01", "P03" ], "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": "The camera is low and very close above 수리영’s hands, tightened beyond the flow’s default into an insert so the fingertip pressing the girl’s face and the old black-and-white paper are almost the only readable forms. The top angle flattens the photograph against the wheelhouse surface, with only a thin sliver of grime and shadow around the edges to keep it physically grounded.", "lighting_mood": "Low amber dusk leaks across the paper while the wheelhouse shadows turn the photograph’s whites gray and the fingertip slightly warm, making the family evidence feel fragile and ominous.", "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", "P03" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" } ], "must_use_composite_character_ids": true, "common_noun_required_when": [ "a reproduced face inside any declared reproduction surface", "unregistered extra" ], "id_use_required_when": [ "a physical photo/poster/document/picture-frame/map/key prop appearing in the scene — use its P## ID and preserve canonical content from entity_canon" ], "demographic_fallback_required": true, "face_identifiability_rule": { "use_entity_id_when": [ "face identifiable: front, profile, three-quarter, or eyes closed", "OTS framing with any visible face" ], "common_noun_required_when": [ "back to camera with no visible face", "silhouette or blurred outline only", "OTS with only back of head/shoulder, face entirely hidden" ], "id_use_summary": "C##/C##O## allowed only when face is identifiable AND framing is not dominated by an isolated body part", "rationale_summary": "C## 사용 시 얼굴 ref 가 자동 결합 — 얼굴이 안 보이는 인물에 C## 부착하면 ref 와 본문 묘사가 충돌" }, "reproduction_surface_rule": { "applies": true, "id_use": "forbidden — generic descriptor with demographic only", "rationale_summary": "C##O## 은 얼굴 reference 이미지를 원본 해상도로 inject 하므로 사진/화면/반사 표면 안 얼굴이 표면 밖 실물 크기로 합성됨. 이 rule 은 표면 안 얼굴 (face) 에만 적용 — 물리적 사진/액자/문서 prop 자체는 P## ID 보존 (Patch A)." }, "demographic_descriptor_policy": { "required_on_first_appearance": true, "applies_to_id_forms": [ "C##", "C##O##" ], "token_count_range": [ 1, 2 ], "components": { "ethnicity": [ "Asian", "East Asian", "South Asian", "Southeast Asian", "Black", "Middle Eastern", "Hispanic", "Latina", "Latino", "Caucasian" ], "gender": [ "man", "woman", "figure" ], "age_band": [ "in his/her 20s", "in his/her 30s", "in his/her 40s", "young", "middle-aged", "elderly" ], "role_hint_from_outfit": [ "in worker uniform", "in fisher workwear", "in detective coat", "in business suit" ] }, "format_template_a": "C##O## in <옷 1-3 단어>, ", "format_template_b": "C##O##, , <자세 표현>", "source_priority": [ "visual_world_rules.region", "visual_world_rules.era", "shot context outfit hint" ], "scenario_dependency_ban": "no work-specific proper nouns (character names, place names) — common nouns only", "regional_consistency": "한 작품 내 region 표기는 일관 — 선택된 세분 ethnicity (East Asian / Southeast Asian / South Asian / Caucasian / Black / Hispanic / Latina|o / Middle Eastern) 를 작품 단위 고정" }, "subject_reference_policy": [ { "subject_id": "C01", "policy_type": "identity_reference", "policy": "base_id_required", "reason": "Only 수리영’s hand and fingertip are framed in an extreme insert, so a full outlook reference would not be visually meaningful." } ], "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": "L20B02", "reference_usage": "exact_background", "owned_objects": [ "control console", "helm wheel", "instrument gauge", "navigation display", "radio", "window", "work surface" ], "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": "wheelhouse_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the wheelhouse is deeply engulfed in heavy, cold gray shadows. A warm, muted golden-amber dusk light leaks in from the side, casting a faint glow on the surfaces it touches and creating a stark, moody contrast with the dark surroundings.", "applies_to_shots": [ 3, 8 ], "element_scope": "full", "source_facts": [ "Low amber dusk leaks across the paper", "wheelhouse shadows turn the photograph’s whites gray", "Muted gold from the side window", "cabin remains heavy with gray shadow" ], "visual_inferences": [ "The setting sun provides the only illumination, entering obliquely from a side window and leaving much of the cabin unlit." ], "creative_decisions": [ "Maintaining a consistent high-contrast amber and gray lighting across different framings to emphasize the ominous, fragile mood of the scene before complete nightfall." ], "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": "L20B02", "policy": "required" }, { "kind": "prop", "id": "P03", "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": "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": 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": "79c3f04d2bd28c6c" }, { "scene_index": 29, "heading": "S#29. 조타실 / 해질 무렵", "beat_title": "혼미해진 수리영과 차가운 인우", "representative_moment": "의식을 잃은 C01이 C05의 어깨에 쓰러지고, C05가 그 모습을 굳은 표정으로 서늘하게 내려다본다.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "close frontal two-shot, high contrast dusk lighting", "t2i_prompt": "Photorealistic cinematic still. At shoulder height, torso and shoulders occupying the frame, lower bodies implied beyond the edges. C05O05 in a khaki waterproof jumper, a Korean man in his 20s, facing the camera, chin lowered rigidly, eyes fixed down on C01O01, his stiffened face in sharp focus. C01O01 in a gray casual windbreaker, a Korean woman in her 20s, at three-quarter angle from the right, unconscious, slack and limp with eyes closed, her softly-focused collapsed head slumped heavily into the lower-left portion of his chest line, both sharing the same narrow cabin space. Deeply engulfed in heavy cold gray shadows, a warm muted golden-amber dusk light leaks in from the side, catching C01O01's hair and C05O05's cheekbone. [L20B01: dim grimy rear wall partially obscured by their shoulders, a faint red circular mark visible among darker stains on the wall face angled slightly toward the camera].", "outfit_assignments": [ { "character_id": "C01", "outlook_id": "O01" }, { "character_id": "C05", "outlook_id": "O05" } ], "source_facts": [ "S#29 action: '의식이 혼미해지는 수리영. 수리영을 노려보는 인우의 어깨에 그대로 쓰러진다.'", "shot description: 의식을 잃은 수리영의 머리가 인우의 어깨에 툭 얹혀진 상태와 그 모습을 서늘하게 내려다보는 인우의 굳은 정면." ], "visual_inferences": [ "fixed_element wheelhouse_dusk_lighting: heavy cold gray shadows with warm muted golden-amber dusk light leaking from the side.", "To comply with the close framing rule prohibiting two sharp faces, C05O05 is in sharp focus while the collapsed head of C01O01 is softly focused." ], "creative_decisions": [ "Close frontal two-shot stacking their heads vertically to emphasize the chilling dynamic.", "Included a faint red circular mark on the frame-edge background wall to anchor the occult threat without breaking the skipped background mode." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [ "character" ], "owned_object_usage": [ { "owned_token": "control console", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "gauge", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "helm wheel", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "navigation display", "usage_kind": "absent", "source_phrase": "" }, { "owned_token": "radio", "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": "007ed938bfbdd4d5", "owned_hash": "dfdea5388fb24895", "camera_direction_hash": "f7d915522d98e181", "owned_usage_hash": "59e782f2209af0c1", "validator": "scene_detail_owned_objects.v1.close_skip", "violations": [] } } ], "scene_type": "normal", "dependent_scene_index": -1, "dependency_reason": "", "_shot_index": 8, "visible_entities": [ "C01", "C05" ], "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 and about three feet away, the camera holds a close frontal two-shot from the final reaction position, slightly low so 수리영’s collapsed head and 인우’s fixed face stack on the same vertical axis. The frame becomes nearly frozen after her head lands on his shoulder: her weight slumps into the lower-left portion of his chest line while his rigid face remains centered above, cold and still.", "lighting_mood": "Muted gold from the side window catches 수리영’s hair and 인우’s cheekbone, but the cabin remains heavy with gray shadow and a small red mark burns faintly behind them.", "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", "C05" ], "allowed_outlook_pairs": [ { "character_id": "C01", "outlook_id": "O01" }, { "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": [], "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": "L20B01", "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [ { "element_id": "wheelhouse_dusk_lighting", "element_type": "environment_state", "character_name": "", "description": "The interior of the wheelhouse is deeply engulfed in heavy, cold gray shadows. A warm, muted golden-amber dusk light leaks in from the side, casting a faint glow on the surfaces it touches and creating a stark, moody contrast with the dark surroundings.", "applies_to_shots": [ 3, 8 ], "element_scope": "full", "source_facts": [ "Low amber dusk leaks across the paper", "wheelhouse shadows turn the photograph’s whites gray", "Muted gold from the side window", "cabin remains heavy with gray shadow" ], "visual_inferences": [ "The setting sun provides the only illumination, entering obliquely from a side window and leaving much of the cabin unlit." ], "creative_decisions": [ "Maintaining a consistent high-contrast amber and gray lighting across different framings to emphasize the ominous, fragile mood of the scene before complete nightfall." ], "confidence": "high" } ], "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": "C05O05", "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": "18462831ce12629a" }, { "scene_index": 30, "heading": "S#30. 바다 / 해질 무렵", "beat_title": "안개 속으로 사라지는 어선의 뱃머리", "representative_moment": "어선의 뱃머리가 짙은 안개 속으로 절반쯤 진입하며 잔잔한 물살을 가르는 순간.", "t2i_variations": [ { "variant_label": "var_1", "camera_effect": "Water-level close framing tracking the bow, contrasting the golden dusk reflection with the cold gray fog", "t2i_prompt": "Photorealistic cinematic still. A low tracking observer at water level beside the bow, the waterline slicing across the lower third of the frame. The wet, weathered bow of a fishing boat occupies the left-middle of the frame without overpowering it, its forward half completely swallowed by [L21: an opaque gray-white fog wall closing over the upper and forward space]. The camera focuses on the nearer bow edge as it gently parts the calm sea surface into legible ripples. The calm water surface reflects broken gold dusk and the dark underside of the hull, distorted by the parted ripples. The last gold of sunset lies thinly on the water, while the dense fog desaturates the visible hull into cold gray, creating a soundless threshold.", "outfit_assignments": [], "source_facts": [ "S#30 action: 'A thick fog bank seems to erase the boundary with the world. In-woo's boat slowly enters the fog.'", "shot description: 'The bow of the fishing boat half enters the thick fog, parting the calm water in a cold, silent moment.'" ], "visual_inferences": [ "Low camera tracking at water level emphasizes the physical interaction between the hull and the water as it crosses the boundary.", "The thick fog desaturates the boat's hull into a cold gray, contrasting with the fading gold dusk reflected on the calm sea." ], "creative_decisions": [ "Close framing on the bow to capture the tension of entering the unknown threshold without showing the entire boat.", "Focusing on the legible ripples in the foreground to provide a textural contrast against the opaque, featureless fog wall." ], "confidence": "high", "applied_frame_spatial_constraint_ids": [], "reference_phrase_kinds": [], "owned_object_usage": [], "owned_validation": { "schema_version": 2, "t2i_prompt_hash": "a3cda90debb4c22b", "owned_hash": "e3b0c44298fc1c14", "camera_direction_hash": "129757fe2fd862ba", "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": [ "L21" ], "render_prompt_card": { "schema_version": 1, "shot_key": { "scene_index": 30, "shot_index": 3 }, "render_strategy": { "mode": "direct", "primary_subject": "", "framing_scale": "close", "moment_lock": "single still moment from the selected shot only", "camera_direction": "Tracking at water level beside the bow, the camera moves at the boat’s slow speed within a few meters of the hull, keeping the forward half of the fishing boat swallowed by fog while the nearer bow edge and parted ripples stay legible. The bow occupies the left-middle of the frame without overpowering it, the waterline slicing across the lower third as the gray fog wall closes over the upper and forward space.", "lighting_mood": "The last gold of sunset lies thinly on the water while the fog desaturates the hull into cold gray, making the boat seem to enter a soundless threshold.", "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": [ "L21" ], "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": null, "reference_usage": "skipped_close_framing", "owned_objects": [], "camera_reference": null, "close_framing_skips_background_ref": true, "constraints": [ "this shot uses close framing — the background reference is intentionally omitted at composition time", "describe the close subject and its immediate surroundings without relying on the absent background reference", "use [L##: ...] block for frame-edge surface description" ] }, "continuity_elements_used": { "fixed_elements": [], "previous_shot_refs": [], "forward_zoom_targets": [], "cross_shot_id_substitution_rule": { "applies_when": "fixed_elements[i].character_name maps to a C## allowed in id_policy.allowed_base_entity_ids OR a C##O## in id_policy.allowed_outlook_pairs", "substitution": "replace the common-noun person reference inside fixed_elements[i].description (e.g. 'An Asian man' / 'a woman' / 'a figure') with the matched C## or C##O##", "double_description_forbidden": "do not describe the same character with both a C##/C##O## and a separate common-noun person reference in the same t2i_prompt — produces phantom extra figures", "preserve_pose_unchanged": "fixed_elements[i].description 의 자세/위치/상태 keyword 는 그대로 유지 — 인물 참조 토큰만 교체", "no_repeat_after_pose": "이미 t2i_prompt 안에서 같은 인물의 자세를 묘사했다면, 같은 fixed_element description 을 별도 문장으로 반복하지 마라", "rationale_summary": "fixed_element description 이 보통명사로 작성되었고 같은 인물의 C## 묘사가 t2i_prompt 에 따로 있으면 모델이 두 인물로 인식해 인원 수가 부풀려진다" }, "ref_usage_constraints": { "zoom_in_detail": { "scope_summary": "same moment, same space, same primary subject; camera moved/zoomed onto a sub-region only", "forbidden_additions": [ "new persons not in the prior shot", "new props not in the prior shot", "new background elements not in the prior shot", "new poses, gestures, or expressions not in the prior shot", "different moments of action" ], "required_phrasings": [ "describe only the focused sub-region in concrete terms", "explicitly state the same environment (lighting, background) is preserved", "use continuity tokens such as 'the same figure', 'the same room', 'the same surface'" ], "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy applies per-subject; policy enum (id_and_outlook_required / base_id_required / generic_descriptor_allowed) determines C##/C##O## use", "rationale_summary": "zoom_in_detail = 앞 샷의 특정 영역을 확대한 '같은 사진'. 새 정보 추가 금지 — 현재 프레임이 보여주는 부분만 정확히 묘사" }, "exact_background": { "scope_summary": "same room, possibly different moment or angle; background carries through, persons/actions are described fresh for this shot", "permitted_additions": [ "new persons, poses, expressions, actions for this shot's moment", "different camera angle within the same room" ], "background_consistency_rule": "background elements established in the prior shot must remain consistent (no contradictory walls, doors, furniture)", "ignore_keep_handled_elsewhere": "ignore/keep directives are auto-applied via reference image processing — do not echo them inside t2i_prompt", "rationale_summary": "같은 공간, 새 순간 — 인물·동작은 이 샷의 instantaneous moment 로 묘사하되 배경은 prior shot 과 일관" }, "atmosphere_reference": { "scope_summary": "different room/floor/angle entirely — only mood/tone is shared with the prior shot", "forbidden_imports": [ "furniture layout from the prior shot", "wall positions from the prior shot", "specific architectural features from the prior shot" ], "required_handling": "describe the new space's background directly in this shot; carry through only color tone or lighting mood (e.g. 'cool gray ambient' / 'dim warm pool')", "rationale_summary": "atmosphere_reference = 분위기 reference 만, 레이아웃 reference 아님. 가구·벽 배치를 끌어오면 모델이 두 공간을 한 공간으로 합성" } }, "view_consistency": { "framing_scope_options": [ "third_person", "close_up" ], "single_camera_rule": "one t2i_prompt = one camera position. third-person full body and close-up of the same character must not be combined in a single prompt.", "third_person_handling": "describe persons with full or partial body in frame; single-camera consistency forbids combining a partial close-up and the same character's full body in one prompt", "close_up_handling": "an isolated framed area of a single subject must not coexist with that subject's full-body description in the same prompt", "focus_on_is_focus_area_not_view_switch": "any framing-intent wording designates the area shown by the single camera position; it does not switch the view", "subject_reference_policy_cross_ref": "id_policy.subject_reference_policy — subjects with policy = 'base_id_required' or 'generic_descriptor_allowed' must not use outlook reference (C##O## form); additionally, subjects with policy = 'generic_descriptor_allowed' must not use bare C## either — render with a common/demographic descriptor only", "prop_orientation_rule": "props the character holds must face one direction only — toward camera OR toward character, not both. forbidden contradiction example: 'photo angled partly toward the camera' + 'eyes fixed on the photograph' (양방향 모순)", "mixing_forbidden": true, "rationale_summary": "single t2i_prompt = single camera shutter at 1/1000s. 시점 혼합은 모델이 두 카메라 합성으로 해석해 인물이 두 번 등장 또는 부위 분리" }, "constraints": [ "fixed_elements are continuity inputs derived from prior shots — treat them as binding context, not as source facts to be invented", "when fixed_elements[i].character_name maps to a C## in id_policy.allowed_base_entity_ids or a C##O## in id_policy.allowed_outlook_pairs, replace the common-noun person reference inside fixed_elements[i].description with that C##/C##O## and do not also write a separate common-noun reference for the same character in the same t2i_prompt", "previous_shot_refs with ref_usage='zoom_in_detail' must preserve the same moment, the same space, and the same lighting as the prior shot — no new persons, props, background elements, poses, or moments", "previous_shot_refs with ref_usage='exact_background' may freely describe this shot's persons, poses, and moment, but must keep background elements consistent with the prior shot's established room", "previous_shot_refs with ref_usage='atmosphere_reference' must describe a new space directly — do not import furniture layout or wall positions from the prior shot, only mood/tone", "one t2i_prompt = one camera position — third-person full body and close-up of the same character must not be combined in a single shot. any framing-intent wording designates the area shown by the single camera position, not a view switch (id_policy.subject_reference_policy applies per-subject)", "props held by characters face one direction only — toward camera OR toward character, never both" ] }, "asset_requirements": { "required_refs": [], "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": "e971335ebc623421" } ], "consumed_bg_catalog_hash": "0336b31251efc584", "consumed_shot_binding_hash": "264862411e91579d" }, "step_id": "scene_detail", "run_id": "7403820e-b9f4-4f78-ad39-f288e7995cf5", "resolved_model": "gemini-pro", "updated_at": "2026-06-03T03:35:20.286625+00:00", "project_config_snapshot": {} }