{
 "groupbg::wooded_crime_site": {
  "input_fingerprint": "6f511b6173c2e638",
  "meta": {
   "model": "gpt-image-2",
   "size": "1536x864",
   "pack": "11.202607220237",
   "contract": "bgfirst_full_v3",
   "group_sig": {
    "key": "wooded_crime_site",
    "tags": [
     "S1sh3",
     "S3sh2"
    ]
   },
   "context_sig": "00800b21cb087083"
  },
  "prompt": "Create ONE empty live-action location background photograph — NO PEOPLE, no figures, no body parts, no silhouettes, no shadows or reflections of people anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods and produce in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch of a shot that happens at this location: use it ONLY as spatial evidence — what this place contains, how its ground, structures and landmarks are arranged and proportioned. Ignore the sketched people and arrows entirely, and do NOT copy its line style: render a fully photographic, physically plausible real place that fits THE LOCATION text below.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nTHE LOCATION — South Korea, 2026; all people are Korean unless stated: A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs.\n\nLOCATION DETAIL (descriptive evidence about this place from the production's location records — evidence, not a staging order): use it to understand what kind of place this is and what it permanently contains. Carry over ONLY its enduring physical features — layout, structures, ground and materials, fixed equipment and installations. Do NOT treat any lighting, weather or time-of-day wording, momentary object states, depicted events or subjective impressions in it as fixed facts of the place; the TIME OF DAY lock below is the sole authority for time and lighting.\n우거진 숲속 및 숲속 사건 현장: 수풀이 빽빽하게 자라난 어두운 산속 지대와 훼손된 흙바닥이 있는 장소 (특징: 빽빽한 나무와 우거진 수풀; 거친 흙바닥; 나무와 풀 사이에 쳐진 노란색 폴리스 라인; 해 질 무렵의 어둑하고 붉은 빛 또는 흐린 낮의 채광)\n\nSCENE EVIDENCE (verbatim quotes from the screenplay about this place — treat them as evidence of what the location physically contains and looks like; stage the PLACE those moments happen in, but do NOT depict the momentary actions, people or staged props themselves):\n- S#1. 숲속 / 해 질 무렵\n- 씬1, 2 현장을 감식하는 경찰, 폴리스 라인을 치고 시신을 수거한다.\n\nTIME OF DAY (lock): dusk.\n\nRender ONE photorealistic empty location photograph, 16:9, neutral enough that every shot of this place can be staged from it later. No text or watermarks.",
  "effective_prompt": "Create ONE empty live-action location background photograph — NO PEOPLE, no figures, no body parts, no silhouettes, no shadows or reflections of people anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods and produce in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch of a shot that happens at this location: use it ONLY as spatial evidence — what this place contains, how its ground, structures and landmarks are arranged and proportioned. Ignore the sketched people and arrows entirely, and do NOT copy its line style: render a fully photographic, physically plausible real place that fits THE LOCATION text below.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nTHE LOCATION — South Korea, 2026; all people are Korean unless stated: A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs.\n\nLOCATION DETAIL (descriptive evidence about this place from the production's location records — evidence, not a staging order): use it to understand what kind of place this is and what it permanently contains. Carry over ONLY its enduring physical features — layout, structures, ground and materials, fixed equipment and installations. Do NOT treat any lighting, weather or time-of-day wording, momentary object states, depicted events or subjective impressions in it as fixed facts of the place; the TIME OF DAY lock below is the sole authority for time and lighting.\n우거진 숲속 및 숲속 사건 현장: 수풀이 빽빽하게 자라난 어두운 산속 지대와 훼손된 흙바닥이 있는 장소 (특징: 빽빽한 나무와 우거진 수풀; 거친 흙바닥; 나무와 풀 사이에 쳐진 노란색 폴리스 라인; 해 질 무렵의 어둑하고 붉은 빛 또는 흐린 낮의 채광)\n\nSCENE EVIDENCE (verbatim quotes from the screenplay about this place — treat them as evidence of what the location physically contains and looks like; stage the PLACE those moments happen in, but do NOT depict the momentary actions, people or staged props themselves):\n- S#1. 숲속 / 해 질 무렵\n- 씬1, 2 현장을 감식하는 경찰, 폴리스 라인을 치고 시신을 수거한다.\n\nTIME OF DAY (lock): dusk.\n\nRender ONE photorealistic empty location photograph, 16:9, neutral enough that every shot of this place can be staged from it later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/groupbg_wooded_crime_site_88435c.png",
  "asset_id": "2d51ffc1-dc78-4c5e-91d4-936ed926791c",
  "input_asset_ids": [
   "10afec5f-f0f7-4691-b81d-3c94a3505e02"
  ],
  "origin_tag": "S1sh3",
  "place_text": "A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs.",
  "origin_inputs": {
   "place_text": "A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs.",
   "time_of_day_en": "dusk",
   "conti_asset_id": "10afec5f-f0f7-4691-b81d-3c94a3505e02"
  }
 },
 "S1sh3::bgfirst_bg": {
  "input_fingerprint": "edc037893720a160",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 허공에 뜬 검은 형상의 존재가 금월산 남성 피해자의 등 위를 짓누르며 닿고, 그 충격으로 피해자의 몸이 앞으로 크게 휘청이는 역동적 충돌 찰나의 측면\n\nLOCATION (lock): A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs.\n\nTIME OF DAY (lock): dusk.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At waist height and several meters to the side, the camera tracks parallel in a broad profile, widening only enough to preserve the changed distance and reveal the entire collision. 금월산 남성 피해자 pitches across the right half of frame while 검은 형상의 존재 descends through the upper-left space onto his back, leaving visible lateral room for the full forward stagger.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: 우거진 수풀 (Dense along the forest route); used as The surrounding growth establishes the escape path behind the profile collision without obstructing either figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Dusk ambient light is treated in a cool desaturated palette with restrained contrast and the victim's blood held as a deep-crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 허공에 뜬 검은 형상의 존재가 금월산 남성 피해자의 등 위를 짓누르며 닿고, 그 충격으로 피해자의 몸이 앞으로 크게 휘청이는 역동적 충돌 찰나의 측면\n\nLOCATION (lock): A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs.\n\nTIME OF DAY (lock): dusk.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At waist height and several meters to the side, the camera tracks parallel in a broad profile, widening only enough to preserve the changed distance and reveal the entire collision. 금월산 남성 피해자 pitches across the right half of frame while 검은 형상의 존재 descends through the upper-left space onto his back, leaving visible lateral room for the full forward stagger.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: 우거진 수풀 (Dense along the forest route); used as The surrounding growth establishes the escape path behind the profile collision without obstructing either figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Dusk ambient light is treated in a cool desaturated palette with restrained contrast and the victim's blood held as a deep-crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S1sh3__bgfirst_bg.png",
  "asset_id": "f043abae-f7e3-4dad-ab02-c4925e58531c",
  "input_asset_ids": [
   "10afec5f-f0f7-4691-b81d-3c94a3505e02",
   "2d51ffc1-dc78-4c5e-91d4-936ed926791c"
  ]
 },
 "S1sh3": {
  "input_fingerprint": "cf2b56afdc2628e8",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 허공에 뜬 검은 형상의 존재가 금월산 남성 피해자의 등 위를 짓누르며 닿고, 그 충격으로 피해자의 몸이 앞으로 크게 휘청이는 역동적 충돌 찰나의 측면\n\nLOCATION (lock): A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At waist height and several meters to the side, the camera tracks parallel in a broad profile, widening only enough to preserve the changed distance and reveal the entire collision. 금월산 남성 피해자 pitches across the right half of frame while 검은 형상의 존재 descends through the upper-left space onto his back, leaving visible lateral room for the full forward stagger.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: 우거진 수풀 (Dense along the forest route); used as The surrounding growth establishes the escape path behind the profile collision without obstructing either figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Dusk ambient light is treated in a cool desaturated palette with restrained contrast and the victim's blood held as a deep-crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The male victim is still missing one shoe, with a large bite wound on his thigh bleeding as the black figure strikes him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 금월산 남성 피해자 (한국인 성인 남성, 검은 머리, 평범한 성인 얼굴); 낯선 한국인 남성 (한국인 성인 남성, 검은 머리, 얼굴 생김새 불명) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 허공에 뜬 검은 형상의 존재가 금월산 남성 피해자의 등 위를 짓누르며 닿고, 그 충격으로 피해자의 몸이 앞으로 크게 휘청이는 역동적 충돌 찰나의 측면\n\nLOCATION (lock): A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At waist height and several meters to the side, the camera tracks parallel in a broad profile, widening only enough to preserve the changed distance and reveal the entire collision. 금월산 남성 피해자 pitches across the right half of frame while 검은 형상의 존재 descends through the upper-left space onto his back, leaving visible lateral room for the full forward stagger.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: 우거진 수풀 (Dense along the forest route); used as The surrounding growth establishes the escape path behind the profile collision without obstructing either figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Dusk ambient light is treated in a cool desaturated palette with restrained contrast and the victim's blood held as a deep-crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The male victim is still missing one shoe, with a large bite wound on his thigh bleeding as the black figure strikes him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 금월산 남성 피해자 (한국인 성인 남성, 검은 머리, 평범한 성인 얼굴); 낯선 한국인 남성 (한국인 성인 남성, 검은 머리, 얼굴 생김새 불명) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 허공에 뜬 검은 형상의 존재가 금월산 남성 피해자의 등 위를 짓누르며 닿고, 그 충격으로 피해자의 몸이 앞으로 크게 휘청이는 역동적 충돌 찰나의 측면\n\nLOCATION (lock): A dense patch of undergrowth within a forest at dusk, where the fleeing victim is overtaken among the shrubs. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At waist height and several meters to the side, the camera tracks parallel in a broad profile, widening only enough to preserve the changed distance and reveal the entire collision. 금월산 남성 피해자 pitches across the right half of frame while 검은 형상의 존재 descends through the upper-left space onto his back, leaving visible lateral room for the full forward stagger.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: 우거진 수풀 (Dense along the forest route); used as The surrounding growth establishes the escape path behind the profile collision without obstructing either figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Dusk ambient light is treated in a cool desaturated palette with restrained contrast and the victim's blood held as a deep-crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The male victim is still missing one shoe, with a large bite wound on his thigh bleeding as the black figure strikes him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 금월산 남성 피해자 (한국인 성인 남성, 검은 머리, 평범한 성인 얼굴); 낯선 한국인 남성 (한국인 성인 남성, 검은 머리, 얼굴 생김새 불명) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S1sh3__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S1sh3.png"
    },
    {
     "label": "CHARACTER REFERENCE — 낯선 한국인 남성: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:851118>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/groupbg_wooded_crime_site_88435c.png"
    },
    {
     "label": "CHARACTER REFERENCE — 낯선 한국인 남성: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:851118>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "'허공에 뜬' 공격자의 공중 자세와 피해자의 벗겨진 한쪽 신발, 허벅지 상처 등 세부 지시를 정확히 구현했습니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "공격자의 발이 땅에 닿아 있어 '허공에 뜬' 지시를 위반했으며, 피해자의 신발 상태도 지시문과 다르게 묘사되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "'허공에 뜬' 공격자의 공중 자세와 피해자의 벗겨진 한쪽 신발, 허벅지 상처 등 세부 지시를 정확히 구현했습니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "공격자의 발이 땅에 닿아 있어 '허공에 뜬' 지시를 위반했으며, 피해자의 신발 상태도 지시문과 다르게 묘사되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "공격자가 허공에 떠 있지 않고 바닥을 딛고 있으며, 피해자가 신발을 양쪽 모두 벗고 있어 상태 지침을 위반했습니다."
     },
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "공격자가 허공에 뜬 채 등을 덮치는 구도를 정확히 연출했으며, 피해자의 한쪽 신발 상실과 허벅지 상처 등 세부 조건을 훌륭하게 반영했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "공격자가 허공에 떠 있지 않고 바닥을 딛고 있으며, 피해자가 신발을 양쪽 모두 벗고 있어 상태 지침을 위반했습니다."
     },
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "공격자가 허공에 뜬 채 등을 덮치는 구도를 정확히 연출했으며, 피해자의 한쪽 신발 상실과 허벅지 상처 등 세부 조건을 훌륭하게 반영했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 15,
     "B": 10
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   }
  },
  "totals": {
   "A": 15,
   "B": 10
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 7,
    "verdict_ko": "'허공에 뜬' 공격자의 공중 자세와 피해자의 벗겨진 한쪽 신발, 허벅지 상처 등 세부 지시를 정확히 구현했습니다."
   },
   {
    "label": "B",
    "score": 5,
    "verdict_ko": "공격자의 발이 땅에 닿아 있어 '허공에 뜬' 지시를 위반했으며, 피해자의 신발 상태도 지시문과 다르게 묘사되었습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/groupbg_wooded_crime_site_88435c.png"
   },
   {
    "label": "CHARACTER REFERENCE — 낯선 한국인 남성: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:851118>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "공중에 떠 있는 남성(낯선 한국인 남성)의 목에 캐릭터 레퍼런스에 있는 검은색 목도리가 누락되었습니다.",
     "fix_en": "Add the black scarf wrapped around the hovering man's neck to match his character reference.",
     "unfixable": false
    },
    {
     "issue_ko": "공중에 떠 있는 남성의 오른팔이 신체에서 완전히 누락되어 있습니다.",
     "fix_en": "Render the hovering man's right arm, showing it extending towards the victim or positioned naturally alongside his body.",
     "unfixable": false
    },
    {
     "issue_ko": "피해자의 맨발인 왼쪽 발의 구조와 발가락이 해부학적으로 심하게 일그러져 있습니다.",
     "fix_en": "Reconstruct the victim's bare left foot to correct the anatomy and render normal human toes.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Shift both figures together to the right while preserving their current scale, contact point, and depth order, placing the victim across the right half and the attacker in the upper-left relative to him.",
     "issue_ko": "충돌하는 두 인물이 레이아웃보다 화면 왼쪽에 치우쳐 있어 피해자가 요구된 오른쪽 절반을 충분히 점유하지 못한다.",
     "unfixable": false
    },
    {
     "fix_en": "Repose the airborne attacker into a near-horizontal descending line from the upper left, with his legs trailing leftward and his upper body extending toward the victim's back, while preserving the hand contact.",
     "issue_ko": "공중의 공격자가 스케치의 거의 수평인 하강 자세가 아니라 다리와 몸통이 아래로 처진 웅크린 대각 자세로 배치되어 움직임 축이 다르다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the black scarf wrapped around the hovering man's neck to match his character reference.\n- Render the hovering man's right arm, showing it extending towards the victim or positioned naturally alongside his body.\n- Reconstruct the victim's bare left foot to correct the anatomy and render normal human toes.\n- Shift both figures together to the right while preserving their current scale, contact point, and depth order, placing the victim across the right half and the attacker in the upper-left relative to him.\n- Repose the airborne attacker into a near-horizontal descending line from the upper left, with his legs trailing leftward and his upper body extending toward the victim's back, while preserving the hand contact.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "피해자가 신발 한 짝을 잃어버린 상태를 정확히 반영하였으며, 역동적인 충돌 순간과 프레이밍 지침을 잘 따랐습니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "공중에 뜬 공격자의 자세는 스케치와 더 유사하나, 피해자가 양발 모두 맨발로 묘사되어 '신발 한 짝을 잃어버림'이라는 상태 지침을 위반했습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "피해자가 신발 한 짝을 잃어버린 상태를 정확히 반영하였으며, 역동적인 충돌 순간과 프레이밍 지침을 잘 따랐습니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "공중에 뜬 공격자의 자세는 스케치와 더 유사하나, 피해자가 양발 모두 맨발로 묘사되어 '신발 한 짝을 잃어버림'이라는 상태 지침을 위반했습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "스케치의 구도와 인물 배치를 충실히 구현했으며, 특히 피해자가 신발을 한 짝만 잃어버렸다는 구체적인 설정을 정확하게 반영하여 가장 우수합니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "전반적인 구도는 유사하나, 피해자가 신발을 한 짝만 잃어버렸다는 지시사항을 무시하고 양발 모두 맨발로 묘사하여 디테일에서 감점되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "스케치의 구도와 인물 배치를 충실히 구현했으며, 특히 피해자가 신발을 한 짝만 잃어버렸다는 구체적인 설정을 정확하게 반영하여 가장 우수합니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "전반적인 구도는 유사하나, 피해자가 신발을 한 짝만 잃어버렸다는 지시사항을 무시하고 양발 모두 맨발로 묘사하여 디테일에서 감점되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 15,
     "B": 9
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "A",
   "fix_won": false,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S1sh3__bgfirst_bg.png",
   "bg_asset_id": "f043abae-f7e3-4dad-ab02-c4925e58531c",
   "bg_record_key": "S1sh3::bgfirst_bg",
   "chain_winner": true,
   "authority": "groupbg",
   "group_key": "wooded_crime_site",
   "groupbg_asset_id": "2d51ffc1-dc78-4c5e-91d4-936ed926791c"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S10sh3::variants": {
  "author_fp": "9aec1dc9818cf0aa",
  "author": {
   "variants": [
    {
     "approach_ko": "정면에 가까운 절제된 대칭 구도와 깊은 초점으로 문틀과 미세한 틈의 물리적 긴장을 또렷하게 포착한다.",
     "prompt_en": "Use a restrained, nearly frontal composition with the verticals held rigid and the narrow separation as the visual axis. Keep the door edge, frame, released locking hardware, and worn metal surfaces crisply legible through a deep plane of focus, while the recess remains tonally dense and unreadable. Let the low-key night illumination skim the exterior metal just enough to define weight, texture, and the minute gap."
    },
    {
     "approach_ko": "얕은 초점과 비스듬한 표면 겹침으로 풀린 문 가장자리와 어둠 사이의 불안한 경계를 강조한다.",
     "prompt_en": "Photograph the doorway with a compressed, intimate lens feel and shallow focus concentrated on the released edge where door and frame nearly meet. Allow the nearer frame surfaces and the darkness beyond to fall gently soft, creating layered planes that draw the eye into the thin recess without revealing it. Preserve subdued night color and soft low-key falloff, with only restrained highlights tracing the angled metal edge."
    }
   ]
  },
  "reused": false
 },
 "S10sh3": {
  "input_fingerprint": "701d0e901fb8a559",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 잠금장치가 풀린 채 미세하게 틈이 벌어진 옥탑방의 어두운 현관문\n\nLOCATION (lock): The exterior rooftop landing immediately before the rooftop dwelling's metal entrance door, its lock released and the door barely ajar. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Directly outside the doorway at waist height, the retreat finishes and briefly locks on the narrow separation between the released iron door and its frame. The door occupies the central midground without exceeding a natural portion of the image, while the dark interior is visible only as a thin vertical recess.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: narrow opening of rooftop room door in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옥탑방 철문 (잠금장치가 풀리고 미세하게 열림) — The outer face is angled slightly away from camera, exposing a thin view of the door edge and opening; used as Primary focal structure; its released edge creates the narrow opening; 문틀 (문을 둘러싼 상태) — Its front edge faces the camera around the slightly opened door; used as The frame outlines the opening and makes the separation immediately readable; 옥탑방 내부 (어두움); used as A narrow background recess beyond the released door.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Night ambient light leaves the doorway in restrained low-key contrast, with the interior held dark.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop-room door remains unlocked and slightly ajar after the stranger forces the latch with the flyer.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 잠금장치가 풀린 채 미세하게 틈이 벌어진 옥탑방의 어두운 현관문\n\nLOCATION (lock): The exterior rooftop landing immediately before the rooftop dwelling's metal entrance door, its lock released and the door barely ajar. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Directly outside the doorway at waist height, the retreat finishes and briefly locks on the narrow separation between the released iron door and its frame. The door occupies the central midground without exceeding a natural portion of the image, while the dark interior is visible only as a thin vertical recess.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: narrow opening of rooftop room door in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옥탑방 철문 (잠금장치가 풀리고 미세하게 열림) — The outer face is angled slightly away from camera, exposing a thin view of the door edge and opening; used as Primary focal structure; its released edge creates the narrow opening; 문틀 (문을 둘러싼 상태) — Its front edge faces the camera around the slightly opened door; used as The frame outlines the opening and makes the separation immediately readable; 옥탑방 내부 (어두움); used as A narrow background recess beyond the released door.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Night ambient light leaves the doorway in restrained low-key contrast, with the interior held dark.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop-room door remains unlocked and slightly ajar after the stranger forces the latch with the flyer.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse a restrained, nearly frontal composition with the verticals held rigid and the narrow separation as the visual axis. Keep the door edge, frame, released locking hardware, and worn metal surfaces crisply legible through a deep plane of focus, while the recess remains tonally dense and unreadable. Let the low-key night illumination skim the exterior metal just enough to define weight, texture, and the minute gap.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 잠금장치가 풀린 채 미세하게 틈이 벌어진 옥탑방의 어두운 현관문\n\nLOCATION (lock): The exterior rooftop landing immediately before the rooftop dwelling's metal entrance door, its lock released and the door barely ajar. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Directly outside the doorway at waist height, the retreat finishes and briefly locks on the narrow separation between the released iron door and its frame. The door occupies the central midground without exceeding a natural portion of the image, while the dark interior is visible only as a thin vertical recess.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: narrow opening of rooftop room door in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옥탑방 철문 (잠금장치가 풀리고 미세하게 열림) — The outer face is angled slightly away from camera, exposing a thin view of the door edge and opening; used as Primary focal structure; its released edge creates the narrow opening; 문틀 (문을 둘러싼 상태) — Its front edge faces the camera around the slightly opened door; used as The frame outlines the opening and makes the separation immediately readable; 옥탑방 내부 (어두움); used as A narrow background recess beyond the released door.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Night ambient light leaves the doorway in restrained low-key contrast, with the interior held dark.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop-room door remains unlocked and slightly ajar after the stranger forces the latch with the flyer.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nPhotograph the doorway with a compressed, intimate lens feel and shallow focus concentrated on the released edge where door and frame nearly meet. Allow the nearer frame surfaces and the darkness beyond to fall gently soft, creating layered planes that draw the eye into the thin recess without revealing it. Preserve subdued night color and soft low-key falloff, with only restrained highlights tracing the angled metal edge.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "A": 4,
   "B": 7
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 4,
    "verdict_ko": "요구된 '미세하게 벌어진 틈' 대신 문이 너무 활짝 열려 있어 핵심 묘사를 위반했습니다."
   },
   {
    "label": "B",
    "score": 7,
    "verdict_ko": "요청된 클로즈업 프레이밍을 잘 따랐으며, 잠금장치가 풀려 미세하게 열린 문의 상태를 정확하게 구현했습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L03B01.png"
   },
   {
    "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/background_chain/seed_bg_rooftop_residence_sel.png"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "문이 너무 많이 열려 있어 '미세하게 틈이 벌어진(barely ajar)' 상태가 아니며 틈새가 너무 넓습니다.",
     "fix_en": "Close the door so that it is barely ajar, leaving only a very thin vertical opening between the door and the frame.",
     "unfixable": false
    },
    {
     "issue_ko": "문의 디자인이 STRUCTURE LOOK 레퍼런스와 일치하지 않습니다 (상단 유리창 패널이 없고 전체가 세로줄 철판으로 덮여 있음).",
     "fix_en": "Replace the vertical metal slats on the upper half of the door with a window panel to match the STRUCTURE LOOK reference.",
     "unfixable": false
    },
    {
     "issue_ko": "문 오른쪽 벽면에 STRUCTURE LOOK 레퍼런스에 있는 창문이 누락되고 빈 벽만 보입니다.",
     "fix_en": "Add the window to the concrete wall on the right side of the door, matching its placement and style in the STRUCTURE LOOK reference.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Rotate the door nearly closed in place so the dark interior reads only as a very thin vertical recess, preserving the camera and all other geometry.",
     "issue_ko": "문과 문틀 사이의 틈이 너무 넓어 내부가 큰 검은 면으로 보이며, 중앙에 ‘미세하게 벌어진 얇은 수직 틈’이어야 한다는 구도와 맞지 않습니다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Close the door so that it is barely ajar, leaving only a very thin vertical opening between the door and the frame.\n- Replace the vertical metal slats on the upper half of the door with a window panel to match the STRUCTURE LOOK reference.\n- Add the window to the concrete wall on the right side of the door, matching its placement and style in the STRUCTURE LOOK reference.\n- Rotate the door nearly closed in place so the dark interior reads only as a very thin vertical recess, preserving the camera and all other geometry.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "구도와 어두운 내부 공간을 정확히 구현했으나, 레퍼런스 2의 문 상단 재질(유리/평면)을 반영하지 못했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "열린 문틈 사이에 또 다른 문짝이 복제되어 구조적으로 불가능한 형태를 만들었고 어두운 내부가 보이지 않습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "구도와 어두운 내부 공간을 정확히 구현했으나, 레퍼런스 2의 문 상단 재질(유리/평면)을 반영하지 못했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "열린 문틈 사이에 또 다른 문짝이 복제되어 구조적으로 불가능한 형태를 만들었고 어두운 내부가 보이지 않습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 3,
      "verdict_ko": "레퍼런스에 없는 유리 패널을 철문에 임의로 생성하고, 벌어진 틈새에 물리적으로 불가능한 구조물을 추가해 심각한 오류가 발생했습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "레퍼런스의 철문 디자인과 재질을 정확하게 반영했으며, 미세하게 열린 틈과 어두운 내부를 요청된 클로즈업 구도와 조명으로 충실히 구현했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "레퍼런스에 없는 유리 패널을 철문에 임의로 생성하고, 벌어진 틈새에 물리적으로 불가능한 구조물을 추가해 심각한 오류가 발생했습니다."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "레퍼런스의 철문 디자인과 재질을 정확하게 반영했으며, 미세하게 열린 틈과 어두운 내부를 요청된 클로즈업 구도와 조명으로 충실히 구현했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 14,
     "B": 6
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "A",
   "fix_won": false,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "플레이트+seed만 (배경 전용)",
  "share_plan": {
   "ref_plan": "background"
  },
  "lane_policy": "ab_select_bypass:bg_only",
  "plate_select": {
   "candidates": {
    "A": "L03B01",
    "B": "L03B02",
    "C": "L03B03"
   },
   "assigned": "L03B03",
   "choice": "Candidate C",
   "confident": true,
   "reason_ko": "원문에서 '옥탑방의 어두운 현관문'을 특정하고 있으므로, 건물 전체를 조망하는 구도(B)보다는 옥탑방 현관문 앞 공간을 비추는 뷰가 적합합니다.",
   "kept": "L03B03"
  }
 },
 "S10sh4::variants": {
  "author_fp": "7766133113cb3ac0",
  "author": {
   "variants": [
    {
     "approach_ko": "정면의 창틀과 겹친 그림자를 또렷한 대칭과 강한 명암으로 압축하는 냉혹한 관찰 시점.",
     "prompt_en": "Use a restrained, normal-lens perspective with the frontal window boundary held in crisp, near-symmetrical balance. Keep the pane and both overlapping cast shadows sharply legible, emphasizing the intruder’s enlarged upper form bearing down into Min-sook’s bent lower shadow. Let the hard interior illumination produce dense silhouettes while the surrounding exterior falls into cool, desaturated darkness, creating a stark, contained tableau."
    },
    {
     "approach_ko": "유리의 물성과 빛 번짐을 살려 공격의 그림자가 창 안에서 숨 막히게 포획된 듯 보이게 하는 시점.",
     "prompt_en": "Photograph the frontal pane with a subtly compressed lens feel, making the overlapping shadows seem trapped just behind the glass while preserving the full window boundary and locked screen positions. Favor the physical surface response of the illuminated glass and worn frame, with restrained halation along the brightest edges but clean separation between the man’s descending shadow and Min-sook’s pinned, bent silhouette. Hold the cool exterior subdued and textural so the warm lit pane becomes the sole visual pressure point."
    }
   ]
  },
  "reused": false
 },
 "S10sh4": {
  "input_fingerprint": "785e611f145c81ac",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 불 켜진 창문 너머로, 낯선 한국인 남성의 거대한 그림자가 강민숙(민숙)의 굽은 그림자를 위에서 덮쳐 누르고 있는 찰나의 실루엣\n\nLOCATION (lock): Inside the rooftop dwelling beside its illuminated window, where the attack appears only as overlapping silhouettes seen through the glass from outside. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The lateral track ends at chest height in a stable frontal alignment with the lit window, placing the overlapping shadows in the center while retaining the window boundary around them. The 낯선 한국인 남성’s enlarged upper shadow descends from above and behind 강민숙’s bent shadow, pinning her silhouette into the lower portion of the pane.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 낯선 한국인 남성 shadow in the middle-center of the frame, background, looks toward 강민숙's shadow; 강민숙 shadow in the lower-center of the frame, background; lit rooftop room window in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옥탑방 창문 (불 켜진 방의 그림자가 비침) — The frontal surface shows the large shadow of the man overlapping 강민숙’s bent shadow; used as The entire assault is perceived as overlapping shadows held within this frontal surface; 창틀 (창문을 둘러싼 상태) — Its front edges face the camera symmetrically around the window; used as The surrounding boundary keeps the shadow event spatially contained.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The lit window isolates the figures as stark silhouettes within an otherwise cool, desaturated night exterior.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The forced rooftop-room door remains slightly open behind the intruder. Min-sook still bears the small circular wrist mark established earlier.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 불 켜진 창문 너머로, 낯선 한국인 남성의 거대한 그림자가 강민숙(민숙)의 굽은 그림자를 위에서 덮쳐 누르고 있는 찰나의 실루엣\n\nLOCATION (lock): Inside the rooftop dwelling beside its illuminated window, where the attack appears only as overlapping silhouettes seen through the glass from outside. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The lateral track ends at chest height in a stable frontal alignment with the lit window, placing the overlapping shadows in the center while retaining the window boundary around them. The 낯선 한국인 남성’s enlarged upper shadow descends from above and behind 강민숙’s bent shadow, pinning her silhouette into the lower portion of the pane.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 낯선 한국인 남성 shadow in the middle-center of the frame, background, looks toward 강민숙's shadow; 강민숙 shadow in the lower-center of the frame, background; lit rooftop room window in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옥탑방 창문 (불 켜진 방의 그림자가 비침) — The frontal surface shows the large shadow of the man overlapping 강민숙’s bent shadow; used as The entire assault is perceived as overlapping shadows held within this frontal surface; 창틀 (창문을 둘러싼 상태) — Its front edges face the camera symmetrically around the window; used as The surrounding boundary keeps the shadow event spatially contained.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The lit window isolates the figures as stark silhouettes within an otherwise cool, desaturated night exterior.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The forced rooftop-room door remains slightly open behind the intruder. Min-sook still bears the small circular wrist mark established earlier.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse a restrained, normal-lens perspective with the frontal window boundary held in crisp, near-symmetrical balance. Keep the pane and both overlapping cast shadows sharply legible, emphasizing the intruder’s enlarged upper form bearing down into Min-sook’s bent lower shadow. Let the hard interior illumination produce dense silhouettes while the surrounding exterior falls into cool, desaturated darkness, creating a stark, contained tableau.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 불 켜진 창문 너머로, 낯선 한국인 남성의 거대한 그림자가 강민숙(민숙)의 굽은 그림자를 위에서 덮쳐 누르고 있는 찰나의 실루엣\n\nLOCATION (lock): Inside the rooftop dwelling beside its illuminated window, where the attack appears only as overlapping silhouettes seen through the glass from outside. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The lateral track ends at chest height in a stable frontal alignment with the lit window, placing the overlapping shadows in the center while retaining the window boundary around them. The 낯선 한국인 남성’s enlarged upper shadow descends from above and behind 강민숙’s bent shadow, pinning her silhouette into the lower portion of the pane.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 낯선 한국인 남성 shadow in the middle-center of the frame, background, looks toward 강민숙's shadow; 강민숙 shadow in the lower-center of the frame, background; lit rooftop room window in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옥탑방 창문 (불 켜진 방의 그림자가 비침) — The frontal surface shows the large shadow of the man overlapping 강민숙’s bent shadow; used as The entire assault is perceived as overlapping shadows held within this frontal surface; 창틀 (창문을 둘러싼 상태) — Its front edges face the camera symmetrically around the window; used as The surrounding boundary keeps the shadow event spatially contained.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The lit window isolates the figures as stark silhouettes within an otherwise cool, desaturated night exterior.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The forced rooftop-room door remains slightly open behind the intruder. Min-sook still bears the small circular wrist mark established earlier.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nPhotograph the frontal pane with a subtly compressed lens feel, making the overlapping shadows seem trapped just behind the glass while preserving the full window boundary and locked screen positions. Favor the physical surface response of the illuminated glass and worn frame, with restrained halation along the brightest edges but clean separation between the man’s descending shadow and Min-sook’s pinned, bent silhouette. Hold the cool exterior subdued and textural so the warm lit pane becomes the sole visual pressure point.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "A": 4,
   "B": 2
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 4,
    "verdict_ko": "미디엄 샷 프레이밍을 어기고 와이드 샷으로 연출했으며 건축물 외관(문 삭제, 비정상적으로 커진 창문)을 위반했으나, 그림자가 창문 안에 맺히는 핵심 연출은 올바르게 구현했습니다."
   },
   {
    "label": "B",
    "score": 2,
    "verdict_ko": "여성의 실루엣이 불투명한 콘크리트 외벽 위로 겹쳐서 나타나는 물리적으로 불가능한 연출(또는 프롬프트가 금지한 야외 인물 배치)을 보여주어 심각한 감점입니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L03B01.png"
   },
   {
    "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/background_chain/seed_bg_rooftop_residence_sel.png"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "카메라 앵글이 창문을 가까이서 잡는 미디엄 샷이 아니라 옥상 전체를 보여주는 익스트림 와이드 샷으로, 레퍼런스 이미지의 구도를 그대로 복사했습니다.",
     "fix_en": "Change the framing to a medium shot tightly framing the lit window and the overlapping shadows.",
     "unfixable": true
    },
    {
     "issue_ko": "건물 전면에 있어야 할 녹색 문이 사라졌고, 창문이 비정상적으로 크게 확대되어 구조물 참조 이미지(STRUCTURE LOOK)의 형태와 비율을 완전히 위반했습니다.",
     "fix_en": "Rebuild the structure's facade to include the standard green door and small sliding window shown in the reference.",
     "unfixable": true
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Crop tightly around the illuminated window to create a medium shot, keeping the complete window boundary visible and the overlapping shadows centered.",
     "issue_ko": "현재 프레임은 옥상 전경이 크게 포함된 와이드 숏으로, 지정된 미디엄 숏보다 창문과 겹친 그림자가 지나치게 작게 보인다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Crop tightly around the illuminated window to create a medium shot, keeping the complete window boundary visible and the overlapping shadows centered.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지정된 미디엄 샷 프레이밍을 어기고 주변 옥탑방 배경을 너무 넓게 보여주는 와이드 샷으로 연출되었습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "미디엄 샷 기준에 맞춰 창문과 두 인물의 겹쳐진 그림자 실루엣에 집중하는 카메라 구도를 적절히 구현했습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지정된 미디엄 샷 프레이밍을 어기고 주변 옥탑방 배경을 너무 넓게 보여주는 와이드 샷으로 연출되었습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "미디엄 샷 기준에 맞춰 창문과 두 인물의 겹쳐진 그림자 실루엣에 집중하는 카메라 구도를 적절히 구현했습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "요구된 미디엄 샷 프레이밍과 창문을 중심에 둔 구도를 정확히 구현하여 가장 높은 우선순위를 충족했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "지시된 미디엄 샷을 무시하고 레퍼런스 이미지의 와이드 샷 구도를 그대로 모방하여 프레이밍 규정을 위반했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "요구된 미디엄 샷 프레이밍과 창문을 중심에 둔 구도를 정확히 구현하여 가장 높은 우선순위를 충족했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지시된 미디엄 샷을 무시하고 레퍼런스 이미지의 와이드 샷 구도를 그대로 모방하여 프레이밍 규정을 위반했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 8,
     "B": 14
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "플레이트+seed만 (배경 전용)",
  "share_plan": {
   "ref_plan": "background"
  },
  "lane_policy": "ab_select_bypass:bg_only",
  "plate_select": {
   "candidates": {
    "A": "L03B01",
    "B": "L03B02",
    "C": "L03B03"
   },
   "assigned": "L03B03",
   "choice": "Candidate C",
   "confident": true,
   "reason_ko": "\"불 켜진 창문 너머로\" 일어나는 사건의 실루엣을 외부에서 관찰하는 옥탑방 밖 앵글이 필요합니다. A와 C가 동일한 구도를 보여주며 모두 조건에 부합하므로, 현재 할당된 C를 유지합니다.",
   "kept": "L03B03"
  }
 },
 "S12sh3::bgfirst_bg": {
  "input_fingerprint": "ed79be5e76fb2985",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 커튼 뒤, 목과 어깨에 살점이 뜯긴 거친 상처를 입고 피를 흘린 채 앉아 고개를 떨군 강민숙(민숙)의 끔찍한 모습\n\nLOCATION (lock): Behind the window curtain inside the daughter's bedroom, in the blood-soaked corner where the mother's body is found seated.\n\nTIME OF DAY (lock): night.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From just behind and beside 수리영’s prior position at shoulder height, the push-in stops on a tight view of 강민숙 seated behind the displaced curtain, keeping her lowered head and the torn neck, shoulder, and collarbone wounds together. Her body occupies most of the center and lower frame, while the curtain edge remains as a narrow foreground border rather than obscuring the injuries.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 강민숙 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 커튼 (젖혀짐) — One displaced edge faces the camera as a narrow foreground border beside 강민숙; used as Its drawn-aside edge frames the reveal and marks the concealed location; 방바닥 (혈흔이 흥건함); used as Visible beneath and around the seated body, grounding her collapsed posture; 수리영 방 (핏자국이 있음); used as Provides limited room context beyond the corpse without distracting from the wounds.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The room’s established faint stand light supports cool desaturated low-key illumination, with blood and exposed wounds retained as deep-crimson accents.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 커튼 뒤, 목과 어깨에 살점이 뜯긴 거친 상처를 입고 피를 흘린 채 앉아 고개를 떨군 강민숙(민숙)의 끔찍한 모습\n\nLOCATION (lock): Behind the window curtain inside the daughter's bedroom, in the blood-soaked corner where the mother's body is found seated.\n\nTIME OF DAY (lock): night.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From just behind and beside 수리영’s prior position at shoulder height, the push-in stops on a tight view of 강민숙 seated behind the displaced curtain, keeping her lowered head and the torn neck, shoulder, and collarbone wounds together. Her body occupies most of the center and lower frame, while the curtain edge remains as a narrow foreground border rather than obscuring the injuries.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 강민숙 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 커튼 (젖혀짐) — One displaced edge faces the camera as a narrow foreground border beside 강민숙; used as Its drawn-aside edge frames the reveal and marks the concealed location; 방바닥 (혈흔이 흥건함); used as Visible beneath and around the seated body, grounding her collapsed posture; 수리영 방 (핏자국이 있음); used as Provides limited room context beyond the corpse without distracting from the wounds.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The room’s established faint stand light supports cool desaturated low-key illumination, with blood and exposed wounds retained as deep-crimson accents.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S12sh3__bgfirst_bg.png",
  "asset_id": "4460cdfd-6be1-4a6d-87c6-aeca7da1cc69",
  "input_asset_ids": [
   "417e2ef9-c6a1-429c-9c18-65d88d1df6f2",
   "90f18abf-c6dc-4e3b-90ce-02626ddca34c"
  ]
 },
 "S12sh3": {
  "input_fingerprint": "bf370040a09b5d7a",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 커튼 뒤, 목과 어깨에 살점이 뜯긴 거친 상처를 입고 피를 흘린 채 앉아 고개를 떨군 강민숙(민숙)의 끔찍한 모습\n\nLOCATION (lock): Behind the window curtain inside the daughter's bedroom, in the blood-soaked corner where the mother's body is found seated. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From just behind and beside 수리영’s prior position at shoulder height, the push-in stops on a tight view of 강민숙 seated behind the displaced curtain, keeping her lowered head and the torn neck, shoulder, and collarbone wounds together. Her body occupies most of the center and lower frame, while the curtain edge remains as a narrow foreground border rather than obscuring the injuries.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 강민숙 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 커튼 (젖혀짐) — One displaced edge faces the camera as a narrow foreground border beside 강민숙; used as Its drawn-aside edge frames the reveal and marks the concealed location; 방바닥 (혈흔이 흥건함); used as Visible beneath and around the seated body, grounding her collapsed posture; 수리영 방 (핏자국이 있음); used as Provides limited room context beyond the corpse without distracting from the wounds.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The room’s established faint stand light supports cool desaturated low-key illumination, with blood and exposed wounds retained as deep-crimson accents.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is seated behind the curtain with her torso slumped forward and her head dropped toward her chest. Her limbs hang limp, with one hand loosely holding a crumpled photograph.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Min-sook’s dead body remains seated behind the curtain with her torso slumped forward and her head dropped toward her chest, her torn neck and shoulder bleeding amid the room’s bloodstains and bloody footprints. Her limbs hang limp, with one hand loosely holding a crumpled photograph, and her wrist still bears the circular mark.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 40대 초반 여성, 부드러운 중년 얼굴, 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 커튼 뒤, 목과 어깨에 살점이 뜯긴 거친 상처를 입고 피를 흘린 채 앉아 고개를 떨군 강민숙(민숙)의 끔찍한 모습\n\nLOCATION (lock): Behind the window curtain inside the daughter's bedroom, in the blood-soaked corner where the mother's body is found seated. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From just behind and beside 수리영’s prior position at shoulder height, the push-in stops on a tight view of 강민숙 seated behind the displaced curtain, keeping her lowered head and the torn neck, shoulder, and collarbone wounds together. Her body occupies most of the center and lower frame, while the curtain edge remains as a narrow foreground border rather than obscuring the injuries.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 강민숙 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 커튼 (젖혀짐) — One displaced edge faces the camera as a narrow foreground border beside 강민숙; used as Its drawn-aside edge frames the reveal and marks the concealed location; 방바닥 (혈흔이 흥건함); used as Visible beneath and around the seated body, grounding her collapsed posture; 수리영 방 (핏자국이 있음); used as Provides limited room context beyond the corpse without distracting from the wounds.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The room’s established faint stand light supports cool desaturated low-key illumination, with blood and exposed wounds retained as deep-crimson accents.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is seated behind the curtain with her torso slumped forward and her head dropped toward her chest. Her limbs hang limp, with one hand loosely holding a crumpled photograph.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Min-sook’s dead body remains seated behind the curtain with her torso slumped forward and her head dropped toward her chest, her torn neck and shoulder bleeding amid the room’s bloodstains and bloody footprints. Her limbs hang limp, with one hand loosely holding a crumpled photograph, and her wrist still bears the circular mark.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 40대 초반 여성, 부드러운 중년 얼굴, 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 커튼 뒤, 목과 어깨에 살점이 뜯긴 거친 상처를 입고 피를 흘린 채 앉아 고개를 떨군 강민숙(민숙)의 끔찍한 모습\n\nLOCATION (lock): Behind the window curtain inside the daughter's bedroom, in the blood-soaked corner where the mother's body is found seated. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From just behind and beside 수리영’s prior position at shoulder height, the push-in stops on a tight view of 강민숙 seated behind the displaced curtain, keeping her lowered head and the torn neck, shoulder, and collarbone wounds together. Her body occupies most of the center and lower frame, while the curtain edge remains as a narrow foreground border rather than obscuring the injuries.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 강민숙 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 커튼 (젖혀짐) — One displaced edge faces the camera as a narrow foreground border beside 강민숙; used as Its drawn-aside edge frames the reveal and marks the concealed location; 방바닥 (혈흔이 흥건함); used as Visible beneath and around the seated body, grounding her collapsed posture; 수리영 방 (핏자국이 있음); used as Provides limited room context beyond the corpse without distracting from the wounds.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The room’s established faint stand light supports cool desaturated low-key illumination, with blood and exposed wounds retained as deep-crimson accents.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is seated behind the curtain with her torso slumped forward and her head dropped toward her chest. Her limbs hang limp, with one hand loosely holding a crumpled photograph.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Min-sook’s dead body remains seated behind the curtain with her torso slumped forward and her head dropped toward her chest, her torn neck and shoulder bleeding amid the room’s bloodstains and bloody footprints. Her limbs hang limp, with one hand loosely holding a crumpled photograph, and her wrist still bears the circular mark.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 40대 초반 여성, 부드러운 중년 얼굴, 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S12sh3__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S12sh3.png"
    },
    {
     "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:899407>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B03.png"
    },
    {
     "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:899407>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "지정된 장소(창문, 혈흔)와 전경 커튼 배치는 따랐으나, 핵심인 목과 어깨의 심각한 상처가 누락되었고 프레이밍이 클로즈업보다 넓습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "장소 록(배경 가구 다름)을 위반했고 상처가 전혀 없으며, 사진을 쥔 손에 힘이 들어가 시체의 중력 지침을 어겼습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "지정된 장소(창문, 혈흔)와 전경 커튼 배치는 따랐으나, 핵심인 목과 어깨의 심각한 상처가 누락되었고 프레이밍이 클로즈업보다 넓습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "장소 록(배경 가구 다름)을 위반했고 상처가 전혀 없으며, 사진을 쥔 손에 힘이 들어가 시체의 중력 지침을 어겼습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "지정된 로케이션 배경, 야간 조명, 그리고 축 늘어진 사망자의 자세는 잘 반영되었으나, 가장 중요한 목과 어깨의 유혈 상처 및 기준 이미지의 의상이 완전히 누락되었습니다."
     },
     {
      "label": "A",
      "score": 2,
      "verdict_ko": "의상은 일치하지만 로케이션이 전혀 다르고, 사진을 든 손이 허공에 떠 있어 중력/부동 상태 규칙(Hard Violation)을 위반했으며 상처와 혈흔도 없습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지정된 로케이션 배경, 야간 조명, 그리고 축 늘어진 사망자의 자세는 잘 반영되었으나, 가장 중요한 목과 어깨의 유혈 상처 및 기준 이미지의 의상이 완전히 누락되었습니다."
     },
     {
      "label": "B",
      "score": 2,
      "verdict_ko": "의상은 일치하지만 로케이션이 전혀 다르고, 사진을 든 손이 허공에 떠 있어 중력/부동 상태 규칙(Hard Violation)을 위반했으며 상처와 혈흔도 없습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 10,
     "B": 6
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   }
  },
  "totals": {
   "A": 10,
   "B": 6
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 6,
    "verdict_ko": "지정된 장소(창문, 혈흔)와 전경 커튼 배치는 따랐으나, 핵심인 목과 어깨의 심각한 상처가 누락되었고 프레이밍이 클로즈업보다 넓습니다."
   },
   {
    "label": "B",
    "score": 4,
    "verdict_ko": "장소 록(배경 가구 다름)을 위반했고 상처가 전혀 없으며, 사진을 쥔 손에 힘이 들어가 시체의 중력 지침을 어겼습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B03.png"
   },
   {
    "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:899407>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "제공된 원본 배경 이미지(카메라 위치, 원근감, 방 구조)가 완전히 무시되고 새로운 공간과 구도로 렌더링되었습니다.",
     "fix_en": "Restore the exact room architecture, perspective, and camera position from the provided background reference image.",
     "unfixable": true
    },
    {
     "issue_ko": "목, 쇄골, 어깨에 살점이 뜯기고 피를 흘리는 끔찍한 상처가 없고, 단순히 옷에 구멍만 나 있습니다.",
     "fix_en": "Add severe, bleeding, torn flesh wounds to the exposed neck, collarbone, and shoulder area.",
     "unfixable": false
    },
    {
     "issue_ko": "캐릭터 의상이 레퍼런스(베이지색 반팔, 회색 트레이닝복)와 달리 검은색 긴팔 옷으로 잘못 묘사되었습니다.",
     "fix_en": "Change the character's clothing to a beige short-sleeve t-shirt and grey sweatpants.",
     "unfixable": false
    },
    {
     "issue_ko": "프롬프트나 스케치에 명시되지 않은 붉은색 담요가 캐릭터 밑에 추가되었습니다.",
     "fix_en": "Remove the red blanket from beneath and around the character.",
     "unfixable": false
    },
    {
     "issue_ko": "한 손으로만 사진을 느슨하게 쥐고 축 늘어져야 한다는 지침과 달리, 두 손을 모아 사진을 쥐고 있습니다.",
     "fix_en": "Reposition the hands so the right arm hangs entirely limp and only the left hand loosely rests with the photograph.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "issue_ko": "카메라가 지정된 배경 사진과 반대쪽에 가까운 방향에서 복도 쪽을 바라보고 있어, 계약의 고정된 카메라 축과 원근이 근본적으로 다릅니다.",
     "fix_en": "Retain the current viewpoint and geometry, but minimize the contradictory hallway context with the tightest possible crop around Min-sook.",
     "unfixable": true
    },
    {
     "issue_ko": "프레이밍이 클로즈업보다 훨씬 넓고 민숙이 화면 왼쪽에 치우쳐 있어, 몸이 중앙과 하단 대부분을 차지해야 한다는 구도를 따르지 않습니다.",
     "fix_en": "Crop in tightly and reframe Min-sook at middle-center, keeping her lowered head, neck, shoulders, collarbone, limp hands, and a small area of floor together in frame.",
     "unfixable": false
    },
    {
     "issue_ko": "왼쪽 커튼이 넓은 전경 덩어리로 화면을 크게 차지해, 민숙 옆의 좁은 전경 테두리여야 한다는 배치와 다릅니다.",
     "fix_en": "Reduce the visible curtain to a narrow strip along the left frame edge without covering Min-sook’s head, neck, or shoulders.",
     "unfixable": false
    },
    {
     "issue_ko": "민숙이 바닥에 주저앉아 주변 바닥으로 자세가 고정된 배치가 아니라 높아진 침구 위에 앉은 듯 보여, 계약의 신체 깊이와 지지 관계가 다릅니다.",
     "fix_en": "Lower Min-sook’s seated body onto the visible floor behind the curtain and let her torso and limbs settle fully with gravity while preserving the existing viewpoint.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add severe, bleeding, torn flesh wounds to the exposed neck, collarbone, and shoulder area.\n- Change the character's clothing to a beige short-sleeve t-shirt and grey sweatpants.\n- Remove the red blanket from beneath and around the character.\n- Reposition the hands so the right arm hangs entirely limp and only the left hand loosely rests with the photograph.\n- Crop in tightly and reframe Min-sook at middle-center, keeping her lowered head, neck, shoulders, collarbone, limp hands, and a small area of floor together in frame.\n- Reduce the visible curtain to a narrow strip along the left frame edge without covering Min-sook’s head, neck, or shoulders.\n- Lower Min-sook’s seated body onto the visible floor behind the curtain and let her torso and limbs settle fully with gravity while preserving the existing viewpoint.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "의상은 레퍼런스와 완벽하게 일치하나, 프롬프트에서 요구한 목과 어깨의 심각한 상처 및 유혈 묘사가 누락되었으며 커튼의 위치가 스케치와 반대로 배치되었습니다."
     },
     {
      "label": "A",
      "score": 3,
      "verdict_ko": "지정된 의상(베이지 티셔츠, 회색 바지)을 무시하고 검은 옷을 입혔으며, 상처 묘사가 없고 배경과 커튼 위치 또한 프롬프트와 스케치에 어긋납니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "의상은 레퍼런스와 완벽하게 일치하나, 프롬프트에서 요구한 목과 어깨의 심각한 상처 및 유혈 묘사가 누락되었으며 커튼의 위치가 스케치와 반대로 배치되었습니다."
     },
     {
      "label": "A",
      "score": 3,
      "verdict_ko": "지정된 의상(베이지 티셔츠, 회색 바지)을 무시하고 검은 옷을 입혔으며, 상처 묘사가 없고 배경과 커튼 위치 또한 프롬프트와 스케치에 어긋납니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지정된 의상과 구도, 배경은 대체로 준수했으나, 숏 텍스트의 가장 핵심인 목과 어깨의 끔찍한 상처와 유혈 묘사가 완전히 누락되었습니다."
     },
     {
      "label": "B",
      "score": 2,
      "verdict_ko": "핵심 상처 묘사가 누락되었을 뿐만 아니라, 캐릭터의 의상이 완전히 다르고 프롬프트에 없는 붉은 담요가 추가되어(하드 위반) 감점 요소가 큽니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "지정된 의상과 구도, 배경은 대체로 준수했으나, 숏 텍스트의 가장 핵심인 목과 어깨의 끔찍한 상처와 유혈 묘사가 완전히 누락되었습니다."
     },
     {
      "label": "A",
      "score": 2,
      "verdict_ko": "핵심 상처 묘사가 누락되었을 뿐만 아니라, 캐릭터의 의상이 완전히 다르고 프롬프트에 없는 붉은 담요가 추가되어(하드 위반) 감점 요소가 큽니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 5,
     "B": 8
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S12sh3__bgfirst_bg.png",
   "bg_asset_id": "4460cdfd-6be1-4a6d-87c6-aeca7da1cc69",
   "bg_record_key": "S12sh3::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L05B01",
    "B": "L05B02",
    "C": "L05B03",
    "D": "L05B04",
    "E": "L05B05",
    "F": "L05B06"
   },
   "assigned": "L05B05",
   "choice": "Candidate E",
   "confident": true,
   "reason_ko": "지문에서 '커튼 뒤'라는 공간적 배경을 명시하고 있으므로, 커튼과 창문이 가장 잘 보이고 그 주변 공간을 비추는 현재 배정된 이미지가 적합합니다.",
   "kept": "L05B05"
  }
 },
 "S14sh3::variants": {
  "author_fp": "bfac6480b0b9d37a",
  "author": {
   "variants": [
    {
     "approach_ko": "문턱의 정면적이고 절제된 와이드 시점에서 방 전체를 깊은 초점으로 기록해 비정상적인 정돈과 흔적의 완전한 소멸을 동시에 강조한다.",
     "prompt_en": "Use a restrained, rectilinear wide perspective with deep focus from the standing threshold viewpoint. Keep the tidy bed and spotless wall clearly legible at middle-left while the opened curtain and vacant floor remain equally sharp at middle-right, creating a calm, forensic tableau. Let the cool desaturated ambient light spread evenly across the physical surfaces with minimal shadow drama, making the room’s excessive order and complete lack of evidence feel unnervingly objective."
    },
    {
     "approach_ko": "팬이 멈춘 커튼 뒤 빈 공간에 시선을 고정하고 침대 쪽은 주변부의 부드러운 층으로 남겨, 부재 자체가 장면의 종착점이 되게 한다.",
     "prompt_en": "Frame the held endpoint of the pan with the empty space behind the opened curtain as the strongest visual destination, while the bed area remains offset at middle-left as a slightly softer but fully recognizable peripheral layer. Use a natural wide-lens depth progression rather than a flat survey, with crisp emphasis settling into the vacant background and uninterrupted clean floor. Shape the cool ambient illumination with gentle falloff across the bed side and clearer separation at the curtain edge, turning the revealed absence into the shot’s quiet dramatic center."
    }
   ]
  },
  "reused": false
 },
 "S14sh3": {
  "input_fingerprint": "5229d49fb0ed97aa",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 핏자국 하나 없이 깨끗하게 정돈된 방 안과 텅 빈 커튼 뒤 공간\n\nLOCATION (lock): Inside the daughter's unnaturally tidy bedroom, including the now-empty space behind the curtain. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From standing eye level on the threshold, a wide pan sweeps across the bed area and completes at the cleared curtain, then holds on the empty space behind it. The bed remains offset on one side while the opened curtain and vacant floor occupy the other, allowing the excessive order and total absence of blood, body, or wall mark to read in one coherent layout.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: bed area without red mark in the middle-left of the frame, midground; empty space behind curtain in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 침대 (정돈됨) — Its side and head area are visible from the threshold at an oblique angle; used as One side of the wide room layout; the wall above it confirms the red mark has vanished; 커튼 (뒤 공간이 드러남) — The curtain is displaced enough for the camera to see fully into the space behind it; used as Its cleared edge reveals the vacant area that previously concealed 강민숙; 커튼 뒤 공간 (비어 있음); used as Final focal area of the pan, held long enough to establish the body’s absence; 침대 위 벽면 (붉은 표식 없음) — The wall face is visible to camera and carries no red mark; used as The front-facing surface above the bed verifies that no supernatural mark remains; 방바닥 (핏자국 없이 깨끗함); used as Its uninterrupted visibility establishes the disappearance of all blood evidence.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Interior ambient light is treated in a cool desaturated palette with restrained contrast, emphasizing the room’s unnaturally complete visual order.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room is now unnaturally clean: the blood, red wall circle, and Min-sook’s corpse have all disappeared, leaving the space behind the curtain empty. Su-ri-yeong still has the photograph she took from Min-sook.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 핏자국 하나 없이 깨끗하게 정돈된 방 안과 텅 빈 커튼 뒤 공간\n\nLOCATION (lock): Inside the daughter's unnaturally tidy bedroom, including the now-empty space behind the curtain. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From standing eye level on the threshold, a wide pan sweeps across the bed area and completes at the cleared curtain, then holds on the empty space behind it. The bed remains offset on one side while the opened curtain and vacant floor occupy the other, allowing the excessive order and total absence of blood, body, or wall mark to read in one coherent layout.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: bed area without red mark in the middle-left of the frame, midground; empty space behind curtain in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 침대 (정돈됨) — Its side and head area are visible from the threshold at an oblique angle; used as One side of the wide room layout; the wall above it confirms the red mark has vanished; 커튼 (뒤 공간이 드러남) — The curtain is displaced enough for the camera to see fully into the space behind it; used as Its cleared edge reveals the vacant area that previously concealed 강민숙; 커튼 뒤 공간 (비어 있음); used as Final focal area of the pan, held long enough to establish the body’s absence; 침대 위 벽면 (붉은 표식 없음) — The wall face is visible to camera and carries no red mark; used as The front-facing surface above the bed verifies that no supernatural mark remains; 방바닥 (핏자국 없이 깨끗함); used as Its uninterrupted visibility establishes the disappearance of all blood evidence.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Interior ambient light is treated in a cool desaturated palette with restrained contrast, emphasizing the room’s unnaturally complete visual order.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room is now unnaturally clean: the blood, red wall circle, and Min-sook’s corpse have all disappeared, leaving the space behind the curtain empty. Su-ri-yeong still has the photograph she took from Min-sook.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse a restrained, rectilinear wide perspective with deep focus from the standing threshold viewpoint. Keep the tidy bed and spotless wall clearly legible at middle-left while the opened curtain and vacant floor remain equally sharp at middle-right, creating a calm, forensic tableau. Let the cool desaturated ambient light spread evenly across the physical surfaces with minimal shadow drama, making the room’s excessive order and complete lack of evidence feel unnervingly objective.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 핏자국 하나 없이 깨끗하게 정돈된 방 안과 텅 빈 커튼 뒤 공간\n\nLOCATION (lock): Inside the daughter's unnaturally tidy bedroom, including the now-empty space behind the curtain. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From standing eye level on the threshold, a wide pan sweeps across the bed area and completes at the cleared curtain, then holds on the empty space behind it. The bed remains offset on one side while the opened curtain and vacant floor occupy the other, allowing the excessive order and total absence of blood, body, or wall mark to read in one coherent layout.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: bed area without red mark in the middle-left of the frame, midground; empty space behind curtain in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 침대 (정돈됨) — Its side and head area are visible from the threshold at an oblique angle; used as One side of the wide room layout; the wall above it confirms the red mark has vanished; 커튼 (뒤 공간이 드러남) — The curtain is displaced enough for the camera to see fully into the space behind it; used as Its cleared edge reveals the vacant area that previously concealed 강민숙; 커튼 뒤 공간 (비어 있음); used as Final focal area of the pan, held long enough to establish the body’s absence; 침대 위 벽면 (붉은 표식 없음) — The wall face is visible to camera and carries no red mark; used as The front-facing surface above the bed verifies that no supernatural mark remains; 방바닥 (핏자국 없이 깨끗함); used as Its uninterrupted visibility establishes the disappearance of all blood evidence.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Interior ambient light is treated in a cool desaturated palette with restrained contrast, emphasizing the room’s unnaturally complete visual order.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room is now unnaturally clean: the blood, red wall circle, and Min-sook’s corpse have all disappeared, leaving the space behind the curtain empty. Su-ri-yeong still has the photograph she took from Min-sook.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nFrame the held endpoint of the pan with the empty space behind the opened curtain as the strongest visual destination, while the bed area remains offset at middle-left as a slightly softer but fully recognizable peripheral layer. Use a natural wide-lens depth progression rather than a flat survey, with crisp emphasis settling into the vacant background and uninterrupted clean floor. Shape the cool ambient illumination with gentle falloff across the bed side and clearer separation at the curtain edge, turning the revealed absence into the shot’s quiet dramatic center.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "A": 4,
   "B": 7
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 4,
    "verdict_ko": "커튼 뒤의 빈 공간을 명확히 드러내지 않고 커튼이 닫혀 있어 프롬프트의 핵심 구성을 반영하지 못했습니다."
   },
   {
    "label": "B",
    "score": 7,
    "verdict_ko": "문지방 시점에서 좌측의 정돈된 침대와 우측의 걷힌 커튼, 그리고 그 뒤의 빈 공간을 지시된 프레이밍에 맞게 정확히 구현했습니다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S12sh3_sel.png"
   }
  ],
  "critique": {
   "issues": []
  },
  "composition_critique": {
   "issues": []
  },
  "fix_skipped": true,
  "ref_mode": "prev만 (배경 전용·공유 계획)",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S12sh3"
  },
  "lane_policy": "share_plan_prev_bgonly",
  "plate_select": {
   "candidates": {
    "A": "L05B01",
    "B": "L05B02",
    "C": "L05B03",
    "D": "L05B04",
    "E": "L05B05",
    "F": "L05B06"
   },
   "assigned": "L05B06",
   "choice": "Candidate C",
   "confident": true,
   "reason_ko": "침대가 있는 깨끗한 방 내부와 우측의 커튼(커튼 뒤 공간)이 함께 보이는 앵글이므로 이 서브공간이 적합합니다.",
   "kept": "L05B06"
  }
 },
 "S14sh4::variants": {
  "author_fp": "a692909a08f6368c",
  "author": {
   "variants": [
    {
     "approach_ko": "가까운 광각 감각으로 수리영의 창백한 표정과 그 너머 방의 깊고 불길한 공백을 함께 체감하게 한다.",
     "prompt_en": "Use an intimate, gently wide close-up feel at her eye level, holding her front three-quarter face dominant in the left-center while the bed’s receding long side leads into the empty right side of the room. Keep her eyes and pallid facial tension crisply present, with shallow but gradual focus allowing the ordered bed, bare curtain area, and immaculate negative space to remain softly intelligible. Let the cool ambient night light wrap delicately across her face and fall away into the subdued room, emphasizing stunned stillness and spatial absence."
    },
    {
     "approach_ko": "압축된 망원 감각과 절제된 측면 명암으로 수리영의 얼굴 뒤에 깨끗한 빈방이 차갑게 눌러붙는 구도를 만든다.",
     "prompt_en": "Use a restrained telephoto close-up feel from the fixed eye-level front three-quarter position, compressing the pale face against the meticulously clean room without enlarging any background element. Frame the face as a still, sculptural mass in the middle-left, with her fixed gaze carrying directly into the softly resolved bed and empty curtain boundary on the right. Shape the existing cool night ambience into subtle lateral contrast across her features, leaving the background low in saturation and quiet detail so its impossible cleanliness presses against her expression."
    }
   ]
  },
  "reused": false
 },
 "S14sh4": {
  "input_fingerprint": "ccc3dafe6cf9d0a2",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 텅 빈 방을 멍하니 응시하는 수리영의 창백해진 얼굴\n\nLOCATION (lock): Inside the daughter's cleaned and empty bedroom, facing the curtain where the body had appeared. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From inside the room at 수리영's eye level, finish the dolly-in on her front three-quarter side, letting her pale face occupy most of the left-center frame while the meticulously ordered, empty room remains legible beyond her fixed eyeline. Hold the close composition just before moving alongside her, with shallow focus preserving the absence behind her rather than isolating her as a portrait.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward empty room interior; empty room interior in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bed (Meticulously ordered, with neither the red mark nor blood visible) — Its long side recedes behind 수리영 and the upper surface remains visible to camera; used as Softly resolved background evidence of the impossible disappearance; curtain (Present, with no body visible around it); used as Background boundary associated with the vanished body; empty room interior (Unnaturally clean and empty of the previously seen traces); used as Negative space held beyond 수리영's eyeline.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained ambient night interior light supports a cool, desaturated palette and keeps the empty room subdued behind 수리영's pallor.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same spotless bedroom, empty curtain area, restrained nighttime lighting, and the woman's position facing into the room. Exclude any corpse, blood, red wall mark, or displaced objects, and crop the composition around her pale face while retaining enough of the clean room behind her.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room remains excessively clean and empty, with no blood, red circle, or corpse behind the curtain. Su-ri-yeong retains the photograph taken from Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 텅 빈 방을 멍하니 응시하는 수리영의 창백해진 얼굴\n\nLOCATION (lock): Inside the daughter's cleaned and empty bedroom, facing the curtain where the body had appeared. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From inside the room at 수리영's eye level, finish the dolly-in on her front three-quarter side, letting her pale face occupy most of the left-center frame while the meticulously ordered, empty room remains legible beyond her fixed eyeline. Hold the close composition just before moving alongside her, with shallow focus preserving the absence behind her rather than isolating her as a portrait.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward empty room interior; empty room interior in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bed (Meticulously ordered, with neither the red mark nor blood visible) — Its long side recedes behind 수리영 and the upper surface remains visible to camera; used as Softly resolved background evidence of the impossible disappearance; curtain (Present, with no body visible around it); used as Background boundary associated with the vanished body; empty room interior (Unnaturally clean and empty of the previously seen traces); used as Negative space held beyond 수리영's eyeline.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained ambient night interior light supports a cool, desaturated palette and keeps the empty room subdued behind 수리영's pallor.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same spotless bedroom, empty curtain area, restrained nighttime lighting, and the woman's position facing into the room. Exclude any corpse, blood, red wall mark, or displaced objects, and crop the composition around her pale face while retaining enough of the clean room behind her.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room remains excessively clean and empty, with no blood, red circle, or corpse behind the curtain. Su-ri-yeong retains the photograph taken from Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse an intimate, gently wide close-up feel at her eye level, holding her front three-quarter face dominant in the left-center while the bed’s receding long side leads into the empty right side of the room. Keep her eyes and pallid facial tension crisply present, with shallow but gradual focus allowing the ordered bed, bare curtain area, and immaculate negative space to remain softly intelligible. Let the cool ambient night light wrap delicately across her face and fall away into the subdued room, emphasizing stunned stillness and spatial absence.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 텅 빈 방을 멍하니 응시하는 수리영의 창백해진 얼굴\n\nLOCATION (lock): Inside the daughter's cleaned and empty bedroom, facing the curtain where the body had appeared. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From inside the room at 수리영's eye level, finish the dolly-in on her front three-quarter side, letting her pale face occupy most of the left-center frame while the meticulously ordered, empty room remains legible beyond her fixed eyeline. Hold the close composition just before moving alongside her, with shallow focus preserving the absence behind her rather than isolating her as a portrait.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward empty room interior; empty room interior in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bed (Meticulously ordered, with neither the red mark nor blood visible) — Its long side recedes behind 수리영 and the upper surface remains visible to camera; used as Softly resolved background evidence of the impossible disappearance; curtain (Present, with no body visible around it); used as Background boundary associated with the vanished body; empty room interior (Unnaturally clean and empty of the previously seen traces); used as Negative space held beyond 수리영's eyeline.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained ambient night interior light supports a cool, desaturated palette and keeps the empty room subdued behind 수리영's pallor.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same spotless bedroom, empty curtain area, restrained nighttime lighting, and the woman's position facing into the room. Exclude any corpse, blood, red wall mark, or displaced objects, and crop the composition around her pale face while retaining enough of the clean room behind her.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room remains excessively clean and empty, with no blood, red circle, or corpse behind the curtain. Su-ri-yeong retains the photograph taken from Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse a restrained telephoto close-up feel from the fixed eye-level front three-quarter position, compressing the pale face against the meticulously clean room without enlarging any background element. Frame the face as a still, sculptural mass in the middle-left, with her fixed gaze carrying directly into the softly resolved bed and empty curtain boundary on the right. Shape the existing cool night ambience into subtle lateral contrast across her features, leaving the background low in saturation and quiet detail so its impossible cleanliness presses against her expression.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "A": 7,
   "B": 4
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 7,
    "verdict_ko": "요구된 클로즈업 샷 크기와 인물 배치를 정확히 구현하였고 공간 구조(침대 위치 등)가 레퍼런스와 일치하나, 의상이 레퍼런스와 다릅니다."
   },
   {
    "label": "B",
    "score": 4,
    "verdict_ko": "의상은 레퍼런스와 일치하나 샷 크기가 너무 넓게 잡혔고(무릎까지 보임), 방의 구조(침대 위치)가 좌우 반전되어 공간 규칙을 위반했습니다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S14sh3_sel.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "수리영이 입고 있는 옷이 캐릭터 레퍼런스와 다릅니다(남색 후드집업 누락).",
     "fix_en": "Add the navy blue zip-up hoodie over her shoulders and visible chest area to match the character reference.",
     "unfixable": false
    },
    {
     "issue_ko": "배경의 침구가 이전 장면 레퍼런스나 프롬프트의 지시('meticulously ordered')와 달리 헝클어져서 침대 밖으로 늘어져 있습니다.",
     "fix_en": "Tidy the bed covers in the background so they are meticulously flat and neatly ordered on the mattress, matching the reference image.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": []
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the navy blue zip-up hoodie over her shoulders and visible chest area to match the character reference.\n- Tidy the bed covers in the background so they are meticulously flat and neatly ordered on the mattress, matching the reference image.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "지정된 클로즈업 구도와 창백한 표정을 완벽하게 구현했으며, 레퍼런스의 남색 후드 집업 의상도 정확히 유지했습니다."
     },
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "구도와 표정 연기는 우수하나, 레퍼런스에 명시된 캐릭터의 겉옷(남색 후드)이 누락되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "지정된 클로즈업 구도와 창백한 표정을 완벽하게 구현했으며, 레퍼런스의 남색 후드 집업 의상도 정확히 유지했습니다."
     },
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "구도와 표정 연기는 우수하나, 레퍼런스에 명시된 캐릭터의 겉옷(남색 후드)이 누락되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지정된 클로즈업 프레이밍과 텅 빈 방의 배경을 잘 살렸으며, 레퍼런스의 의상(파란색 후드)을 정확히 유지했습니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "프레이밍과 배경 묘사는 지시사항에 부합하나, 인물 레퍼런스에 있는 겉옷(파란색 후드)이 누락되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "지정된 클로즈업 프레이밍과 텅 빈 방의 배경을 잘 살렸으며, 레퍼런스의 의상(파란색 후드)을 정확히 유지했습니다."
     },
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "프레이밍과 배경 묘사는 지시사항에 부합하나, 인물 레퍼런스에 있는 겉옷(파란색 후드)이 누락되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 14,
     "B": 17
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S14sh3"
  }
 },
 "S14sh8::variants": {
  "author_fp": "b57970a360f44b7d",
  "author": {
   "variants": [
    {
     "approach_ko": "얕은 심도로 두 손과 긴장된 어깨의 접촉을 중심에 모으는 친밀한 정면 축 구도.",
     "prompt_en": "Use an intimate, gently compressed lens feel with very shallow depth of field, holding the shared axis steady at upper-torso height. Make Hye-su’s two hands closing softly around Su-ri-yeong’s tense shoulders the crisp visual anchor in the lower third, while their faces and the clean room recede subtly from that tactile point. Shape the neutral interior ambience into soft lateral modeling across fingers, fabric creases, and shoulder tension, with restrained contrast and cool desaturated color."
    },
    {
     "approach_ko": "조금 더 넓은 공간감과 비대칭 레이어링으로 포옹의 고립감과 빈 방의 정적을 함께 살리는 구도.",
     "prompt_en": "Use a more open, natural-perspective lens feel while preserving the completed dolly-in distance, medium scale, and upper-torso camera height. Arrange the partially back-to-camera Su-ri-yeong as a near layer and Hye-su’s upper torso immediately beyond her, with the clasping hands forming a strong low-third bridge between them; retain enough depth for the orderly interior to remain quietly legible without competing. Let broad neutral ambient light fall evenly through the layered figures, adding only delicate edge separation around hands and shoulders in a cool, muted, low-contrast finish."
    }
   ]
  },
  "reused": false
 },
 "S14sh8": {
  "input_fingerprint": "c8696a9e63e2beda",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 수리영의 어깨를 두 손으로 부드럽게 감싸 쥔 혜수의 상체\n\nLOCATION (lock): Inside the rooftop dwelling near the daughter's bedroom and living area, after the rooms have been searched and found empty. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At upper-torso height and one arm's length away, complete the gentle dolly-in toward the shared axis, framing 혜수 from the upper torso while both hands close around 수리영's shoulders in the lower third. Keep 수리영 partially back-to-camera so the physical reassurance, rather than either face alone, becomes the compositional center before the downward tilt begins.\n- FRAMING SCALE: medium shot\n- KEY BACKGROUND ELEMENTS: room interior (Clean and orderly); used as Muted spatial context behind the consoling figures.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient interior light is shaped with restrained contrast and a cool, desaturated finish, keeping attention on the hands and tense shoulders.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop room and dining table remain completely cleaned, with the corpse, blood, wall mark, and two teacups gone. Su-ri-yeong still possesses the photograph removed from her mother’s hand.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리); 혜수 (한국인 40대 중반 여성, 성숙한 타원형 얼굴, 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 수리영의 어깨를 두 손으로 부드럽게 감싸 쥔 혜수의 상체\n\nLOCATION (lock): Inside the rooftop dwelling near the daughter's bedroom and living area, after the rooms have been searched and found empty. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At upper-torso height and one arm's length away, complete the gentle dolly-in toward the shared axis, framing 혜수 from the upper torso while both hands close around 수리영's shoulders in the lower third. Keep 수리영 partially back-to-camera so the physical reassurance, rather than either face alone, becomes the compositional center before the downward tilt begins.\n- FRAMING SCALE: medium shot\n- KEY BACKGROUND ELEMENTS: room interior (Clean and orderly); used as Muted spatial context behind the consoling figures.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient interior light is shaped with restrained contrast and a cool, desaturated finish, keeping attention on the hands and tense shoulders.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop room and dining table remain completely cleaned, with the corpse, blood, wall mark, and two teacups gone. Su-ri-yeong still possesses the photograph removed from her mother’s hand.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리); 혜수 (한국인 40대 중반 여성, 성숙한 타원형 얼굴, 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse an intimate, gently compressed lens feel with very shallow depth of field, holding the shared axis steady at upper-torso height. Make Hye-su’s two hands closing softly around Su-ri-yeong’s tense shoulders the crisp visual anchor in the lower third, while their faces and the clean room recede subtly from that tactile point. Shape the neutral interior ambience into soft lateral modeling across fingers, fabric creases, and shoulder tension, with restrained contrast and cool desaturated color.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 수리영의 어깨를 두 손으로 부드럽게 감싸 쥔 혜수의 상체\n\nLOCATION (lock): Inside the rooftop dwelling near the daughter's bedroom and living area, after the rooms have been searched and found empty. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At upper-torso height and one arm's length away, complete the gentle dolly-in toward the shared axis, framing 혜수 from the upper torso while both hands close around 수리영's shoulders in the lower third. Keep 수리영 partially back-to-camera so the physical reassurance, rather than either face alone, becomes the compositional center before the downward tilt begins.\n- FRAMING SCALE: medium shot\n- KEY BACKGROUND ELEMENTS: room interior (Clean and orderly); used as Muted spatial context behind the consoling figures.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient interior light is shaped with restrained contrast and a cool, desaturated finish, keeping attention on the hands and tense shoulders.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop room and dining table remain completely cleaned, with the corpse, blood, wall mark, and two teacups gone. Su-ri-yeong still possesses the photograph removed from her mother’s hand.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리); 혜수 (한국인 40대 중반 여성, 성숙한 타원형 얼굴, 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse a more open, natural-perspective lens feel while preserving the completed dolly-in distance, medium scale, and upper-torso camera height. Arrange the partially back-to-camera Su-ri-yeong as a near layer and Hye-su’s upper torso immediately beyond her, with the clasping hands forming a strong low-third bridge between them; retain enough depth for the orderly interior to remain quietly legible without competing. Let broad neutral ambient light fall evenly through the layered figures, adding only delicate edge separation around hands and shoulders in a cool, muted, low-contrast finish.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "B": 7,
   "A": 4
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 7,
    "verdict_ko": "혜수의 가죽 재킷 복장을 정확히 유지했으며, 두 손으로 수리영의 어깨를 감싸는 프레이밍과 후면 구도를 지시대로 잘 구현했습니다."
   },
   {
    "label": "A",
    "score": 4,
    "verdict_ko": "혜수의 가죽 재킷이 누락되었으며, 텍스트에 명시된 '두 손' 중 한 손만 화면에 보여 연출 지시를 온전히 충족하지 못했습니다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S14sh4_sel.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   },
   {
    "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:887414>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "수리영의 왼쪽 어깨(화면 좌측)를 감싸 쥔 혜수의 오른손 손가락이 앞쪽으로 5개나 보여 해부학적으로 기형적입니다.",
     "fix_en": "Correct the anatomy of the hand on the left shoulder by removing the extra bottom finger, leaving exactly four fingers wrapping over the front.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": []
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Correct the anatomy of the hand on the left shoulder by removing the extra bottom finger, leaving exactly four fingers wrapping over the front.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지정된 미디엄 샷 구도, 인물의 배치 및 의상, 배경 조건을 모두 충실히 반영했습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "A와 시각적으로 동일한 이미지로, 프롬프트의 요구 사항(구도, 인물, 배경)을 잘 충족합니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지정된 미디엄 샷 구도, 인물의 배치 및 의상, 배경 조건을 모두 충실히 반영했습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "A와 시각적으로 동일한 이미지로, 프롬프트의 요구 사항(구도, 인물, 배경)을 잘 충족합니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 10,
      "verdict_ko": "요구된 미디엄 샷 스케일과 구도를 정확히 반영했으며, 어깨를 감싸는 손의 형태가 해부학적 오류 없이 자연스럽게 구현되었습니다."
     },
     {
      "label": "B",
      "score": 2,
      "verdict_ko": "화면 좌측에 묘사된 혜수의 오른손 손가락이 6개로 렌더링되는 치명적인 신체 해부학적 오류(Hard Violation)가 발생했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "요구된 미디엄 샷 스케일과 구도를 정확히 반영했으며, 어깨를 감싸는 손의 형태가 해부학적 오류 없이 자연스럽게 구현되었습니다."
     },
     {
      "label": "A",
      "score": 2,
      "verdict_ko": "화면 좌측에 묘사된 혜수의 오른손 손가락이 6개로 렌더링되는 치명적인 신체 해부학적 오류(Hard Violation)가 발생했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 9,
     "B": 17
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": false,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S14sh4"
  }
 },
 "S18sh2::bgfirst_bg": {
  "input_fingerprint": "1ce3b9aaec9bdbbc",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 감았던 눈을 번쩍 뜬 채 굳은 표정으로 한 곳을 응시하는 수리영\n\nLOCATION (lock): Inside the main bedroom of the rooftop dwelling, at the bed where the daughter has dozed while waiting through the night. Early daylight is beginning outside.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At seated eye level beside the bed, finish the dolly-in on a tight three-quarter view of 수리영 as her eyes snap open and lock toward the unseen concern. Let her face occupy most of the frame while a small portion of the bed remains at the lower edge, holding the rigid stare before the camera eases back for her rise.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 isolated in awakening close view in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: bed (수리영 has been resting against it) — Only the near upper edge and the area beside her head remain visible; used as Lower-frame support showing where 수리영 had fallen asleep while waiting.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural ambient daytime light appropriate to the room is treated with a cool, desaturated palette and restrained contrast around her suddenly opened eyes.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 감았던 눈을 번쩍 뜬 채 굳은 표정으로 한 곳을 응시하는 수리영\n\nLOCATION (lock): Inside the main bedroom of the rooftop dwelling, at the bed where the daughter has dozed while waiting through the night. Early daylight is beginning outside.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At seated eye level beside the bed, finish the dolly-in on a tight three-quarter view of 수리영 as her eyes snap open and lock toward the unseen concern. Let her face occupy most of the frame while a small portion of the bed remains at the lower edge, holding the rigid stare before the camera eases back for her rise.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 isolated in awakening close view in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: bed (수리영 has been resting against it) — Only the near upper edge and the area beside her head remain visible; used as Lower-frame support showing where 수리영 had fallen asleep while waiting.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural ambient daytime light appropriate to the room is treated with a cool, desaturated palette and restrained contrast around her suddenly opened eyes.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S18sh2__bgfirst_bg.png",
  "asset_id": "87ce1765-6c14-49cf-b9e8-b4a7f47273cd",
  "input_asset_ids": [
   "7de5e7f0-004a-42e2-ba2e-7fd890dab160",
   "37e7d1c7-6801-4d1f-b144-a42aab8bab61"
  ]
 },
 "S18sh2": {
  "input_fingerprint": "a6c49ae73bd717df",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 감았던 눈을 번쩍 뜬 채 굳은 표정으로 한 곳을 응시하는 수리영\n\nLOCATION (lock): Inside the main bedroom of the rooftop dwelling, at the bed where the daughter has dozed while waiting through the night. Early daylight is beginning outside. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At seated eye level beside the bed, finish the dolly-in on a tight three-quarter view of 수리영 as her eyes snap open and lock toward the unseen concern. Let her face occupy most of the frame while a small portion of the bed remains at the lower edge, holding the rigid stare before the camera eases back for her rise.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 isolated in awakening close view in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: bed (수리영 has been resting against it) — Only the near upper edge and the area beside her head remain visible; used as Lower-frame support showing where 수리영 had fallen asleep while waiting.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural ambient daytime light appropriate to the room is treated with a cool, desaturated palette and restrained contrast around her suddenly opened eyes.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong still possesses the photograph taken from Min-sook and remains in the lit rooftop room after waiting there all night.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 감았던 눈을 번쩍 뜬 채 굳은 표정으로 한 곳을 응시하는 수리영\n\nLOCATION (lock): Inside the main bedroom of the rooftop dwelling, at the bed where the daughter has dozed while waiting through the night. Early daylight is beginning outside. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At seated eye level beside the bed, finish the dolly-in on a tight three-quarter view of 수리영 as her eyes snap open and lock toward the unseen concern. Let her face occupy most of the frame while a small portion of the bed remains at the lower edge, holding the rigid stare before the camera eases back for her rise.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 isolated in awakening close view in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: bed (수리영 has been resting against it) — Only the near upper edge and the area beside her head remain visible; used as Lower-frame support showing where 수리영 had fallen asleep while waiting.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural ambient daytime light appropriate to the room is treated with a cool, desaturated palette and restrained contrast around her suddenly opened eyes.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong still possesses the photograph taken from Min-sook and remains in the lit rooftop room after waiting there all night.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 감았던 눈을 번쩍 뜬 채 굳은 표정으로 한 곳을 응시하는 수리영\n\nLOCATION (lock): Inside the main bedroom of the rooftop dwelling, at the bed where the daughter has dozed while waiting through the night. Early daylight is beginning outside. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At seated eye level beside the bed, finish the dolly-in on a tight three-quarter view of 수리영 as her eyes snap open and lock toward the unseen concern. Let her face occupy most of the frame while a small portion of the bed remains at the lower edge, holding the rigid stare before the camera eases back for her rise.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 isolated in awakening close view in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: bed (수리영 has been resting against it) — Only the near upper edge and the area beside her head remain visible; used as Lower-frame support showing where 수리영 had fallen asleep while waiting.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural ambient daytime light appropriate to the room is treated with a cool, desaturated palette and restrained contrast around her suddenly opened eyes.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong still possesses the photograph taken from Min-sook and remains in the lit rooftop room after waiting there all night.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S18sh2__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S18sh2.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B01.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "얼굴이 화면 대부분을 차지하는 정확한 클로즈업 프레이밍과 서늘한 새벽 조명, 번쩍 뜬 눈의 표정을 매우 충실하게 구현했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "클로즈업 지시를 어기고 화면을 너무 넓게 잡아 인물의 비중이 작으며, 배경의 방 구조가 레퍼런스와 다르게 왜곡되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "얼굴이 화면 대부분을 차지하는 정확한 클로즈업 프레이밍과 서늘한 새벽 조명, 번쩍 뜬 눈의 표정을 매우 충실하게 구현했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "클로즈업 지시를 어기고 화면을 너무 넓게 잡아 인물의 비중이 작으며, 배경의 방 구조가 레퍼런스와 다르게 왜곡되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "요구된 클로즈업 프레이밍과 인물의 번쩍 뜬 눈 등 표정 연기를 지시대로 정확히 구현했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "지시된 클로즈업 프레이밍을 무시하고 배경 레퍼런스의 넓은 앵글을 그대로 답습하여 심각한 감점 요소가 되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "요구된 클로즈업 프레이밍과 인물의 번쩍 뜬 눈 등 표정 연기를 지시대로 정확히 구현했습니다."
     },
     {
      "label": "A",
      "score": 3,
      "verdict_ko": "지시된 클로즈업 프레이밍을 무시하고 배경 레퍼런스의 넓은 앵글을 그대로 답습하여 심각한 감점 요소가 되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 7,
     "B": 14
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   }
  },
  "totals": {
   "A": 7,
   "B": 14
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 7,
    "verdict_ko": "얼굴이 화면 대부분을 차지하는 정확한 클로즈업 프레이밍과 서늘한 새벽 조명, 번쩍 뜬 눈의 표정을 매우 충실하게 구현했습니다."
   },
   {
    "label": "A",
    "score": 4,
    "verdict_ko": "클로즈업 지시를 어기고 화면을 너무 넓게 잡아 인물의 비중이 작으며, 배경의 방 구조가 레퍼런스와 다르게 왜곡되었습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B01.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "수리영이 입고 있는 후드 집업의 색상이 레퍼런스의 짙은 남색이 아닌 회색입니다.",
     "fix_en": "Change the color of the zip-up hoodie from grey to navy blue to match the character reference.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "issue_ko": "프레이밍이 요구된 타이트한 클로즈업보다 넓어 상반신과 주변 공간이 과도하게 보이고, 얼굴이 화면 대부분을 차지하지 않는다.",
     "fix_en": "Crop in tightly around her three-quarter face so it occupies most of the frame while keeping her middle-center.",
     "unfixable": false
    },
    {
     "issue_ko": "침대가 화면 하단과 좌측에 넓게 드러나 배경 요소로서 지나치게 큰 비중을 차지한다.",
     "fix_en": "Reduce the visible bed to only a narrow lower-edge support and a small area beside her head through the tighter crop.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the color of the zip-up hoodie from grey to navy blue to match the character reference.\n- Crop in tightly around her three-quarter face so it occupies most of the frame while keeping her middle-center.\n- Reduce the visible bed to only a narrow lower-edge support and a small area beside her head through the tighter crop.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "구도와 표정은 지시사항을 잘 따랐으나, 캐릭터 레퍼런스의 파란색 후드티 대신 회색 후드티를 입고 있어 복장 일치도가 떨어집니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "클로즈업 구도, 차가운 톤의 조명, 번쩍 뜬 눈의 연기가 훌륭하며 캐릭터의 복장(파란색 후드티)도 레퍼런스와 정확히 일치합니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "구도와 표정은 지시사항을 잘 따랐으나, 캐릭터 레퍼런스의 파란색 후드티 대신 회색 후드티를 입고 있어 복장 일치도가 떨어집니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "클로즈업 구도, 차가운 톤의 조명, 번쩍 뜬 눈의 연기가 훌륭하며 캐릭터의 복장(파란색 후드티)도 레퍼런스와 정확히 일치합니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "요구된 클로즈업 프레이밍, 긴장된 표정 연기, 그리고 레퍼런스와 일치하는 의상(남색 후드)을 정확하게 구현했습니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "표정과 프레이밍은 지시를 잘 따랐으나, 레퍼런스와 다른 색상(회색)의 후드 집업을 입고 있어 감점되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "요구된 클로즈업 프레이밍, 긴장된 표정 연기, 그리고 레퍼런스와 일치하는 의상(남색 후드)을 정확하게 구현했습니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "표정과 프레이밍은 지시를 잘 따랐으나, 레퍼런스와 다른 색상(회색)의 후드 집업을 입고 있어 감점되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 11,
     "B": 15
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S18sh2__bgfirst_bg.png",
   "bg_asset_id": "87ce1765-6c14-49cf-b9e8-b4a7f47273cd",
   "bg_record_key": "S18sh2::bgfirst_bg",
   "chain_winner": false,
   "authority": "plate"
  },
  "ref_mode": "플레이트+엔티티 (2택1: 무콘티 승)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L05B01",
    "B": "L05B02",
    "C": "L05B03",
    "D": "L05B04",
    "E": "L05B05",
    "F": "L05B06"
   },
   "assigned": "L05B01",
   "choice": "Candidate A",
   "confident": false,
   "reason_ko": "지문에서 인물이 눈을 뜨고 한 곳을 응시한다는 행동만 묘사되어 있을 뿐, 침대나 거울 등 특정 서브공간을 특정할 수 있는 단서가 전혀 없습니다. 따라서 공간을 변경할 명확한 근거가 없으므로 현재 배정된 후보를 유지합니다.",
   "kept": "L05B01"
  }
 },
 "S18sh8::bgfirst_bg": {
  "input_fingerprint": "439e2c2c2c85e7c4",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 거울 속 반사된 수리영의 등 뒤 욕실 타일 벽면 위로 선명하게 나타난 붉은색 원 형태\n\nLOCATION (lock): Inside the small bathroom, on the tiled wall reflected behind the daughter in the mirror.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Dolly in from directly behind 수리영 at shoulder height toward the bathroom mirror, keeping her physical back low and left while her reflected figure occupies the middle-left portion of the mirror. Give the reflected wall more space on the right so the vivid red circle remains cleanly separated from her reflection, and retain at least one mirror edge to establish the left-right-reversed reflected view.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-left of the frame, foreground; reflected bathroom wall with red circle in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bathroom mirror (The red circle is visible only within the reflected view at this moment) — The mirror reflects 수리영's front and the bathroom wall behind her, with the wall given more reflected area on the right; used as Primary reflective field containing 수리영 and the supernatural wall mark; bathroom tile wall (A vivid red circle has suddenly appeared on it in the mirror view) — Its reflected camera-visible face bears a clear red circular form positioned away from 수리영's reflected body; used as Separated reflected background plane carrying the sudden mark.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Ambient bathroom light is kept cool and desaturated, allowing the reflected deep-crimson circle to register as the single forceful color accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 거울 속 반사된 수리영의 등 뒤 욕실 타일 벽면 위로 선명하게 나타난 붉은색 원 형태\n\nLOCATION (lock): Inside the small bathroom, on the tiled wall reflected behind the daughter in the mirror.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Dolly in from directly behind 수리영 at shoulder height toward the bathroom mirror, keeping her physical back low and left while her reflected figure occupies the middle-left portion of the mirror. Give the reflected wall more space on the right so the vivid red circle remains cleanly separated from her reflection, and retain at least one mirror edge to establish the left-right-reversed reflected view.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-left of the frame, foreground; reflected bathroom wall with red circle in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bathroom mirror (The red circle is visible only within the reflected view at this moment) — The mirror reflects 수리영's front and the bathroom wall behind her, with the wall given more reflected area on the right; used as Primary reflective field containing 수리영 and the supernatural wall mark; bathroom tile wall (A vivid red circle has suddenly appeared on it in the mirror view) — Its reflected camera-visible face bears a clear red circular form positioned away from 수리영's reflected body; used as Separated reflected background plane carrying the sudden mark.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Ambient bathroom light is kept cool and desaturated, allowing the reflected deep-crimson circle to register as the single forceful color accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S18sh8__bgfirst_bg.png",
  "asset_id": "a62ee12f-c5c9-4f93-8bea-f36b201064a1",
  "input_asset_ids": [
   "a63613fb-df0d-4d9d-a8e7-c164fbe5bfe0",
   "37e7d1c7-6801-4d1f-b144-a42aab8bab61"
  ]
 },
 "S18sh8": {
  "input_fingerprint": "7c4863d10e234b28",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 거울 속 반사된 수리영의 등 뒤 욕실 타일 벽면 위로 선명하게 나타난 붉은색 원 형태\n\nLOCATION (lock): Inside the small bathroom, on the tiled wall reflected behind the daughter in the mirror. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Dolly in from directly behind 수리영 at shoulder height toward the bathroom mirror, keeping her physical back low and left while her reflected figure occupies the middle-left portion of the mirror. Give the reflected wall more space on the right so the vivid red circle remains cleanly separated from her reflection, and retain at least one mirror edge to establish the left-right-reversed reflected view.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-left of the frame, foreground; reflected bathroom wall with red circle in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bathroom mirror (The red circle is visible only within the reflected view at this moment) — The mirror reflects 수리영's front and the bathroom wall behind her, with the wall given more reflected area on the right; used as Primary reflective field containing 수리영 and the supernatural wall mark; bathroom tile wall (A vivid red circle has suddenly appeared on it in the mirror view) — Its reflected camera-visible face bears a clear red circular form positioned away from 수리영's reflected body; used as Separated reflected background plane carrying the sudden mark.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Ambient bathroom light is kept cool and desaturated, allowing the reflected deep-crimson circle to register as the single forceful color accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong now carries her doll-adorned backpack and still has the photograph taken from Min-sook. The red circle is visible only in the mirror on the bathroom wall at this moment.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 거울 속 반사된 수리영의 등 뒤 욕실 타일 벽면 위로 선명하게 나타난 붉은색 원 형태\n\nLOCATION (lock): Inside the small bathroom, on the tiled wall reflected behind the daughter in the mirror. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Dolly in from directly behind 수리영 at shoulder height toward the bathroom mirror, keeping her physical back low and left while her reflected figure occupies the middle-left portion of the mirror. Give the reflected wall more space on the right so the vivid red circle remains cleanly separated from her reflection, and retain at least one mirror edge to establish the left-right-reversed reflected view.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-left of the frame, foreground; reflected bathroom wall with red circle in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bathroom mirror (The red circle is visible only within the reflected view at this moment) — The mirror reflects 수리영's front and the bathroom wall behind her, with the wall given more reflected area on the right; used as Primary reflective field containing 수리영 and the supernatural wall mark; bathroom tile wall (A vivid red circle has suddenly appeared on it in the mirror view) — Its reflected camera-visible face bears a clear red circular form positioned away from 수리영's reflected body; used as Separated reflected background plane carrying the sudden mark.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Ambient bathroom light is kept cool and desaturated, allowing the reflected deep-crimson circle to register as the single forceful color accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong now carries her doll-adorned backpack and still has the photograph taken from Min-sook. The red circle is visible only in the mirror on the bathroom wall at this moment.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 거울 속 반사된 수리영의 등 뒤 욕실 타일 벽면 위로 선명하게 나타난 붉은색 원 형태\n\nLOCATION (lock): Inside the small bathroom, on the tiled wall reflected behind the daughter in the mirror. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Dolly in from directly behind 수리영 at shoulder height toward the bathroom mirror, keeping her physical back low and left while her reflected figure occupies the middle-left portion of the mirror. Give the reflected wall more space on the right so the vivid red circle remains cleanly separated from her reflection, and retain at least one mirror edge to establish the left-right-reversed reflected view.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-left of the frame, foreground; reflected bathroom wall with red circle in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: bathroom mirror (The red circle is visible only within the reflected view at this moment) — The mirror reflects 수리영's front and the bathroom wall behind her, with the wall given more reflected area on the right; used as Primary reflective field containing 수리영 and the supernatural wall mark; bathroom tile wall (A vivid red circle has suddenly appeared on it in the mirror view) — Its reflected camera-visible face bears a clear red circular form positioned away from 수리영's reflected body; used as Separated reflected background plane carrying the sudden mark.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Ambient bathroom light is kept cool and desaturated, allowing the reflected deep-crimson circle to register as the single forceful color accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong now carries her doll-adorned backpack and still has the photograph taken from Min-sook. The red circle is visible only in the mirror on the bathroom wall at this moment.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S18sh8__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S18sh8.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    },
    {
     "label": "PROP REFERENCE — 붉은 원 표식: the exact object appearing in this shot; match its look, material and wear exactly.",
     "path": "<bytes:1135490>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B01.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    },
    {
     "label": "PROP REFERENCE — 붉은 원 표식: the exact object appearing in this shot; match its look, material and wear exactly.",
     "path": "<bytes:1135490>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "지시된 숄더 앵글과 프레이밍을 정확히 구현했으며, 차갑고 채도가 낮은 조명 속 붉은 원의 대비와 캐릭터의 의상 디테일(남색 후드티)이 매우 훌륭합니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "캐릭터가 쥔 사진의 앞면(기능면)이 카메라를 향해 있어 소품 방향 규칙을 위반했으며, 기준 이미지의 남색 후드티 의상을 누락했습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "지시된 숄더 앵글과 프레이밍을 정확히 구현했으며, 차갑고 채도가 낮은 조명 속 붉은 원의 대비와 캐릭터의 의상 디테일(남색 후드티)이 매우 훌륭합니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "캐릭터가 쥔 사진의 앞면(기능면)이 카메라를 향해 있어 소품 방향 규칙을 위반했으며, 기준 이미지의 남색 후드티 의상을 누락했습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "지시된 구도, 쿨톤 조명, 의상(파란 후드)을 정확히 구현하고 거울 속 벽면의 붉은 원을 잘 표현했으나 손에 든 사진 소품이 누락되었습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "전경에서는 사진의 앞면이 거울을 향해 있으나 거울 속 반사에서는 뒷면(백지)이 보이는 물리적 오류(하드 위반)가 발생했으며 지정된 겉옷 의상도 누락되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "지시된 구도, 쿨톤 조명, 의상(파란 후드)을 정확히 구현하고 거울 속 벽면의 붉은 원을 잘 표현했으나 손에 든 사진 소품이 누락되었습니다."
     },
     {
      "label": "A",
      "score": 3,
      "verdict_ko": "전경에서는 사진의 앞면이 거울을 향해 있으나 거울 속 반사에서는 뒷면(백지)이 보이는 물리적 오류(하드 위반)가 발생했으며 지정된 겉옷 의상도 누락되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 8,
     "B": 17
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   }
  },
  "totals": {
   "A": 8,
   "B": 17
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 9,
    "verdict_ko": "지시된 숄더 앵글과 프레이밍을 정확히 구현했으며, 차갑고 채도가 낮은 조명 속 붉은 원의 대비와 캐릭터의 의상 디테일(남색 후드티)이 매우 훌륭합니다."
   },
   {
    "label": "A",
    "score": 5,
    "verdict_ko": "캐릭터가 쥔 사진의 앞면(기능면)이 카메라를 향해 있어 소품 방향 규칙을 위반했으며, 기준 이미지의 남색 후드티 의상을 누락했습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B01.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   },
   {
    "label": "PROP REFERENCE — 붉은 원 표식: the exact object appearing in this shot; match its look, material and wear exactly.",
    "path": "<bytes:1135490>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "캐릭터 레퍼런스 이미지에 있는 파란색 야구 모자가 전경의 인물과 거울 속 뒷모습 모두에서 누락되었습니다.",
     "fix_en": "Add the blue baseball cap from the character reference to the character's head in both the foreground physical body and her mirror reflection.",
     "unfixable": false
    },
    {
     "issue_ko": "거울 속에 비친 붉은색 원이 타일 벽면에 칠해진 것이 아니라 화면에 덧붙여진 평면 그래픽(스티커)처럼 보이며, 반사된 타일의 줄눈과 원근감을 무시하고 있습니다.",
     "fix_en": "Embed the red circle into the reflected tiled wall behind the character, ensuring the paint stroke interacts realistically with the wall's perspective, lighting, and tile grout lines.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": []
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the blue baseball cap from the character reference to the character's head in both the foreground physical body and her mirror reflection.\n- Embed the red circle into the reflected tiled wall behind the character, ensuring the paint stroke interacts realistically with the wall's perspective, lighting, and tile grout lines.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "요구된 구도와 조명, 피사체의 위치를 완벽히 구현했으며 캐릭터 레퍼런스의 모자와 인형 달린 백팩까지 정확하게 재현했습니다."
     },
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "구도와 붉은 원의 위치, 백팩 등은 훌륭하게 구현했으나 캐릭터 레퍼런스에 있는 파란색 모자를 누락했습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "요구된 구도와 조명, 피사체의 위치를 완벽히 구현했으며 캐릭터 레퍼런스의 모자와 인형 달린 백팩까지 정확하게 재현했습니다."
     },
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "구도와 붉은 원의 위치, 백팩 등은 훌륭하게 구현했으나 캐릭터 레퍼런스에 있는 파란색 모자를 누락했습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 10,
      "verdict_ko": "지시된 카메라 구도와 거울 반사 위치를 완벽하게 구현했으며, 레퍼런스의 파란색 모자 등 의상 디테일을 정확히 반영했습니다."
     },
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "구도와 붉은 원의 배치 등 훌륭하게 연출되었으나, 캐릭터 레퍼런스에 착용된 파란색 모자가 누락되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "지시된 카메라 구도와 거울 반사 위치를 완벽하게 구현했으며, 레퍼런스의 파란색 모자 등 의상 디테일을 정확히 반영했습니다."
     },
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "구도와 붉은 원의 배치 등 훌륭하게 연출되었으나, 캐릭터 레퍼런스에 착용된 파란색 모자가 누락되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 17,
     "B": 20
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S18sh8__bgfirst_bg.png",
   "bg_asset_id": "a62ee12f-c5c9-4f93-8bea-f36b201064a1",
   "bg_record_key": "S18sh8::bgfirst_bg",
   "chain_winner": false,
   "authority": "plate"
  },
  "ref_mode": "플레이트+엔티티 (2택1: 무콘티 승)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L05B01",
    "B": "L05B02",
    "C": "L05B03",
    "D": "L05B04",
    "E": "L05B05",
    "F": "L05B06"
   },
   "assigned": "L05B01",
   "choice": "Candidate A",
   "confident": true,
   "reason_ko": "욕실 타일 벽면 위로 나타난 붉은색 원 형태와 거울이라는 명확한 언급이 있으므로 욕실 내부가 보이는 Candidate A가 적합합니다.",
   "kept": "L05B01"
  }
 },
 "S18sh12::variants": {
  "author_fp": "1c3d46d7df93f548",
  "author": {
   "variants": [
    {
     "approach_ko": "얕은 초점과 촉각적인 인화지 질감으로 웃는 얼굴의 붕괴를 주관적이고 불안하게 포착한다.",
     "prompt_en": "Use an intimate, subjective close-focus treatment with very shallow depth, holding the eyes and collapsing smile at the sharpest plane while the thin frame boundary falls gently soft. Let the existing cool morning ambience skim the physical photograph, revealing subtle print texture as the facial features drag downward with damp, tactile realism; keep the restrained shadows dense and the peripheral composition claustrophobic."
    },
    {
     "approach_ko": "정면성이 강한 깊은 초점과 엄격한 대칭 구도로 사진 속 이상 현상을 차갑고 임상적으로 제시한다.",
     "prompt_en": "Photograph the image with rigid, nearly frontal geometry and comparatively deep focus, keeping both the distorted printed face and the narrow frame boundary crisply legible. Favor a centered, forensic composition with even cool morning illumination and controlled low-key falloff, making the downward smear feel starkly embedded in the photographic image rather than created by soft focus or motion blur."
    }
   ]
  },
  "reused": false
 },
 "S18sh12": {
  "input_fingerprint": "ae5b49f62996719c",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 환하게 웃고 있는 액자 사진 속 강민숙(민숙)의 얼굴 이미지가 뭉개지듯 일그러진 순간\n\nLOCATION (lock): The living and dining area of the rooftop dwelling, at the table holding a framed family photograph. Daylight has reached the home by morning. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: End the inherited dolly-in directly over the framed photograph, slightly above and behind where 수리영's shoulder had been, until 강민숙's smiling face occupies most of the close frame while a narrow strip of the photograph edge remains visible. Hold at the peak of the hallucinated facial collapse, rendering the printed features as if they are smearing downward within the image rather than transforming into a physically present person.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: 액자 속 사진 (식탁 위 액자 안에 놓여 있으며, 얼굴 이미지만 환각적으로 일그러져 보임) — The printed front faces the camera; 강민숙's smiling portrait is visible and appears to smear downward at the distortion peak; used as Primary focal surface carrying the hallucinated transformation, with its boundary retained for photographic context; 사진 액자 (식탁 위에 놓여 있음) — Its front edge is nearly square to the camera, with only a narrow border remaining in frame; used as A thin peripheral boundary proving that the face belongs to a framed photograph.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use restrained low-key contrast and a cool desaturated grade, without introducing a specific unestablished light source.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong continues to have her doll-adorned backpack and the old photograph. The framed image of Min-sook returns to normal immediately after its momentary distortion.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 환하게 웃고 있는 액자 사진 속 강민숙(민숙)의 얼굴 이미지가 뭉개지듯 일그러진 순간\n\nLOCATION (lock): The living and dining area of the rooftop dwelling, at the table holding a framed family photograph. Daylight has reached the home by morning. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: End the inherited dolly-in directly over the framed photograph, slightly above and behind where 수리영's shoulder had been, until 강민숙's smiling face occupies most of the close frame while a narrow strip of the photograph edge remains visible. Hold at the peak of the hallucinated facial collapse, rendering the printed features as if they are smearing downward within the image rather than transforming into a physically present person.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: 액자 속 사진 (식탁 위 액자 안에 놓여 있으며, 얼굴 이미지만 환각적으로 일그러져 보임) — The printed front faces the camera; 강민숙's smiling portrait is visible and appears to smear downward at the distortion peak; used as Primary focal surface carrying the hallucinated transformation, with its boundary retained for photographic context; 사진 액자 (식탁 위에 놓여 있음) — Its front edge is nearly square to the camera, with only a narrow border remaining in frame; used as A thin peripheral boundary proving that the face belongs to a framed photograph.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use restrained low-key contrast and a cool desaturated grade, without introducing a specific unestablished light source.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong continues to have her doll-adorned backpack and the old photograph. The framed image of Min-sook returns to normal immediately after its momentary distortion.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse an intimate, subjective close-focus treatment with very shallow depth, holding the eyes and collapsing smile at the sharpest plane while the thin frame boundary falls gently soft. Let the existing cool morning ambience skim the physical photograph, revealing subtle print texture as the facial features drag downward with damp, tactile realism; keep the restrained shadows dense and the peripheral composition claustrophobic.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 환하게 웃고 있는 액자 사진 속 강민숙(민숙)의 얼굴 이미지가 뭉개지듯 일그러진 순간\n\nLOCATION (lock): The living and dining area of the rooftop dwelling, at the table holding a framed family photograph. Daylight has reached the home by morning. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: End the inherited dolly-in directly over the framed photograph, slightly above and behind where 수리영's shoulder had been, until 강민숙's smiling face occupies most of the close frame while a narrow strip of the photograph edge remains visible. Hold at the peak of the hallucinated facial collapse, rendering the printed features as if they are smearing downward within the image rather than transforming into a physically present person.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: 액자 속 사진 (식탁 위 액자 안에 놓여 있으며, 얼굴 이미지만 환각적으로 일그러져 보임) — The printed front faces the camera; 강민숙's smiling portrait is visible and appears to smear downward at the distortion peak; used as Primary focal surface carrying the hallucinated transformation, with its boundary retained for photographic context; 사진 액자 (식탁 위에 놓여 있음) — Its front edge is nearly square to the camera, with only a narrow border remaining in frame; used as A thin peripheral boundary proving that the face belongs to a framed photograph.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use restrained low-key contrast and a cool desaturated grade, without introducing a specific unestablished light source.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong continues to have her doll-adorned backpack and the old photograph. The framed image of Min-sook returns to normal immediately after its momentary distortion.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nPhotograph the image with rigid, nearly frontal geometry and comparatively deep focus, keeping both the distorted printed face and the narrow frame boundary crisply legible. Favor a centered, forensic composition with even cool morning illumination and controlled low-key falloff, making the downward smear feel starkly embedded in the photographic image rather than created by soft focus or motion blur.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "A": 9,
   "B": 4
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 9,
    "verdict_ko": "지정된 클로즈업 샷 크기를 완벽히 준수하여 얼굴이 프레임 대부분을 차지하고 액자 가장자리만 좁게 보이도록 연출함."
   },
   {
    "label": "B",
    "score": 4,
    "verdict_ko": "지시된 클로즈업을 무시하고 액자 전체와 뒷배경 공간까지 넓게 보여주어 프레이밍 규칙을 위반함."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L05B01.png"
   }
  ],
  "critique": {
   "issues": []
  },
  "composition_critique": {
   "issues": [
    {
     "issue_ko": "액자 전면이 카메라에 거의 정면으로 놓여야 하지만, 현재는 강한 사선 원근으로 기울어져 오른쪽 위 프레임 경계가 과도하게 넓게 보인다.",
     "fix_en": "Apply perspective correction and a slightly tighter crop so the photograph front reads nearly square to the camera with only a uniformly narrow frame boundary visible.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Apply perspective correction and a slightly tighter crop so the photograph front reads nearly square to the camera with only a uniformly narrow frame boundary visible.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "액자 속 사진의 인물 얼굴이 아래로 뭉개지듯 일그러지는 환각적인 순간과 클로즈업 구도를 지시문대로 정확히 구현했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "클로즈업 구도와 톤은 맞으나, 가장 중요한 핵심 지시인 '얼굴이 뭉개지듯 일그러지는 효과'가 전혀 표현되지 않았습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "액자 속 사진의 인물 얼굴이 아래로 뭉개지듯 일그러지는 환각적인 순간과 클로즈업 구도를 지시문대로 정확히 구현했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "클로즈업 구도와 톤은 맞으나, 가장 중요한 핵심 지시인 '얼굴이 뭉개지듯 일그러지는 효과'가 전혀 표현되지 않았습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "카메라 각도가 정면에서 약간 벗어났으나, 프롬프트의 핵심인 액자 속 미소 짓는 얼굴이 아래로 뭉개지듯 일그러지는 환각적인 순간을 완벽하게 구현했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "정면 클로즈업 구도와 차가운 색감은 잘 살렸으나, 가장 중요한 샷의 핵심인 '얼굴이 일그러지는 왜곡 현상'이 완전히 누락되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "카메라 각도가 정면에서 약간 벗어났으나, 프롬프트의 핵심인 액자 속 미소 짓는 얼굴이 아래로 뭉개지듯 일그러지는 환각적인 순간을 완벽하게 구현했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "정면 클로즈업 구도와 차가운 색감은 잘 살렸으나, 가장 중요한 샷의 핵심인 '얼굴이 일그러지는 왜곡 현상'이 완전히 누락되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 14,
     "B": 7
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "A",
   "fix_won": false,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "플레이트만 (배경 전용)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L05B01",
    "B": "L05B02",
    "C": "L05B03",
    "D": "L05B04",
    "E": "L05B05",
    "F": "L05B06"
   },
   "assigned": "L05B01",
   "choice": "A",
   "confident": true,
   "reason_ko": "원문에서 '액자 사진'과 그 이미지가 '뭉개지듯 일그러진 순간'을 묘사하고 있으며, 후보 A만이 테이블 위에 일그러진(녹아내리는 듯한) 효과가 들어간 사진이 놓여 있는 모습을 명확히 보여줍니다.",
   "chosen": "L05B01"
  }
 },
 "S19sh1::bgfirst_bg": {
  "input_fingerprint": "602cc682941a9955",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 이른 아침 한산한 마트 안, 진열대 사이에서 주위를 유심히 살피는 수리영의 전신\n\nLOCATION (lock): Inside the quiet supermarket in the early morning, among the merchandise aisles on the main sales floor.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From chest height five meters ahead in the aisle, track backward at 수리영's pace while maintaining a full-body view between the displays. Keep her slightly off-center with the aisle extending behind her, catching her mid-stride as her head and eyes search from one side to the other.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, midground, moves toward far end of the store aisle; far end of the store aisle in the upper-center of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 진열대 (진열대 사이 통로를 이루고 있음) — Their aisle-facing sides flank the frame and recede toward the rear of the store; used as The displays form parallel side boundaries around 수리영 and guide the backward tracking path; 마트 통로 (이른 아침이라 한산함); used as Provides the long search axis and visible spatial context around the moving figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use ambient store illumination appropriate to the early-morning interior, graded cool and desaturated with restrained contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 이른 아침 한산한 마트 안, 진열대 사이에서 주위를 유심히 살피는 수리영의 전신\n\nLOCATION (lock): Inside the quiet supermarket in the early morning, among the merchandise aisles on the main sales floor.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From chest height five meters ahead in the aisle, track backward at 수리영's pace while maintaining a full-body view between the displays. Keep her slightly off-center with the aisle extending behind her, catching her mid-stride as her head and eyes search from one side to the other.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, midground, moves toward far end of the store aisle; far end of the store aisle in the upper-center of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 진열대 (진열대 사이 통로를 이루고 있음) — Their aisle-facing sides flank the frame and recede toward the rear of the store; used as The displays form parallel side boundaries around 수리영 and guide the backward tracking path; 마트 통로 (이른 아침이라 한산함); used as Provides the long search axis and visible spatial context around the moving figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use ambient store illumination appropriate to the early-morning interior, graded cool and desaturated with restrained contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S19sh1__bgfirst_bg.png",
  "asset_id": "081ffa2f-ae80-4057-b1e9-f1db9c42e760",
  "input_asset_ids": [
   "4c573b7b-dbf9-4a95-bd2d-ec876bf4438c",
   "3b6c9161-15bf-4f6a-b131-f9995e713788"
  ]
 },
 "S19sh1": {
  "input_fingerprint": "3509972bdd01566d",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 이른 아침 한산한 마트 안, 진열대 사이에서 주위를 유심히 살피는 수리영의 전신\n\nLOCATION (lock): Inside the quiet supermarket in the early morning, among the merchandise aisles on the main sales floor. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From chest height five meters ahead in the aisle, track backward at 수리영's pace while maintaining a full-body view between the displays. Keep her slightly off-center with the aisle extending behind her, catching her mid-stride as her head and eyes search from one side to the other.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, midground, moves toward far end of the store aisle; far end of the store aisle in the upper-center of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 진열대 (진열대 사이 통로를 이루고 있음) — Their aisle-facing sides flank the frame and recede toward the rear of the store; used as The displays form parallel side boundaries around 수리영 and guide the backward tracking path; 마트 통로 (이른 아침이라 한산함); used as Provides the long search axis and visible spatial context around the moving figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use ambient store illumination appropriate to the early-morning interior, graded cool and desaturated with restrained contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong’s bicycle remains parked outside the mart while she searches inside; she continues to have her backpack and Min-sook’s old photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 이른 아침 한산한 마트 안, 진열대 사이에서 주위를 유심히 살피는 수리영의 전신\n\nLOCATION (lock): Inside the quiet supermarket in the early morning, among the merchandise aisles on the main sales floor. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From chest height five meters ahead in the aisle, track backward at 수리영's pace while maintaining a full-body view between the displays. Keep her slightly off-center with the aisle extending behind her, catching her mid-stride as her head and eyes search from one side to the other.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, midground, moves toward far end of the store aisle; far end of the store aisle in the upper-center of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 진열대 (진열대 사이 통로를 이루고 있음) — Their aisle-facing sides flank the frame and recede toward the rear of the store; used as The displays form parallel side boundaries around 수리영 and guide the backward tracking path; 마트 통로 (이른 아침이라 한산함); used as Provides the long search axis and visible spatial context around the moving figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use ambient store illumination appropriate to the early-morning interior, graded cool and desaturated with restrained contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong’s bicycle remains parked outside the mart while she searches inside; she continues to have her backpack and Min-sook’s old photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 이른 아침 한산한 마트 안, 진열대 사이에서 주위를 유심히 살피는 수리영의 전신\n\nLOCATION (lock): Inside the quiet supermarket in the early morning, among the merchandise aisles on the main sales floor. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From chest height five meters ahead in the aisle, track backward at 수리영's pace while maintaining a full-body view between the displays. Keep her slightly off-center with the aisle extending behind her, catching her mid-stride as her head and eyes search from one side to the other.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, midground, moves toward far end of the store aisle; far end of the store aisle in the upper-center of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 진열대 (진열대 사이 통로를 이루고 있음) — Their aisle-facing sides flank the frame and recede toward the rear of the store; used as The displays form parallel side boundaries around 수리영 and guide the backward tracking path; 마트 통로 (이른 아침이라 한산함); used as Provides the long search axis and visible spatial context around the moving figure.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use ambient store illumination appropriate to the early-morning interior, graded cool and desaturated with restrained contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong’s bicycle remains parked outside the mart while she searches inside; she continues to have her backpack and Min-sook’s old photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S19sh1__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S19sh1.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L09B04.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "요구된 약간 비스듬한(off-center) 프레이밍을 잘 구현하였으며, 공간 레퍼런스의 구도를 그대로 복사하지 않고 새로운 카메라 시점을 설정해 지침을 준수했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "레퍼런스 이미지의 카메라 구도를 절대 복사하지 말라는 지침을 위반하고 배경을 그대로 차용했으며, 인물을 정중앙에 배치하여 프레이밍 요구사항에 실패했습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "요구된 약간 비스듬한(off-center) 프레이밍을 잘 구현하였으며, 공간 레퍼런스의 구도를 그대로 복사하지 않고 새로운 카메라 시점을 설정해 지침을 준수했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "레퍼런스 이미지의 카메라 구도를 절대 복사하지 말라는 지침을 위반하고 배경을 그대로 차용했으며, 인물을 정중앙에 배치하여 프레이밍 요구사항에 실패했습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지시된 약간 중심에서 벗어난(off-center) 구도를 잘 구현했으며, 천장 조명과 표지판 등 공간의 디테일과 의상(에코백 포함)이 레퍼런스와 충실히 일치합니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "인물이 프레임 정중앙에 배치되어 구도 지시를 어겼으며, 천장 조명의 방향이 어긋나고 에코백 소품이 누락되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "지시된 약간 중심에서 벗어난(off-center) 구도를 잘 구현했으며, 천장 조명과 표지판 등 공간의 디테일과 의상(에코백 포함)이 레퍼런스와 충실히 일치합니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "인물이 프레임 정중앙에 배치되어 구도 지시를 어겼으며, 천장 조명의 방향이 어긋나고 에코백 소품이 누락되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 12,
     "B": 11
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": false,
    "policy": 1
   }
  },
  "totals": {
   "A": 12,
   "B": 11
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 7,
    "verdict_ko": "요구된 약간 비스듬한(off-center) 프레이밍을 잘 구현하였으며, 공간 레퍼런스의 구도를 그대로 복사하지 않고 새로운 카메라 시점을 설정해 지침을 준수했습니다."
   },
   {
    "label": "B",
    "score": 4,
    "verdict_ko": "레퍼런스 이미지의 카메라 구도를 절대 복사하지 말라는 지침을 위반하고 배경을 그대로 차용했으며, 인물을 정중앙에 배치하여 프레이밍 요구사항에 실패했습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L09B04.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "손에 든 사진의 앞면이 카메라를 향하고 있습니다. 지침에 따라 사진의 앞면은 캐릭터를 향하고 카메라에는 뒷면이 보여야 합니다.",
     "fix_en": "Flip the photograph in her hand so the blank back faces the camera and the image side faces the character.",
     "unfixable": false
    },
    {
     "issue_ko": "레이아웃 스케치와 달리 다리의 위치가 반대로 되어 있습니다(스케치는 오른발이 앞, 생성된 이미지는 왼발이 앞).",
     "fix_en": "Redraw her legs to match the layout sketch, with her right foot stepping forward flat on the ground and her left leg trailing behind.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": []
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Flip the photograph in her hand so the blank back faces the camera and the image side faces the character.\n- Redraw her legs to match the layout sketch, with her right foot stepping forward flat on the ground and her left leg trailing behind.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "지시된 구도와 배경을 정확히 유지하며, 인물이 보는 사진의 뒷면이 카메라를 향하도록 소품 방향을 올바르게 연출했습니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "전반적인 구도와 배경은 좋으나, 인물이 보는 사진의 앞면(기능면)이 카메라를 향하고 있어 소품 방향 지침을 위반했습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "지시된 구도와 배경을 정확히 유지하며, 인물이 보는 사진의 뒷면이 카메라를 향하도록 소품 방향을 올바르게 연출했습니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "전반적인 구도와 배경은 좋으나, 인물이 보는 사진의 앞면(기능면)이 카메라를 향하고 있어 소품 방향 지침을 위반했습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "캐릭터와 배경이 프롬프트 및 스케치와 정확히 일치하며, 들고 있는 사진의 뒷면이 카메라를 향하도록 소품 방향 규칙을 올바르게 준수했습니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "전반적인 구도와 배경은 좋으나, 인물이 들고 있는 사진의 앞면(기능적 측면)이 카메라를 향하고 있어 소품 방향 규칙을 위반했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "캐릭터와 배경이 프롬프트 및 스케치와 정확히 일치하며, 들고 있는 사진의 뒷면이 카메라를 향하도록 소품 방향 규칙을 올바르게 준수했습니다."
     },
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "전반적인 구도와 배경은 좋으나, 인물이 들고 있는 사진의 앞면(기능적 측면)이 카메라를 향하고 있어 소품 방향 규칙을 위반했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 11,
     "B": 15
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S19sh1__bgfirst_bg.png",
   "bg_asset_id": "081ffa2f-ae80-4057-b1e9-f1db9c42e760",
   "bg_record_key": "S19sh1::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L09B01",
    "B": "L09B02",
    "C": "L09B03",
    "D": "L09B04"
   },
   "assigned": "L09B04",
   "choice": "Candidate D",
   "confident": true,
   "reason_ko": "\"마트 안, 진열대 사이\"라는 묘사에 정확히 부합하는 실내 매대 통로 공간입니다.",
   "kept": "L09B04"
  }
 },
 "S19sh4::bgfirst_bg": {
  "input_fingerprint": "56a04a6750f5a679",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 마트 건물 밖, 고개를 뒤로 젖혀 건물 외벽의 CCTV 카메라를 올려다보는 수리영의 얼굴\n\nLOCATION (lock): Outside the supermarket, directly beneath the section of the building facade fitted with a surveillance camera.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Continue the reveal tilt from slightly below 수리영's eye level, tightening within one meter so her upturned face occupies most of the lower and middle frame while the CCTV camera remains above along the same sightline. Use a restrained low angle and shallow focus that preserves the CCTV silhouette clearly enough to establish what she has found.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward exterior CCTV camera; exterior CCTV camera in the upper-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 외벽 CCTV 카메라 (마트 외벽에 설치되어 있음) — Its lens-bearing front is angled toward the area outside the store and remains visible above 수리영; used as The upper-frame visual target completing 수리영's upward eyeline; 마트 건물 외벽 (CCTV가 설치되어 있음) — The exterior face carrying the CCTV is visible behind and above her; used as Anchors the CCTV physically above 수리영 without widening away from her face.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use natural ambient light appropriate to the early-morning exterior with a cool desaturated grade and controlled facial contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 마트 건물 밖, 고개를 뒤로 젖혀 건물 외벽의 CCTV 카메라를 올려다보는 수리영의 얼굴\n\nLOCATION (lock): Outside the supermarket, directly beneath the section of the building facade fitted with a surveillance camera.\n\nTIME OF DAY (lock): early morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Continue the reveal tilt from slightly below 수리영's eye level, tightening within one meter so her upturned face occupies most of the lower and middle frame while the CCTV camera remains above along the same sightline. Use a restrained low angle and shallow focus that preserves the CCTV silhouette clearly enough to establish what she has found.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward exterior CCTV camera; exterior CCTV camera in the upper-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 외벽 CCTV 카메라 (마트 외벽에 설치되어 있음) — Its lens-bearing front is angled toward the area outside the store and remains visible above 수리영; used as The upper-frame visual target completing 수리영's upward eyeline; 마트 건물 외벽 (CCTV가 설치되어 있음) — The exterior face carrying the CCTV is visible behind and above her; used as Anchors the CCTV physically above 수리영 without widening away from her face.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use natural ambient light appropriate to the early-morning exterior with a cool desaturated grade and controlled facial contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S19sh4__bgfirst_bg.png",
  "asset_id": "b8e8a27d-1cae-43d7-a7d7-cb9a50004dd5",
  "input_asset_ids": [
   "adc7810b-b519-403e-b1a6-9a52897267d0",
   "c804af48-4ad1-417d-acb4-9e8d5dcbd3f5"
  ]
 },
 "S19sh4": {
  "input_fingerprint": "695edfe8f46f284a",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 마트 건물 밖, 고개를 뒤로 젖혀 건물 외벽의 CCTV 카메라를 올려다보는 수리영의 얼굴\n\nLOCATION (lock): Outside the supermarket, directly beneath the section of the building facade fitted with a surveillance camera. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Continue the reveal tilt from slightly below 수리영's eye level, tightening within one meter so her upturned face occupies most of the lower and middle frame while the CCTV camera remains above along the same sightline. Use a restrained low angle and shallow focus that preserves the CCTV silhouette clearly enough to establish what she has found.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward exterior CCTV camera; exterior CCTV camera in the upper-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 외벽 CCTV 카메라 (마트 외벽에 설치되어 있음) — Its lens-bearing front is angled toward the area outside the store and remains visible above 수리영; used as The upper-frame visual target completing 수리영's upward eyeline; 마트 건물 외벽 (CCTV가 설치되어 있음) — The exterior face carrying the CCTV is visible behind and above her; used as Anchors the CCTV physically above 수리영 without widening away from her face.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use natural ambient light appropriate to the early-morning exterior with a cool desaturated grade and controlled facial contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong remains outside the mart with her bicycle parked there, still carrying her backpack and the photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 마트 건물 밖, 고개를 뒤로 젖혀 건물 외벽의 CCTV 카메라를 올려다보는 수리영의 얼굴\n\nLOCATION (lock): Outside the supermarket, directly beneath the section of the building facade fitted with a surveillance camera. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Continue the reveal tilt from slightly below 수리영's eye level, tightening within one meter so her upturned face occupies most of the lower and middle frame while the CCTV camera remains above along the same sightline. Use a restrained low angle and shallow focus that preserves the CCTV silhouette clearly enough to establish what she has found.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward exterior CCTV camera; exterior CCTV camera in the upper-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 외벽 CCTV 카메라 (마트 외벽에 설치되어 있음) — Its lens-bearing front is angled toward the area outside the store and remains visible above 수리영; used as The upper-frame visual target completing 수리영's upward eyeline; 마트 건물 외벽 (CCTV가 설치되어 있음) — The exterior face carrying the CCTV is visible behind and above her; used as Anchors the CCTV physically above 수리영 without widening away from her face.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use natural ambient light appropriate to the early-morning exterior with a cool desaturated grade and controlled facial contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong remains outside the mart with her bicycle parked there, still carrying her backpack and the photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning.\n\nSHOT TEXT (authoritative, Korean): 마트 건물 밖, 고개를 뒤로 젖혀 건물 외벽의 CCTV 카메라를 올려다보는 수리영의 얼굴\n\nLOCATION (lock): Outside the supermarket, directly beneath the section of the building facade fitted with a surveillance camera. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Continue the reveal tilt from slightly below 수리영's eye level, tightening within one meter so her upturned face occupies most of the lower and middle frame while the CCTV camera remains above along the same sightline. Use a restrained low angle and shallow focus that preserves the CCTV silhouette clearly enough to establish what she has found.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-left of the frame, foreground, looks toward exterior CCTV camera; exterior CCTV camera in the upper-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 마트 외벽 CCTV 카메라 (마트 외벽에 설치되어 있음) — Its lens-bearing front is angled toward the area outside the store and remains visible above 수리영; used as The upper-frame visual target completing 수리영's upward eyeline; 마트 건물 외벽 (CCTV가 설치되어 있음) — The exterior face carrying the CCTV is visible behind and above her; used as Anchors the CCTV physically above 수리영 without widening away from her face.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use natural ambient light appropriate to the early-morning exterior with a cool desaturated grade and controlled facial contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong remains outside the mart with her bicycle parked there, still carrying her backpack and the photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S19sh4__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S19sh4.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L09B03.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지정된 앵글과 프레이밍을 정확히 구현했으며, 요구된 소품(사진)과 인물의 외양을 자연스럽게 묘사했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "가방 끈과 어깨 사이에 끼워진 사진이 포스터처럼 말려 있으며 물리적으로 불가능한 형태(클리핑)로 렌더링되어 감점되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지정된 앵글과 프레이밍을 정확히 구현했으며, 요구된 소품(사진)과 인물의 외양을 자연스럽게 묘사했습니다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "가방 끈과 어깨 사이에 끼워진 사진이 포스터처럼 말려 있으며 물리적으로 불가능한 형태(클리핑)로 렌더링되어 감점되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "요구된 얼굴 중심의 클로즈업 프레이밍과 시선을 정확히 구현했으나, 소지품(사진)이 둥글게 말린 형태로 다소 어색하게 묘사되었습니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "캐릭터와 배경의 구현은 훌륭하나, 손에 든 사진의 앞면을 카메라 렌즈 방향으로 부자연스럽게 노출하여 소품 방향 규칙을 위반했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "요구된 얼굴 중심의 클로즈업 프레이밍과 시선을 정확히 구현했으나, 소지품(사진)이 둥글게 말린 형태로 다소 어색하게 묘사되었습니다."
     },
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "캐릭터와 배경의 구현은 훌륭하나, 손에 든 사진의 앞면을 카메라 렌즈 방향으로 부자연스럽게 노출하여 소품 방향 규칙을 위반했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 13,
     "B": 10
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": false,
    "policy": 1
   }
  },
  "totals": {
   "A": 13,
   "B": 10
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 7,
    "verdict_ko": "지정된 앵글과 프레이밍을 정확히 구현했으며, 요구된 소품(사진)과 인물의 외양을 자연스럽게 묘사했습니다."
   },
   {
    "label": "B",
    "score": 3,
    "verdict_ko": "가방 끈과 어깨 사이에 끼워진 사진이 포스터처럼 말려 있으며 물리적으로 불가능한 형태(클리핑)로 렌더링되어 감점되었습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L09B03.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "캐릭터 레퍼런스에 있는 파란색 캡 모자가 누락되었고, 머리 묶은 위치가 다르게 표현되었습니다.",
     "fix_en": "Add the blue baseball cap from the character reference and adjust the hairstyle to a high bun protruding from the back of the cap.",
     "unfixable": false
    },
    {
     "issue_ko": "손에 든 사진의 앞면(이미지)이 카메라를 향하고 있어 '소품 방향(PROPS FACE THE RIGHT WAY)' 규칙을 위반했습니다.",
     "fix_en": "Flip the photograph in her hand so its blank back faces the camera, reflecting a natural carrying position rather than presenting it to the viewer.",
     "unfixable": false
    },
    {
     "issue_ko": "지시된 '얕은 심도(shallow focus)'가 적용되지 않아 인물부터 배경의 건물과 CCTV까지 모두 너무 선명합니다.",
     "fix_en": "Apply a shallow depth of field to soften the background architecture, keeping the character's face sharply in focus while leaving the CCTV slightly blurred but recognizable.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Crop in slightly around Su-ri-yeong and the CCTV so her upturned face occupies most of the lower and middle frame, while keeping her middle-left and the CCTV upper-right.",
     "issue_ko": "현재 프레이밍이 너무 넓어 수리영의 상반신과 외벽이 과도하게 보이며, 얼굴이 하단과 중단 대부분을 차지해야 하는 클로즈업 스케일보다 작다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the blue baseball cap from the character reference and adjust the hairstyle to a high bun protruding from the back of the cap.\n- Flip the photograph in her hand so its blank back faces the camera, reflecting a natural carrying position rather than presenting it to the viewer.\n- Apply a shallow depth of field to soften the background architecture, keeping the character's face sharply in focus while leaving the CCTV slightly blurred but recognizable.\n- Crop in slightly around Su-ri-yeong and the CCTV so her upturned face occupies most of the lower and middle frame, while keeping her middle-left and the CCTV upper-right.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "레퍼런스의 의상(모자)을 정확히 반영했으며, 사진 소품의 뒷면이 카메라를 향하게 하여 소품 방향 지침을 올바르게 준수했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지시문과 달리 사진 소품의 앞면이 카메라를 향해 있어 규칙을 위반했으며, 캐릭터 레퍼런스의 모자도 누락되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "레퍼런스의 의상(모자)을 정확히 반영했으며, 사진 소품의 뒷면이 카메라를 향하게 하여 소품 방향 지침을 올바르게 준수했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "지시문과 달리 사진 소품의 앞면이 카메라를 향해 있어 규칙을 위반했으며, 캐릭터 레퍼런스의 모자도 누락되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "인물 레퍼런스의 복장(모자)을 유지하고 지침에 따라 사진의 뒷면이 보이도록 소품 방향을 정확히 연출했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "샷 텍스트의 지시가 없음에도 사진의 기능적 면(앞면)이 카메라를 향하도록 연출하여 소품 방향 지침을 위반했습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "인물 레퍼런스의 복장(모자)을 유지하고 지침에 따라 사진의 뒷면이 보이도록 소품 방향을 정확히 연출했습니다."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "샷 텍스트의 지시가 없음에도 사진의 기능적 면(앞면)이 카메라를 향하도록 연출하여 소품 방향 지침을 위반했습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 8,
     "B": 14
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S19sh4__bgfirst_bg.png",
   "bg_asset_id": "b8e8a27d-1cae-43d7-a7d7-cb9a50004dd5",
   "bg_record_key": "S19sh4::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L09B01",
    "B": "L09B02",
    "C": "L09B03",
    "D": "L09B04"
   },
   "assigned": "L09B03",
   "choice": "Candidate C",
   "confident": true,
   "reason_ko": "\"건물 외벽의 CCTV 카메라\"라는 지문에 따라, 외벽에 설치된 CCTV가 명확하게 보이는 공간이 필요함.",
   "kept": "L09B03"
  }
 },
 "S21sh3::bgfirst_bg": {
  "input_fingerprint": "35c3713c92f61fc0",
  "prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 혜수의 빈 책상 앞에 앉아 어지럽게 널린 사건 현장 사진들을 내려다보는 수리영\n\nLOCATION (lock): Inside the police office at the detective's vacant desk, covered with enlarged crime-scene photographs.\n\nTIME OF DAY (lock): late morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From the high three-quarter position beside 혜수's desk, continue the dolly-in until the camera hovers just above 수리영's seated shoulder. Arrange her bent upper body in the lower side of frame and spread the crime-scene photographs across the central desk area, letting her downward gaze lead directly into their disturbing details.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-right of the frame, foreground, looks toward scattered crime-scene photographs; scattered crime-scene photographs in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 사건 현장 확대 사진들 (혜수의 책상 위에 어지럽게 널려 있음) — Their image-bearing fronts face upward toward the camera, showing the photographed neck, shoulder, and facial injuries; used as Primary downward focal field beneath 수리영's gaze; 혜수의 책상 (주인이 자리를 비운 상태이며 사건 사진들이 놓여 있음) — Its top surface faces the high three-quarter camera position; used as Creates the high-angle plane across which the photographs and 수리영's hands are organized; 경찰서 사무실 (형사들이 회의로 빠져나가 수리영만 남아 있음); used as Provides negative space beyond the desk to emphasize that 수리영 has been left alone.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use neutral office illumination in a cool desaturated grade, with restrained contrast that keeps the printed evidence legible without stylization.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "effective_prompt": "Create the EMPTY BACKGROUND PLATE for one film shot — NO PEOPLE, no figures, no body parts, no sketch lines, no arrows anywhere.\n\"Empty\" means no people only: KEEP the location's inherent occupants and stock that define the place — animals in an animal shelter, pen or farm, goods in a market, moored boats in a harbour — unless the shot text explicitly removes them.\nThe FIRST attached image is a thin-line storyboard sketch: use ONLY its camera angle, horizon, perspective and the placement/size of buildings and set masses — ignore the sketched people and arrows entirely. The SECOND attached image (LOCATION PHOTOGRAPH) is the real place: take its architecture, materials, signage and fixed features, and RE-PROJECT them into the sketch's camera. If the photograph's camera differs from the sketch's, the sketch's camera wins.\nHUMAN-SCALE CALIBRATION: derive every structure's true size from human-scale elements — a door ≈ 2m, a window ≈ 1–1.5m wide, one storey ≈ 2.5–3m; never inflate a small structure or shrink a large one.\n\nSHOT TEXT this background must serve (Korean): 혜수의 빈 책상 앞에 앉아 어지럽게 널린 사건 현장 사진들을 내려다보는 수리영\n\nLOCATION (lock): Inside the police office at the detective's vacant desk, covered with enlarged crime-scene photographs.\n\nTIME OF DAY (lock): late morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From the high three-quarter position beside 혜수's desk, continue the dolly-in until the camera hovers just above 수리영's seated shoulder. Arrange her bent upper body in the lower side of frame and spread the crime-scene photographs across the central desk area, letting her downward gaze lead directly into their disturbing details.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-right of the frame, foreground, looks toward scattered crime-scene photographs; scattered crime-scene photographs in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 사건 현장 확대 사진들 (혜수의 책상 위에 어지럽게 널려 있음) — Their image-bearing fronts face upward toward the camera, showing the photographed neck, shoulder, and facial injuries; used as Primary downward focal field beneath 수리영's gaze; 혜수의 책상 (주인이 자리를 비운 상태이며 사건 사진들이 놓여 있음) — Its top surface faces the high three-quarter camera position; used as Creates the high-angle plane across which the photographs and 수리영's hands are organized; 경찰서 사무실 (형사들이 회의로 빠져나가 수리영만 남아 있음); used as Provides negative space beyond the desk to emphasize that 수리영 has been left alone.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use neutral office illumination in a cool desaturated grade, with restrained contrast that keeps the printed evidence legible without stylization.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. No text or watermarks.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S21sh3__bgfirst_bg.png",
  "asset_id": "e5f3cee3-9807-4497-ae9f-8ecd88206ca5",
  "input_asset_ids": [
   "d3afc3c4-c7ac-41c2-877f-014c41f0bf2c",
   "c4fb79c0-ddce-4711-a426-e55717057140"
  ]
 },
 "S21sh3": {
  "input_fingerprint": "0579fd1712391709",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 혜수의 빈 책상 앞에 앉아 어지럽게 널린 사건 현장 사진들을 내려다보는 수리영\n\nLOCATION (lock): Inside the police office at the detective's vacant desk, covered with enlarged crime-scene photographs. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From the high three-quarter position beside 혜수's desk, continue the dolly-in until the camera hovers just above 수리영's seated shoulder. Arrange her bent upper body in the lower side of frame and spread the crime-scene photographs across the central desk area, letting her downward gaze lead directly into their disturbing details.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-right of the frame, foreground, looks toward scattered crime-scene photographs; scattered crime-scene photographs in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 사건 현장 확대 사진들 (혜수의 책상 위에 어지럽게 널려 있음) — Their image-bearing fronts face upward toward the camera, showing the photographed neck, shoulder, and facial injuries; used as Primary downward focal field beneath 수리영's gaze; 혜수의 책상 (주인이 자리를 비운 상태이며 사건 사진들이 놓여 있음) — Its top surface faces the high three-quarter camera position; used as Creates the high-angle plane across which the photographs and 수리영's hands are organized; 경찰서 사무실 (형사들이 회의로 빠져나가 수리영만 남아 있음); used as Provides negative space beyond the desk to emphasize that 수리영 has been left alone.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use neutral office illumination in a cool desaturated grade, with restrained contrast that keeps the printed evidence legible without stylization.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The enlarged murder-scene photographs remain scattered across Hye-su’s desk. Su-ri-yeong continues carrying her backpack and Min-sook’s old photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 혜수의 빈 책상 앞에 앉아 어지럽게 널린 사건 현장 사진들을 내려다보는 수리영\n\nLOCATION (lock): Inside the police office at the detective's vacant desk, covered with enlarged crime-scene photographs. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From the high three-quarter position beside 혜수's desk, continue the dolly-in until the camera hovers just above 수리영's seated shoulder. Arrange her bent upper body in the lower side of frame and spread the crime-scene photographs across the central desk area, letting her downward gaze lead directly into their disturbing details.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-right of the frame, foreground, looks toward scattered crime-scene photographs; scattered crime-scene photographs in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 사건 현장 확대 사진들 (혜수의 책상 위에 어지럽게 널려 있음) — Their image-bearing fronts face upward toward the camera, showing the photographed neck, shoulder, and facial injuries; used as Primary downward focal field beneath 수리영's gaze; 혜수의 책상 (주인이 자리를 비운 상태이며 사건 사진들이 놓여 있음) — Its top surface faces the high three-quarter camera position; used as Creates the high-angle plane across which the photographs and 수리영's hands are organized; 경찰서 사무실 (형사들이 회의로 빠져나가 수리영만 남아 있음); used as Provides negative space beyond the desk to emphasize that 수리영 has been left alone.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use neutral office illumination in a cool desaturated grade, with restrained contrast that keeps the printed evidence legible without stylization.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The enlarged murder-scene photographs remain scattered across Hye-su’s desk. Su-ri-yeong continues carrying her backpack and Min-sook’s old photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 혜수의 빈 책상 앞에 앉아 어지럽게 널린 사건 현장 사진들을 내려다보는 수리영\n\nLOCATION (lock): Inside the police office at the detective's vacant desk, covered with enlarged crime-scene photographs. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From the high three-quarter position beside 혜수's desk, continue the dolly-in until the camera hovers just above 수리영's seated shoulder. Arrange her bent upper body in the lower side of frame and spread the crime-scene photographs across the central desk area, letting her downward gaze lead directly into their disturbing details.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 수리영 in the lower-right of the frame, foreground, looks toward scattered crime-scene photographs; scattered crime-scene photographs in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 사건 현장 확대 사진들 (혜수의 책상 위에 어지럽게 널려 있음) — Their image-bearing fronts face upward toward the camera, showing the photographed neck, shoulder, and facial injuries; used as Primary downward focal field beneath 수리영's gaze; 혜수의 책상 (주인이 자리를 비운 상태이며 사건 사진들이 놓여 있음) — Its top surface faces the high three-quarter camera position; used as Creates the high-angle plane across which the photographs and 수리영's hands are organized; 경찰서 사무실 (형사들이 회의로 빠져나가 수리영만 남아 있음); used as Provides negative space beyond the desk to emphasize that 수리영 has been left alone.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Use neutral office illumination in a cool desaturated grade, with restrained contrast that keeps the printed evidence legible without stylization.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The enlarged murder-scene photographs remain scattered across Hye-su’s desk. Su-ri-yeong continues carrying her backpack and Min-sook’s old photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S21sh3__bgfirst_bg.png"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/conti/conti_S21sh3.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L12B03.png"
    },
    {
     "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:874664>"
    }
   ]
  },
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "지정된 숄더뷰의 카메라 구도와 사무실 배경을 정확히 구현했으며, 요구된 배낭과 '민숙의 오래된 사진'을 들고 있는 상태까지 충실히 반영했습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "전반적인 카메라 앵글과 공간적 배경은 지침을 잘 따랐으나, 수리영이 손에 들고 있어야 할 '민숙의 오래된 사진'이 누락되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "지정된 숄더뷰의 카메라 구도와 사무실 배경을 정확히 구현했으며, 요구된 배낭과 '민숙의 오래된 사진'을 들고 있는 상태까지 충실히 반영했습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "전반적인 카메라 앵글과 공간적 배경은 지침을 잘 따랐으나, 수리영이 손에 들고 있어야 할 '민숙의 오래된 사진'이 누락되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "요청된 하이 앵글 오버 숄더 구도를 정확히 구현했으며, 배낭을 멘 채 '민숙의 옛날 사진'을 손에 쥐고 있는 유지 상태(Carried state) 지시를 완벽하게 반영했습니다."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "전반적인 구도와 배경 배치는 적절하나, 지시문에 명시된 '민숙의 옛날 사진'을 들고 있지 않아 소품 유지 상태 조건에서 감점되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 10,
      "verdict_ko": "요청된 하이 앵글 오버 숄더 구도를 정확히 구현했으며, 배낭을 멘 채 '민숙의 옛날 사진'을 손에 쥐고 있는 유지 상태(Carried state) 지시를 완벽하게 반영했습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "전반적인 구도와 배경 배치는 적절하나, 지시문에 명시된 '민숙의 옛날 사진'을 들고 있지 않아 소품 유지 상태 조건에서 감점되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 19,
     "B": 14
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   }
  },
  "totals": {
   "A": 19,
   "B": 14
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 9,
    "verdict_ko": "지정된 숄더뷰의 카메라 구도와 사무실 배경을 정확히 구현했으며, 요구된 배낭과 '민숙의 오래된 사진'을 들고 있는 상태까지 충실히 반영했습니다."
   },
   {
    "label": "B",
    "score": 7,
    "verdict_ko": "전반적인 카메라 앵글과 공간적 배경은 지침을 잘 따랐으나, 수리영이 손에 들고 있어야 할 '민숙의 오래된 사진'이 누락되었습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/episodes/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/images/background_chain/L12B03.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "캐릭터가 메고 있는 백팩의 왼쪽 어깨끈이 누락되었습니다.",
     "fix_en": "Add the black left strap of the backpack over the character's left shoulder, matching the right side.",
     "unfixable": false
    },
    {
     "issue_ko": "배경 원본에 있던 책상 위 사건 현장 사진들의 구체적인 이미지와 배열이 유지되지 않고 다르게 변형되었습니다.",
     "fix_en": "Restore the exact imagery and arrangement of the crime scene photographs on the desk to match the provided background reference.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Crop and scale the existing frame to bring the foreground desk boundaries and screen occupancy as close as possible to the shot-background reference while preserving the current viewpoint.",
     "issue_ko": "선택 이미지의 시야가 기준 배경보다 더 넓고 책상 상판과 전면부가 화면에서 더 작고 높게 배치되어, 고정된 기준 카메라와 원근 구도가 일치하지 않는다.",
     "unfixable": true
    },
    {
     "fix_en": "Turn her head and lower her gaze toward the central field of scattered crime-scene photographs while keeping the held photograph and all other staging unchanged.",
     "issue_ko": "수리영의 시선이 책상 중앙의 사건 현장 사진들이 아니라 손에 든 오래된 사진으로 향해 있어 지정된 시선 축이 어긋난다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the black left strap of the backpack over the character's left shoulder, matching the right side.\n- Restore the exact imagery and arrangement of the crime scene photographs on the desk to match the provided background reference.\n- Turn her head and lower her gaze toward the central field of scattered crime-scene photographs while keeping the held photograph and all other staging unchanged.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "스케치에 제시된 머리 각도와 시선 방향을 정확하게 반영했으며, 배경과 소품 등 제약 조건을 충실히 구현했습니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "전반적인 구성은 우수하나, 고개가 스케치보다 측면으로 더 돌아가 있어 인물의 자세가 다소 다르게 표현되었습니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "스케치에 제시된 머리 각도와 시선 방향을 정확하게 반영했으며, 배경과 소품 등 제약 조건을 충실히 구현했습니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "전반적인 구성은 우수하나, 고개가 스케치보다 측면으로 더 돌아가 있어 인물의 자세가 다소 다르게 표현되었습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "레이아웃 스케치에 제시된 측면 얼굴 각도와 책상을 향한 시선을 정확히 구현했으며, 캐릭터의 정체성도 잘 나타납니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "지정된 소품과 배경은 잘 유지되었으나, 고개가 과도하게 돌아가 있어 스케치의 얼굴 각도 및 시선 처리와 다소 차이가 있습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "레이아웃 스케치에 제시된 측면 얼굴 각도와 책상을 향한 시선을 정확히 구현했으며, 캐릭터의 정체성도 잘 나타납니다."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지정된 소품과 배경은 잘 유지되었으나, 고개가 과도하게 돌아가 있어 스케치의 얼굴 각도 및 시선 처리와 다소 차이가 있습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 14,
     "B": 14
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": false,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S21sh3__bgfirst_bg.png",
   "bg_asset_id": "e5f3cee3-9807-4497-ae9f-8ecd88206ca5",
   "bg_record_key": "S21sh3::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  },
  "plate_select": {
   "candidates": {
    "A": "L12B01",
    "B": "L12B02",
    "C": "L12B03"
   },
   "assigned": "L12B02",
   "choice": "Candidate B",
   "confident": true,
   "reason_ko": "\"혜수의 빈 책상 앞에 앉아\"라는 지문을 통해 경찰서 내부의 사무 공간(책상)이 필요함을 알 수 있으며, 현재 할당된 Candidate B가 이를 잘 보여줍니다.",
   "kept": "L12B02"
  }
 },
 "S21sh5::variants": {
  "author_fp": "10862c47cb7a4540",
  "author": {
   "variants": [
    {
     "approach_ko": "가까운 시점과 극도로 얕은 심도로 충격에 얼어붙은 표정과 고립감을 밀착해 포착한다.",
     "prompt_en": "Use an intimate, close-perspective lens feel with extremely shallow depth, resolving the tension around her widened eyes, locked jaw, and bloodless stillness while the photograph at the lower edge falls gently out of focus. Let cool late-morning daylight rake softly across one side of her face, leaving restrained falloff on the other; reduce the empty office behind her to subdued shapes and muted tonal layers, making her isolation feel immediate and airless."
    },
    {
     "approach_ko": "압축된 관찰자적 시선과 정돈된 층위로 굳어 버린 얼굴을 차갑고 객관적으로 응시한다.",
     "prompt_en": "Use a more compressed, observational lens feel, holding her face with clinical steadiness while flattening the desk-side layers into a controlled composition. Keep enough definition in the clenched photograph along the lower edge to register its physical paper texture and bent grip without revealing it fully, then let the empty office recede into soft but structured depth. Favor broad, even daylight with modest contrast and small natural eye reflections, creating a cold suspended beat rather than an expressive chiaroscuro."
    }
   ]
  },
  "reused": false
 },
 "S21sh5": {
  "input_fingerprint": "38927b062c0561ba",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 사진을 쥔 채 사색이 되어 두 눈을 크게 뜬 수리영의 굳은 얼굴\n\nLOCATION (lock): At the photo-covered detective's desk inside the police office, where the daughter studies the gruesome evidence at close range. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: End the desk-side arc at 수리영's seated eye level in a tight three-quarter facial frame, keeping her rigid face and widened eyes sharply resolved. Her clenched photograph enters only along the lower edge as contextual evidence, while open office space remains soft behind her in the direction the camera will next track.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: 수리영이 쥔 사건 현장 사진 (수리영의 손에 움켜쥐어져 있음) — The image-bearing front is partly turned toward 수리영 and only partially visible to the camera; used as A limited lower-frame cue linking her shock to what she has just examined; 비어 있는 경찰서 사무실 (형사들이 회의로 빠져나간 상태); used as Soft background context that reinforces her isolation before attention moves to 김형사의 desk.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Keep the light cool, desaturated, and restrained, using modest facial contrast to hold the rigidity of 수리영's shocked expression.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same daylight office lighting, the cluttered desk, scattered crime-scene photographs, clothing, and seated position. Exclude other officers and most of the wider office, retaining only enough desk and photographs behind the frightened face to preserve the setting.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The murder-scene photographs remain spread over the desk as Su-ri-yeong holds one, and she still has her backpack and Min-sook’s photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 사진을 쥔 채 사색이 되어 두 눈을 크게 뜬 수리영의 굳은 얼굴\n\nLOCATION (lock): At the photo-covered detective's desk inside the police office, where the daughter studies the gruesome evidence at close range. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: End the desk-side arc at 수리영's seated eye level in a tight three-quarter facial frame, keeping her rigid face and widened eyes sharply resolved. Her clenched photograph enters only along the lower edge as contextual evidence, while open office space remains soft behind her in the direction the camera will next track.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: 수리영이 쥔 사건 현장 사진 (수리영의 손에 움켜쥐어져 있음) — The image-bearing front is partly turned toward 수리영 and only partially visible to the camera; used as A limited lower-frame cue linking her shock to what she has just examined; 비어 있는 경찰서 사무실 (형사들이 회의로 빠져나간 상태); used as Soft background context that reinforces her isolation before attention moves to 김형사의 desk.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Keep the light cool, desaturated, and restrained, using modest facial contrast to hold the rigidity of 수리영's shocked expression.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same daylight office lighting, the cluttered desk, scattered crime-scene photographs, clothing, and seated position. Exclude other officers and most of the wider office, retaining only enough desk and photographs behind the frightened face to preserve the setting.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The murder-scene photographs remain spread over the desk as Su-ri-yeong holds one, and she still has her backpack and Min-sook’s photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse an intimate, close-perspective lens feel with extremely shallow depth, resolving the tension around her widened eyes, locked jaw, and bloodless stillness while the photograph at the lower edge falls gently out of focus. Let cool late-morning daylight rake softly across one side of her face, leaving restrained falloff on the other; reduce the empty office behind her to subdued shapes and muted tonal layers, making her isolation feel immediate and airless.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 사진을 쥔 채 사색이 되어 두 눈을 크게 뜬 수리영의 굳은 얼굴\n\nLOCATION (lock): At the photo-covered detective's desk inside the police office, where the daughter studies the gruesome evidence at close range. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: End the desk-side arc at 수리영's seated eye level in a tight three-quarter facial frame, keeping her rigid face and widened eyes sharply resolved. Her clenched photograph enters only along the lower edge as contextual evidence, while open office space remains soft behind her in the direction the camera will next track.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 수리영 in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: 수리영이 쥔 사건 현장 사진 (수리영의 손에 움켜쥐어져 있음) — The image-bearing front is partly turned toward 수리영 and only partially visible to the camera; used as A limited lower-frame cue linking her shock to what she has just examined; 비어 있는 경찰서 사무실 (형사들이 회의로 빠져나간 상태); used as Soft background context that reinforces her isolation before attention moves to 김형사의 desk.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Keep the light cool, desaturated, and restrained, using modest facial contrast to hold the rigidity of 수리영's shocked expression.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same daylight office lighting, the cluttered desk, scattered crime-scene photographs, clothing, and seated position. Exclude other officers and most of the wider office, retaining only enough desk and photographs behind the frightened face to preserve the setting.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The murder-scene photographs remain spread over the desk as Su-ri-yeong holds one, and she still has her backpack and Min-sook’s photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 21세 여성, 타원형의 젊은 얼굴, 긴 검은 생머리, 평소 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nUse a more compressed, observational lens feel, holding her face with clinical steadiness while flattening the desk-side layers into a controlled composition. Keep enough definition in the clenched photograph along the lower edge to register its physical paper texture and bent grip without revealing it fully, then let the empty office recede into soft but structured depth. Favor broad, even daylight with modest contrast and small natural eye reflections, creating a cold suspended beat rather than an expressive chiaroscuro.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "B": 9,
   "A": 6
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 9,
    "verdict_ko": "클로즈업 앵글, 크게 뜬 두 눈과 굳은 표정, 화면 하단 가장자리에 걸쳐진 사진 등 프레이밍과 소품 지시를 매우 정확히 구현했습니다."
   },
   {
    "label": "A",
    "score": 6,
    "verdict_ko": "표정과 배경은 적절하나, 들고 있는 사진이 화면 하단 가장자리가 아닌 중앙을 가리고 있으며 이미지 면이 전혀 보이지 않는 백지로 묘사되었습니다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S21sh3_sel.png"
   },
   {
    "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:874664>"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "수리영이 손에 아무것도 쥐지 않은 채 빈 주먹을 쥐고 있으며, 사진은 손과 분리되어 화면 좌측 하단에 떠 있습니다.",
     "fix_en": "Redraw the hand to physically grip the photograph at the bottom left, removing the empty fist from the chest area.",
     "unfixable": false
    },
    {
     "issue_ko": "지시문에 수리영이 '사건 현장 사진'을 움켜쥐고 있어야 한다고 명시되었으나, 현재 화면에 보이는 사진은 노부부의 사진입니다.",
     "fix_en": "Replace the image on the held photograph with one of the gruesome crime scene photos.",
     "unfixable": false
    },
    {
     "issue_ko": "레퍼런스 이미지에서 후드티 어깨 위로 보이던 검은색 백팩 끈이 누락되었습니다.",
     "fix_en": "Add the thick black backpack straps over her shoulders, resting on top of the navy hoodie.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Lower the held photograph so that only a narrow, partially visible portion enters along the bottom edge, preserving the face and all other framing.",
     "issue_ko": "손에 든 사진이 좌하단에서 너무 크게 솟아올라 제한적인 하단 프레임 단서가 아니라 두드러진 전경 요소가 되었다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redraw the hand to physically grip the photograph at the bottom left, removing the empty fist from the chest area.\n- Replace the image on the held photograph with one of the gruesome crime scene photos.\n- Add the thick black backpack straps over her shoulders, resting on top of the navy hoodie.\n- Lower the held photograph so that only a narrow, partially visible portion enters along the bottom edge, preserving the face and all other framing.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 2,
      "verdict_ko": "이전 샷에 있던 배낭이 누락되었으며, 가슴 부근에 몸통과 연결되지 않은 기형적인 주먹이 생성되어 신체 구조상 치명적인 오류(Hard Violation)가 발생했습니다."
     },
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "지시된 클로즈업 앵글과 사색이 된 표정을 매우 충실하게 구현했으며, 이전 샷에서 이어지는 배낭 끈과 책상 위 사진들의 디테일까지 정확하게 유지한 우수한 결과물입니다."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 2,
      "verdict_ko": "이전 샷에 있던 배낭이 누락되었으며, 가슴 부근에 몸통과 연결되지 않은 기형적인 주먹이 생성되어 신체 구조상 치명적인 오류(Hard Violation)가 발생했습니다."
     },
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "지시된 클로즈업 앵글과 사색이 된 표정을 매우 충실하게 구현했으며, 이전 샷에서 이어지는 배낭 끈과 책상 위 사진들의 디테일까지 정확하게 유지한 우수한 결과물입니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "요구된 클로즈업 구도와 두 눈을 크게 뜬 굳은 표정을 잘 살렸으며, 지시된 대로 백팩을 멘 상태와 사건 현장 사진을 쥐고 있는 모습을 정확히 구현했습니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "표정과 카메라 구도는 좋으나, 이전 샷에서 유지되어야 할 백팩이 보이지 않고, 프롬프트에서 명시한 '사건 현장 사진'이 아닌 인물 사진을 들고 있어 디테일 묘사에서 감점되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "요구된 클로즈업 구도와 두 눈을 크게 뜬 굳은 표정을 잘 살렸으며, 지시된 대로 백팩을 멘 상태와 사건 현장 사진을 쥐고 있는 모습을 정확히 구현했습니다."
     },
     {
      "label": "A",
      "score": 6,
      "verdict_ko": "표정과 카메라 구도는 좋으나, 이전 샷에서 유지되어야 할 백팩이 보이지 않고, 프롬프트에서 명시한 '사건 현장 사진'이 아닌 인물 사진을 들고 있어 디테일 묘사에서 감점되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 8,
     "B": 18
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S21sh3"
  }
 },
 "S21sh8::variants": {
  "author_fp": "213eaa7e80bf73ec",
  "author": {
   "variants": [
    {
     "approach_ko": "정면의 화면 전체를 균일하고 예리하게 포착해 좌표 발견의 수사적 명료함을 강조한다.",
     "prompt_en": "Photograph the display with restrained procedural precision: a perfectly settled frontal insert, edge-to-edge clarity across the map and identifying fields, and only the narrow bezel anchoring it as a physical monitor. Keep exposure even and cool, with crisp screen detail and controlled contrast so the isolated crimson point commands attention without blooming."
    },
    {
     "approach_ko": "실제 모니터의 표면감과 미세한 광학 흔적을 살려 붉은 좌표점에 불길한 긴장감을 집중한다.",
     "prompt_en": "Treat the insert as a tense optical encounter with a real screen: hold the square-on composition while rendering faint pixel structure, subtle glass sheen, and a gentle falloff toward the peripheral bezel, never compromising readability. Let cool late-morning ambient light skim the monitor surface softly as the red offshore point remains the sharpest, most saturated visual beat."
    }
   ]
  },
  "reused": false
 },
 "S21sh8": {
  "input_fingerprint": "22c00156462f7606",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 바다 위 특정 해상 좌표가 붉은 점으로 표시된 모니터 지도 화면 클로즈업\n\nLOCATION (lock): Inside the police office at the duty detective's computer monitor, displaying an offshore last-location map. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Terminate the frontal dolly-in square to 김형사의 monitor, with the maritime map and red coordinate marker spanning the frame while a thin monitor boundary remains visible. Center the offshore point and keep the identifying fields—강민숙's name, last-location coordinates, timestamp, source, and offline status—readable around it.\n- FRAMING SCALE: insert close-up on a detail\n- FRAME LAYOUT: monitor tracking-result display in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: 김형사의 모니터 지도 화면 (위치추적 결과가 표시되어 있음) — The active front display faces the camera nearly square-on, showing an offshore map, a red coordinate point, 강민숙's tracking information, and Status: offline; used as The sole informational focus, presenting the location discovery through a device display; 모니터 테두리 (화면 가장자리에 일부 보임) — Its front bezel is aligned nearly perpendicular to the camera axis; used as A narrow peripheral edge establishes that the map is being viewed on a monitor.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Let the active display remain clear within a cool desaturated treatment, using the documented red coordinate point as the selective crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The monitor continues displaying Min-sook’s last phone location at the offshore coordinates. Su-ri-yeong still carries her backpack and the old photograph; immediately afterward she copies the coordinates onto a note and pockets it.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 바다 위 특정 해상 좌표가 붉은 점으로 표시된 모니터 지도 화면 클로즈업\n\nLOCATION (lock): Inside the police office at the duty detective's computer monitor, displaying an offshore last-location map. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Terminate the frontal dolly-in square to 김형사의 monitor, with the maritime map and red coordinate marker spanning the frame while a thin monitor boundary remains visible. Center the offshore point and keep the identifying fields—강민숙's name, last-location coordinates, timestamp, source, and offline status—readable around it.\n- FRAMING SCALE: insert close-up on a detail\n- FRAME LAYOUT: monitor tracking-result display in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: 김형사의 모니터 지도 화면 (위치추적 결과가 표시되어 있음) — The active front display faces the camera nearly square-on, showing an offshore map, a red coordinate point, 강민숙's tracking information, and Status: offline; used as The sole informational focus, presenting the location discovery through a device display; 모니터 테두리 (화면 가장자리에 일부 보임) — Its front bezel is aligned nearly perpendicular to the camera axis; used as A narrow peripheral edge establishes that the map is being viewed on a monitor.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Let the active display remain clear within a cool desaturated treatment, using the documented red coordinate point as the selective crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The monitor continues displaying Min-sook’s last phone location at the offshore coordinates. Su-ri-yeong still carries her backpack and the old photograph; immediately afterward she copies the coordinates onto a note and pockets it.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nPhotograph the display with restrained procedural precision: a perfectly settled frontal insert, edge-to-edge clarity across the map and identifying fields, and only the narrow bezel anchoring it as a physical monitor. Keep exposure even and cool, with crisp screen detail and controlled contrast so the isolated crimson point commands attention without blooming.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 바다 위 특정 해상 좌표가 붉은 점으로 표시된 모니터 지도 화면 클로즈업\n\nLOCATION (lock): Inside the police office at the duty detective's computer monitor, displaying an offshore last-location map. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: Terminate the frontal dolly-in square to 김형사의 monitor, with the maritime map and red coordinate marker spanning the frame while a thin monitor boundary remains visible. Center the offshore point and keep the identifying fields—강민숙's name, last-location coordinates, timestamp, source, and offline status—readable around it.\n- FRAMING SCALE: insert close-up on a detail\n- FRAME LAYOUT: monitor tracking-result display in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: 김형사의 모니터 지도 화면 (위치추적 결과가 표시되어 있음) — The active front display faces the camera nearly square-on, showing an offshore map, a red coordinate point, 강민숙's tracking information, and Status: offline; used as The sole informational focus, presenting the location discovery through a device display; 모니터 테두리 (화면 가장자리에 일부 보임) — Its front bezel is aligned nearly perpendicular to the camera axis; used as A narrow peripheral edge establishes that the map is being viewed on a monitor.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Let the active display remain clear within a cool desaturated treatment, using the documented red coordinate point as the selective crimson accent.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The monitor continues displaying Min-sook’s last phone location at the offshore coordinates. Su-ri-yeong still carries her backpack and the old photograph; immediately afterward she copies the coordinates onto a note and pockets it.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.\n\nVISUAL APPROACH — this take (camera, framing, composition):\nTreat the insert as a tense optical encounter with a real screen: hold the square-on composition while rendering faint pixel structure, subtle glass sheen, and a gentle falloff toward the peripheral bezel, never compromising readability. Let cool late-morning ambient light skim the monitor surface softly as the red offshore point remains the sharpest, most saturated visual beat.\n\nPRIORITY OF INSTRUCTIONS: the VISUAL APPROACH above governs camera,\nframing and composition ONLY. Every other instruction in this prompt —\nlocation, time of day, people and their identity, poses, props,\ncarried state, physical rules, and the no-text rule — is binding\nexactly as written earlier. If the VISUAL APPROACH appears to conflict\nwith any of them, the earlier instruction wins."
  },
  "totals": {
   "A": 8,
   "B": 3
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 8,
    "verdict_ko": "요청된 모든 정보(강민숙, 오프라인 등)를 화면 UI 내에 한국어로 완벽하게 구현했으며, 모니터 테두리를 남긴 클로즈업 구도도 정확합니다."
   },
   {
    "label": "B",
    "score": 3,
    "verdict_ko": "프롬프트의 지시문 텍스트가 화면 겉면에 캡션처럼 그대로 인쇄되어 나오는 치명적인 오류(주석/텍스트 누출)가 있어 감점되었습니다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/88362757-1436-49d1-91c3-4bcc56fad1b2/images/5ecb3f04-00aa-436b-9ae7-232c4aa9f150/scene/recipe/S21sh5_sel.png"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "화면 상단의 창 제목과 탭, 그리고 지도 위에 의미 없는 한글(‘자찰의 타잭’ 등)이 생성되었습니다.",
     "fix_en": "Remove the gibberish text from the window title bar, tab, and map geography, leaving only the requested tracking information panels.",
     "unfixable": false
    }
   ]
  },
  "composition_critique": {
   "issues": [
    {
     "fix_en": "Apply a subtle perspective correction so the monitor face and bezel appear square-on, with their edges parallel to the image frame, while preserving the existing crop and content placement.",
     "issue_ko": "모니터가 정면에 수직으로 놓인 구도보다 약간 비스듬히 보여 화면과 베젤에 원근 왜곡이 생겼다.",
     "unfixable": false
    },
    {
     "fix_en": "Crop the narrow office sliver beyond the monitor’s right edge so only the thin bezel remains around the display.",
     "issue_ko": "오른쪽 베젤 바깥의 사무실 배경이 보여 모니터 화면만을 정보적 초점으로 삼는 밀착 인서트 구도에서 벗어난다.",
     "unfixable": false
    }
   ]
  },
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the gibberish text from the window title bar, tab, and map geography, leaving only the requested tracking information panels.\n- Apply a subtle perspective correction so the monitor face and bezel appear square-on, with their edges parallel to the image frame, while preserving the existing crop and content placement.\n- Crop the narrow office sliver beyond the monitor’s right edge so only the thin bezel remains around the display.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "요청된 모니터 화면과 텍스트 필드를 완벽하게 구현했으며, 불필요하게 왜곡된 글자가 생성되지 않아 더 깔끔함."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지시된 텍스트 필드를 훌륭하게 구현했으나, 상단 바에 요청하지 않은 의미 불명의 글자가 포함되어 아쉬움."
     }
    ]
   },
   "forward_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "요청된 모니터 화면과 텍스트 필드를 완벽하게 구현했으며, 불필요하게 왜곡된 글자가 생성되지 않아 더 깔끔함."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "지시된 텍스트 필드를 훌륭하게 구현했으나, 상단 바에 요청하지 않은 의미 불명의 글자가 포함되어 아쉬움."
     }
    ]
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 10,
      "verdict_ko": "요청된 텍스트(이름, 좌표, 시간, 상태 등)를 화면 내에 완벽하고 깔끔하게 구현했으며, 불필요한 외계어 텍스트 없이 클로즈업 샷의 구도와 조명 조건을 정확히 준수했습니다."
     },
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "전반적인 구도와 화면 UI는 프롬프트와 잘 맞으나, 창 상단과 지도 위에 의미 없는 인공지능 생성 문자(외계어)가 노출되어 디테일에서 감점되었습니다."
     }
    ]
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 10,
      "verdict_ko": "요청된 텍스트(이름, 좌표, 시간, 상태 등)를 화면 내에 완벽하고 깔끔하게 구현했으며, 불필요한 외계어 텍스트 없이 클로즈업 샷의 구도와 조명 조건을 정확히 준수했습니다."
     },
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "전반적인 구도와 화면 UI는 프롬프트와 잘 맞으나, 창 상단과 지도 위에 의미 없는 인공지능 생성 문자(외계어)가 노출되어 디테일에서 감점되었습니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 15,
     "B": 18
    },
    "ranking": [
     "B",
     "A"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "B",
   "fix_won": true,
   "policy": "fix_rejudge_v1_flip"
  },
  "ref_mode": "prev만 (배경 전용·공유 계획)",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S21sh5"
  },
  "lane_policy": "share_plan_prev_bgonly",
  "plate_select": {
   "candidates": {
    "A": "L12B01",
    "B": "L12B02",
    "C": "L12B03"
   },
   "assigned": "L12B02",
   "choice": "B",
   "confident": true,
   "reason_ko": "\"모니터 지도 화면 클로즈업\"이라는 지문에 따라 모니터가 위치한 경찰서 내부 책상 공간이 필요하므로, 해당 서브공간을 보여주는 현재 할당된 이미지를 유지합니다.",
   "chosen": "L12B02"
  }
 }
}