{
 "S1sh2::signage": {
  "fp": "2585b16902247e52",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "era_assess::7255272bbe359f0b": {
  "subjects": [
   {
    "subject_native": "현대 한국의 오래된 동네 세탁소 내부",
    "search_terms_native": [
     "한국 동네 세탁소 내부",
     "골목 세탁소 다림질 작업대",
     "오래된 세탁소 풍경",
     "동네 세탁소 인테리어"
    ],
    "language_lock_native": "한국어 검색어만 사용하고 다른 언어로 번역하거나 추가하지 마십시오.",
    "reason_ko": "한국 특유의 동네 세탁소 내부 구조(공업용 스팀 다리미, 다림질판, 비닐 씌운 의류 레일 등) 대신 서구식 빨래방이나 일반 매장 형태로 잘못 생성될 수 있습니다."
   }
  ],
  "subject_text": "낡고 생활감 있는 세탁소 내부\n낡고 생활감 있는 소형 세탁소 내부. 오래된 형광등 아래 다림질대와 카운터가 한 공간에 놓인 실용적인 작업장.",
  "identity": "canonical",
  "scope_id": "L01",
  "scope_role": "location_interior",
  "scope_sha": "7252603e3f309a9a"
 },
 "era_fail::f10504b0cc9b51d4": {
  "stage": "research",
  "subject": "현대 한국의 오래된 동네 세탁소 내부",
  "terms": [
   "한국 동네 세탁소 내부",
   "골목 세탁소 다림질 작업대",
   "오래된 세탁소 풍경",
   "동네 세탁소 인테리어"
  ],
  "status": "no_usable",
  "queries": [
   [
    "한국 동네 세탁소 내부 골목 세탁소 다림질 작업대 오래된 세탁소 풍경 동네 세탁소 인테리어",
    "현대 한국 오래된 동네 세탁소 내부 다림질 작업대"
   ]
  ],
  "candidate_urls": [
   "https://t1.daumcdn.net/news/202504/19/chosun/20250419064010520ixvb.jpg",
   "https://ojsfile.ohmynews.com/STD_IMG_FILE/2016/1017/IE002036074_STD.jpg",
   "https://blog.kakaocdn.net/dna/v3mON/btsQdBDo2MP/AAAAAAAAAAAAAAAAAAAAALhY2m82aAvxp-w93gOwNW-xtS_SmnrXjfX7QBVXRMz-/img.png?allow_ip=&allow_referer=&credential=yqXZFxpELC7KVnFOS48ylbz2pIh7yKj8&expires=1777561199&signature=UjyTFX26T2h8Fgk8Qn2mjWv82iA%3D",
   "https://ojsfile.ohmynews.com/STD_IMG_FILE/2016/1017/IE002036070_STD.jpg"
  ],
  "coarse": {
   "eligible": [],
   "chosen_index": 0,
   "reason": "종류·보임·기준을 다 만족하는 후보가 없다 — 이 라운드에선 안 고른다",
   "single_judge": true,
   "rejected_judges": {}
  },
  "verdicts": [
   {
    "index": 1,
    "object_type_match": "yes",
    "visible": true,
    "criteria_match": "unsure",
    "similarity": 95
   },
   {
    "index": 2,
    "object_type_match": "yes",
    "visible": true,
    "criteria_match": "unsure",
    "similarity": 90
   },
   {
    "index": 3,
    "object_type_match": "yes",
    "visible": true,
    "criteria_match": "unsure",
    "similarity": 95
   },
   {
    "index": 4,
    "object_type_match": "yes",
    "visible": true,
    "criteria_match": "unsure",
    "similarity": 85
   }
  ],
  "chosen_reason_ko": "",
  "attempts": 2
 },
 "S1sh2::bgfirst_bg": {
  "input_fingerprint": "9f8032d6a75580db",
  "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): 열린 세탁소 문 틈으로 가방을 멘 태욱이 빗물에 젖은 어깨를 한 채 안으로 막 한 걸음 내디딘 순간.\n\nLOCATION (lock): The interior side of the laundry shop entrance, directly across the open doorway from the rain outside. A worn fluorescent light flickers overhead.\n\nTIME OF DAY (lock): day, rainy.\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 several steps inside the shop, the camera looks obliquely across the open threshold as 태욱 enters from the left background toward the right-side interior. His backpack and rain-wet shoulder remain visible as he is caught with his weight transferring onto the forward foot; the tracking move briefly settles on that first step, leaving open space ahead for the camera's continued retreat.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 태욱 in the middle-left of the frame, midground, moves toward shop interior at frame right; open shop doorway in the upper-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입문 (open) — The threshold is seen diagonally, with the exterior behind 태욱 and the shop interior extending toward frame right; used as Frames 태욱's entry and establishes his inward direction of travel; 다림질대 (steam rising above it) — Its long side recedes obliquely in the deeper shop interior; used as Provides spatial continuity with the camera's preceding position near the work area.\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 flickering fluorescent illumination renders the shop in subdued brightness, muted worn colors, and gentle-to-moderate 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 readable writing anywhere: surfaces that would carry writing may be present, but stage any wording out of legibility — an oblique angle, distance, shallow focus. No captions, watermarks or overlay text.",
  "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): 열린 세탁소 문 틈으로 가방을 멘 태욱이 빗물에 젖은 어깨를 한 채 안으로 막 한 걸음 내디딘 순간.\n\nLOCATION (lock): The interior side of the laundry shop entrance, directly across the open doorway from the rain outside. A worn fluorescent light flickers overhead.\n\nTIME OF DAY (lock): day, rainy.\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 several steps inside the shop, the camera looks obliquely across the open threshold as 태욱 enters from the left background toward the right-side interior. His backpack and rain-wet shoulder remain visible as he is caught with his weight transferring onto the forward foot; the tracking move briefly settles on that first step, leaving open space ahead for the camera's continued retreat.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 태욱 in the middle-left of the frame, midground, moves toward shop interior at frame right; open shop doorway in the upper-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입문 (open) — The threshold is seen diagonally, with the exterior behind 태욱 and the shop interior extending toward frame right; used as Frames 태욱's entry and establishes his inward direction of travel; 다림질대 (steam rising above it) — Its long side recedes obliquely in the deeper shop interior; used as Provides spatial continuity with the camera's preceding position near the work area.\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 flickering fluorescent illumination renders the shop in subdued brightness, muted worn colors, and gentle-to-moderate 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 readable writing anywhere: surfaces that would carry writing may be present, but stage any wording out of legibility — an oblique angle, distance, shallow focus. No captions, watermarks or overlay text.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S1sh2__bgfirst_bg.png",
  "asset_id": "1bb542c1-29d7-44c4-8d77-bcd5b1051760",
  "input_asset_ids": [
   "dd91305e-d6ce-498d-8ec1-4f25ad785d83",
   "083a0d35-5206-434f-a06b-6fbb4f35eac8"
  ]
 },
 "S1sh2": {
  "input_fingerprint": "6fa5d9b138c3b65c",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 열린 세탁소 문 틈으로 가방을 멘 태욱이 빗물에 젖은 어깨를 한 채 안으로 막 한 걸음 내디딘 순간.\n\nLOCATION (lock): The interior side of the laundry shop entrance, directly across the open doorway from the rain outside. A worn fluorescent light flickers overhead. 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 several steps inside the shop, the camera looks obliquely across the open threshold as 태욱 enters from the left background toward the right-side interior. His backpack and rain-wet shoulder remain visible as he is caught with his weight transferring onto the forward foot; the tracking move briefly settles on that first step, leaving open space ahead for the camera's continued retreat.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 태욱 in the middle-left of the frame, midground, moves toward shop interior at frame right; open shop doorway in the upper-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입문 (open) — The threshold is seen diagonally, with the exterior behind 태욱 and the shop interior extending toward frame right; used as Frames 태욱's entry and establishes his inward direction of travel; 다림질대 (steam rising above it) — Its long side recedes obliquely in the deeper shop interior; used as Provides spatial continuity with the camera's preceding position near the work area.\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 flickering fluorescent illumination renders the shop in subdued brightness, muted worn colors, and gentle-to-moderate 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The laundry door is open, the old fluorescent light flickers, and steam rises above the ironing board. The old yellow umbrella is still stored beneath the counter. 태욱: He steps inside wearing a backpack, with rain soaking his shoulder.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 태욱 (한국인 남성, 20세의 젊은 얼굴, 자연스러운 얼굴선, 짙은색 짧은 머리카락, 꾸밈없는 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
  "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 — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 열린 세탁소 문 틈으로 가방을 멘 태욱이 빗물에 젖은 어깨를 한 채 안으로 막 한 걸음 내디딘 순간.\n\nLOCATION (lock): The interior side of the laundry shop entrance, directly across the open doorway from the rain outside. A worn fluorescent light flickers overhead. The shot takes place here — the FIRST attached image (SHOT BACKGROUND) is this exact place, already built: its ground, structures, horizon, materials and lighting are the finished truth of this location and must not be redesigned or replaced. No location photograph is attached — read the place from that image alone, and add no scenery, structure, vehicle or fixture that it does not already show. This lock governs the place only; the figures in the shot follow the staging and pose instructions.\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 several steps inside the shop, the camera looks obliquely across the open threshold as 태욱 enters from the left background toward the right-side interior. His backpack and rain-wet shoulder remain visible as he is caught with his weight transferring onto the forward foot; the tracking move briefly settles on that first step, leaving open space ahead for the camera's continued retreat.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 태욱 in the middle-left of the frame, midground, moves toward shop interior at frame right; open shop doorway in the upper-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입문 (open) — The threshold is seen diagonally, with the exterior behind 태욱 and the shop interior extending toward frame right; used as Frames 태욱's entry and establishes his inward direction of travel; 다림질대 (steam rising above it) — Its long side recedes obliquely in the deeper shop interior; used as Provides spatial continuity with the camera's preceding position near the work area.\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 flickering fluorescent illumination renders the shop in subdued brightness, muted worn colors, and gentle-to-moderate 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The laundry door is open, the old fluorescent light flickers, and steam rises above the ironing board. The old yellow umbrella is still stored beneath the counter. 태욱: He steps inside wearing a backpack, with rain soaking his shoulder.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 태욱 (한국인 남성, 20세의 젊은 얼굴, 자연스러운 얼굴선, 짙은색 짧은 머리카락, 꾸밈없는 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 열린 세탁소 문 틈으로 가방을 멘 태욱이 빗물에 젖은 어깨를 한 채 안으로 막 한 걸음 내디딘 순간.\n\nLOCATION (lock): The interior side of the laundry shop entrance, directly across the open doorway from the rain outside. A worn fluorescent light flickers overhead. 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 several steps inside the shop, the camera looks obliquely across the open threshold as 태욱 enters from the left background toward the right-side interior. His backpack and rain-wet shoulder remain visible as he is caught with his weight transferring onto the forward foot; the tracking move briefly settles on that first step, leaving open space ahead for the camera's continued retreat.\n- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 태욱 in the middle-left of the frame, midground, moves toward shop interior at frame right; open shop doorway in the upper-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입문 (open) — The threshold is seen diagonally, with the exterior behind 태욱 and the shop interior extending toward frame right; used as Frames 태욱's entry and establishes his inward direction of travel; 다림질대 (steam rising above it) — Its long side recedes obliquely in the deeper shop interior; used as Provides spatial continuity with the camera's preceding position near the work area.\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 flickering fluorescent illumination renders the shop in subdued brightness, muted worn colors, and gentle-to-moderate 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The laundry door is open, the old fluorescent light flickers, and steam rises above the ironing board. The old yellow umbrella is still stored beneath the counter. 태욱: He steps inside wearing a backpack, with rain soaking his shoulder.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 태욱 (한국인 남성, 20세의 젊은 얼굴, 자연스러운 얼굴선, 짙은색 짧은 머리카락, 꾸밈없는 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, and keep from the CAMERA & FRAME contract its subject scale, screen placement and key background elements exactly. For this candidate only, its camera angle and height are deliberately left open: choose a camera position distinctly different from the obvious one, as a film director picking a second setup on the same blocking. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S1sh2__bgfirst_bg.png",
     "asset_id": "1bb542c1-29d7-44c4-8d77-bcd5b1051760",
     "role": "bgfirst_bg"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/conti/conti_S1sh2.png",
     "asset_id": "dd91305e-d6ce-498d-8ec1-4f25ad785d83",
     "role": "conti_light"
    },
    {
     "label": "CHARACTER REFERENCE — 태욱: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:796111>",
     "asset_id": "1be132ff-248c-4c27-a41e-af0a9e7b52b9",
     "role": "character_ref"
    }
   ],
   "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/b1cefa45-c59c-4402-bcf2-afff78af62a5/episodes/45cce53b-1dbe-4248-b920-c8467186e5dc/images/background_chain/L01B01.png",
     "asset_id": "083a0d35-5206-434f-a06b-6fbb4f35eac8",
     "role": "location_plate"
    },
    {
     "label": "CHARACTER REFERENCE — 태욱: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:796111>",
     "asset_id": "1be132ff-248c-4c27-a41e-af0a9e7b52b9",
     "role": "character_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "judge_flip_skipped": "judge_fn_owns_order",
  "readings": [
   {
    "label": "A",
    "direction": "태욱의 몸과 시선이 화면 우측 세탁소 안쪽을 향해 이동하고 있음.",
    "built_space": "좌측의 열린 출입문, 우측의 카운터 및 다림질대, 후면 벽의 세탁기 배치가 레퍼런스 사진과 정확히 일치함.",
    "entities": "인물의 얼굴, 헤어스타일, 복장이 레퍼런스와 일치하며 백팩과 젖은 어깨가 잘 묘사됨. 노란 우산은 보이지 않음.",
    "hard_violations": [],
    "physics": "왼발이 바닥을 딛고 체중을 지탱하고 있으며 오른발이 자연스럽게 뒤따라오는 보행 자세가 안정적임. 한 손은 가방 끈을 쥐고 있음."
   },
   {
    "label": "B",
    "direction": "태욱의 몸과 시선이 화면 우측 세탁소 안쪽을 향해 이동하고 있음.",
    "built_space": "카운터 전면에 노란 우산이 걸려 있으며, 후면 세탁기 앞쪽 통로에 레퍼런스에 없는 비닐 덮인 옷걸이 행거가 임의로 추가되어 공간을 가로막음.",
    "entities": "인물의 얼굴과 복장이 레퍼런스와 일치하고 백팩과 젖은 어깨가 묘사됨. 지시된 노란 우산이 등장함.",
    "hard_violations": [
     "[gemini-pro] 세탁소 후면에 레퍼런스에 존재하지 않는 옷걸이 구조물을 임의로 추가하여 지정된 공간 배치를 위반함"
    ],
    "physics": "왼발이 바닥을 딛고 오른발의 뒤꿈치가 들려 있는 걷는 동작이 물리적으로 자연스러움."
   }
  ],
  "cross_model_order": {
   "policy": "cross_model_order_v2_slot0_fwd_slot1_rev_parallel2",
   "slots": [
    {
     "model": "gemini-pro",
     "order": "forward",
     "display_to_canonical": {
      "A": "A",
      "B": "B"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "지정된 공간 구조와 인물의 동작, 젖은 어깨 등의 디테일을 매우 사실적이고 정확하게 구현했으나, 카운터 아래의 노란 우산이 누락되었습니다."
       },
       {
        "label": "B",
        "score": 6,
        "verdict_ko": "후면 세탁기 앞 공간에 레퍼런스에 없는 옷걸이 구조물을 추가하여 공간을 왜곡했으며, 우산을 카운터 아래가 아닌 전면에 배치하여 지시를 어겼습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "태욱의 몸과 시선이 화면 우측 세탁소 안쪽을 향해 이동하고 있음.",
        "built_space": "좌측의 열린 출입문, 우측의 카운터 및 다림질대, 후면 벽의 세탁기 배치가 레퍼런스 사진과 정확히 일치함.",
        "entities": "인물의 얼굴, 헤어스타일, 복장이 레퍼런스와 일치하며 백팩과 젖은 어깨가 잘 묘사됨. 노란 우산은 보이지 않음.",
        "hard_violations": [],
        "physics": "왼발이 바닥을 딛고 체중을 지탱하고 있으며 오른발이 자연스럽게 뒤따라오는 보행 자세가 안정적임. 한 손은 가방 끈을 쥐고 있음."
       },
       {
        "label": "B",
        "direction": "태욱의 몸과 시선이 화면 우측 세탁소 안쪽을 향해 이동하고 있음.",
        "built_space": "카운터 전면에 노란 우산이 걸려 있으며, 후면 세탁기 앞쪽 통로에 레퍼런스에 없는 비닐 덮인 옷걸이 행거가 임의로 추가되어 공간을 가로막음.",
        "entities": "인물의 얼굴과 복장이 레퍼런스와 일치하고 백팩과 젖은 어깨가 묘사됨. 지시된 노란 우산이 등장함.",
        "hard_violations": [
         "세탁소 후면에 레퍼런스에 존재하지 않는 옷걸이 구조물을 임의로 추가하여 지정된 공간 배치를 위반함"
        ],
        "physics": "왼발이 바닥을 딛고 오른발의 뒤꿈치가 들려 있는 걷는 동작이 물리적으로 자연스러움."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "지정된 공간 구조와 인물의 동작, 젖은 어깨 등의 디테일을 매우 사실적이고 정확하게 구현했으나, 카운터 아래의 노란 우산이 누락되었습니다."
       },
       {
        "label": "B",
        "score": 6,
        "verdict_ko": "후면 세탁기 앞 공간에 레퍼런스에 없는 옷걸이 구조물을 추가하여 공간을 왜곡했으며, 우산을 카운터 아래가 아닌 전면에 배치하여 지시를 어겼습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "태욱의 몸과 시선이 화면 우측 세탁소 안쪽을 향해 이동하고 있음.",
        "built_space": "좌측의 열린 출입문, 우측의 카운터 및 다림질대, 후면 벽의 세탁기 배치가 레퍼런스 사진과 정확히 일치함.",
        "entities": "인물의 얼굴, 헤어스타일, 복장이 레퍼런스와 일치하며 백팩과 젖은 어깨가 잘 묘사됨. 노란 우산은 보이지 않음.",
        "hard_violations": [],
        "physics": "왼발이 바닥을 딛고 체중을 지탱하고 있으며 오른발이 자연스럽게 뒤따라오는 보행 자세가 안정적임. 한 손은 가방 끈을 쥐고 있음."
       },
       {
        "label": "B",
        "direction": "태욱의 몸과 시선이 화면 우측 세탁소 안쪽을 향해 이동하고 있음.",
        "built_space": "카운터 전면에 노란 우산이 걸려 있으며, 후면 세탁기 앞쪽 통로에 레퍼런스에 없는 비닐 덮인 옷걸이 행거가 임의로 추가되어 공간을 가로막음.",
        "entities": "인물의 얼굴과 복장이 레퍼런스와 일치하고 백팩과 젖은 어깨가 묘사됨. 지시된 노란 우산이 등장함.",
        "hard_violations": [
         "세탁소 후면에 레퍼런스에 존재하지 않는 옷걸이 구조물을 임의로 추가하여 지정된 공간 배치를 위반함"
        ],
        "physics": "왼발이 바닥을 딛고 오른발의 뒤꿈치가 들려 있는 걷는 동작이 물리적으로 자연스러움."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 8,
        "verdict_ko": "문턱에 남은 뒷발에서 실내의 앞발로 체중을 옮기는 첫걸음, 중간 왼쪽 배치와 오른쪽 진행 방향이 가장 정확하며 젖은 어깨와 배낭도 확인된다."
       },
       {
        "label": "A",
        "score": 7,
        "verdict_ko": "배낭과 젖은 어깨, 우측으로 걷는 방향 및 증기 나는 다림질대는 잘 구현했지만 이미 문턱에서 두 발 모두 멀어진 시점이고 노란 우산도 카운터 아래가 아니라 앞면에 걸려 있다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "태욱의 몸과 발은 왼쪽 출입구에서 화면 오른쪽 실내로 향하고, 시선은 진행 방향보다 약간 아래쪽인 오른쪽 바닥을 향한다. 앞발의 진행선은 카운터 왼쪽 모서리 옆 통로로 이어진다.",
        "built_space": "열린 유리 출입문 1개가 왼쪽 배경에 있고, 오른쪽에는 큰 낡은 작업 카운터 1개가 있다. 천장에는 여러 줄의 형광등이 있으며, 뒤쪽에 세탁기·건조기들과 의류 행거, 카운터 뒤의 파란 다림질대 1개 및 그 위 증기가 보인다. 태욱은 문과 카운터 사이 통로에 서 있다. 노란 우산 1개는 카운터 아래 수납 상태가 아니라 카운터 전면에 걸려 있다.",
        "entities": "태욱은 젊은 한국인 남성으로 보이며 짧은 검은 머리, 자연스러운 얼굴선, 올리브색 재킷·흰 상의·청바지·부츠가 참고 인물과 대체로 맞는다. 등에 배낭이 있고 왼쪽 어깨와 머리카락이 빗물에 뚜렷하게 젖었다. 열린 문, 형광등, 다림질대, 증기와 노란 우산이 모두 보이지만 우산의 보관 위치는 지시와 다르다.",
        "hard_violations": [],
        "physics": "오른발이 바닥에 평평하게 닿아 체중을 받고 왼발은 뒤에서 바닥 접촉을 유지해 보행 자세가 지지된다. 배낭은 양쪽 어깨끈으로 몸에 지지되어 있고, 우산은 손잡이로 카운터 전면에 걸려 있어 떠 있는 물체는 없다. 다만 두 발 모두 문턱보다 상당히 안쪽이라 막 첫발을 들인 순간보다는 다음 보행 단계처럼 읽힌다."
       },
       {
        "label": "B",
        "direction": "태욱의 얼굴과 시선, 몸통, 앞발이 모두 열린 문에서 화면 오른쪽 실내를 향한다. 진행선은 카운터 왼쪽 모서리 안쪽 통로를 향하며, 문밖으로 나가는 방향이 아니라 세탁소 안으로 들어오는 방향이 명확하다.",
        "built_space": "열린 유리 출입문 1개가 왼쪽 배경에 있고 문턱은 대각선으로 보인다. 오른쪽에는 큰 낡은 작업 카운터 1개, 천장에는 여러 줄의 형광등, 후면에는 세탁기·건조기들과 의류 행거가 있다. 카운터 위에는 다리미와 천이 있고 그 뒤 작업 영역 위로 증기가 올라오지만, 별도의 긴 다림질대 윤곽은 A보다 불분명하다. 태욱은 문턱과 카운터 사이 통로의 중간 왼쪽에 있다. 우산은 보이지 않으나 카운터 아래가 프레임에서 가려져 있어 지정된 수납 상태와 모순되지는 않는다.",
        "entities": "태욱은 젊은 한국인 남성으로 보이고 짧은 검은 머리, 체격, 올리브색 재킷·흰 상의·청바지·부츠가 참고 인물과 대체로 일치한다. 배낭을 착용했으며 어깨와 재킷 표면이 비에 젖어 광택이 난다. 열린 문, 형광등, 증기, 작업대와 세탁 설비가 보인다. 노란 우산은 보이지 않으며 읽을 수 있는 글자나 추가 인물은 없다.",
        "hard_violations": [],
        "physics": "실내의 오른발이 바닥에 완전히 닿아 앞으로 체중을 받고, 왼발은 문턱 바로 안쪽에서 뒤꿈치가 들린 채 앞부분으로 바닥을 밀어 첫걸음을 만든다. 오른손은 배낭끈을 잡고 배낭은 어깨끈에 지지된다. 모든 신체와 물체에 현실적인 접촉·지지가 있으며, 문밖에서 실내로 막 체중을 넘기는 동작이 자연스럽다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 8,
        "verdict_ko": "문턱에 남은 뒷발에서 실내의 앞발로 체중을 옮기는 첫걸음, 중간 왼쪽 배치와 오른쪽 진행 방향이 가장 정확하며 젖은 어깨와 배낭도 확인된다."
       },
       {
        "label": "B",
        "score": 7,
        "verdict_ko": "배낭과 젖은 어깨, 우측으로 걷는 방향 및 증기 나는 다림질대는 잘 구현했지만 이미 문턱에서 두 발 모두 멀어진 시점이고 노란 우산도 카운터 아래가 아니라 앞면에 걸려 있다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "B",
        "direction": "태욱의 몸과 발은 왼쪽 출입구에서 화면 오른쪽 실내로 향하고, 시선은 진행 방향보다 약간 아래쪽인 오른쪽 바닥을 향한다. 앞발의 진행선은 카운터 왼쪽 모서리 옆 통로로 이어진다.",
        "built_space": "열린 유리 출입문 1개가 왼쪽 배경에 있고, 오른쪽에는 큰 낡은 작업 카운터 1개가 있다. 천장에는 여러 줄의 형광등이 있으며, 뒤쪽에 세탁기·건조기들과 의류 행거, 카운터 뒤의 파란 다림질대 1개 및 그 위 증기가 보인다. 태욱은 문과 카운터 사이 통로에 서 있다. 노란 우산 1개는 카운터 아래 수납 상태가 아니라 카운터 전면에 걸려 있다.",
        "entities": "태욱은 젊은 한국인 남성으로 보이며 짧은 검은 머리, 자연스러운 얼굴선, 올리브색 재킷·흰 상의·청바지·부츠가 참고 인물과 대체로 맞는다. 등에 배낭이 있고 왼쪽 어깨와 머리카락이 빗물에 뚜렷하게 젖었다. 열린 문, 형광등, 다림질대, 증기와 노란 우산이 모두 보이지만 우산의 보관 위치는 지시와 다르다.",
        "hard_violations": [],
        "physics": "오른발이 바닥에 평평하게 닿아 체중을 받고 왼발은 뒤에서 바닥 접촉을 유지해 보행 자세가 지지된다. 배낭은 양쪽 어깨끈으로 몸에 지지되어 있고, 우산은 손잡이로 카운터 전면에 걸려 있어 떠 있는 물체는 없다. 다만 두 발 모두 문턱보다 상당히 안쪽이라 막 첫발을 들인 순간보다는 다음 보행 단계처럼 읽힌다."
       },
       {
        "label": "A",
        "direction": "태욱의 얼굴과 시선, 몸통, 앞발이 모두 열린 문에서 화면 오른쪽 실내를 향한다. 진행선은 카운터 왼쪽 모서리 안쪽 통로를 향하며, 문밖으로 나가는 방향이 아니라 세탁소 안으로 들어오는 방향이 명확하다.",
        "built_space": "열린 유리 출입문 1개가 왼쪽 배경에 있고 문턱은 대각선으로 보인다. 오른쪽에는 큰 낡은 작업 카운터 1개, 천장에는 여러 줄의 형광등, 후면에는 세탁기·건조기들과 의류 행거가 있다. 카운터 위에는 다리미와 천이 있고 그 뒤 작업 영역 위로 증기가 올라오지만, 별도의 긴 다림질대 윤곽은 A보다 불분명하다. 태욱은 문턱과 카운터 사이 통로의 중간 왼쪽에 있다. 우산은 보이지 않으나 카운터 아래가 프레임에서 가려져 있어 지정된 수납 상태와 모순되지는 않는다.",
        "entities": "태욱은 젊은 한국인 남성으로 보이고 짧은 검은 머리, 체격, 올리브색 재킷·흰 상의·청바지·부츠가 참고 인물과 대체로 일치한다. 배낭을 착용했으며 어깨와 재킷 표면이 비에 젖어 광택이 난다. 열린 문, 형광등, 증기, 작업대와 세탁 설비가 보인다. 노란 우산은 보이지 않으며 읽을 수 있는 글자나 추가 인물은 없다.",
        "hard_violations": [],
        "physics": "실내의 오른발이 바닥에 완전히 닿아 앞으로 체중을 받고, 왼발은 문턱 바로 안쪽에서 뒤꿈치가 들린 채 앞부분으로 바닥을 밀어 첫걸음을 만든다. 오른손은 배낭끈을 잡고 배낭은 어깨끈에 지지된다. 모든 신체와 물체에 현실적인 접촉·지지가 있으며, 문밖에서 실내로 막 체중을 넘기는 동작이 자연스럽다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": true,
   "slot_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "route": "cross_slot_agree"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 2.0,
    "B": 1.542
   },
   "adjusted": {
    "A": 2.0,
    "B": 1.292
   },
   "violations": {
    "B": [
     "[gemini-pro] 세탁소 후면에 레퍼런스에 존재하지 않는 옷걸이 구조물을 임의로 추가하여 지정된 공간 배치를 위반함"
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "agreed": true
  },
  "totals": {
   "A": 2000,
   "B": 1292
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 2000,
    "verdict_ko": "지정된 공간 구조와 인물의 동작, 젖은 어깨 등의 디테일을 매우 사실적이고 정확하게 구현했으나, 카운터 아래의 노란 우산이 누락되었습니다."
   },
   {
    "label": "B",
    "score": 1292,
    "verdict_ko": "후면 세탁기 앞 공간에 레퍼런스에 없는 옷걸이 구조물을 추가하여 공간을 왜곡했으며, 우산을 카운터 아래가 아닌 전면에 배치하여 지시를 어겼습니다.  ★위반: [gemini-pro] 세탁소 후면에 레퍼런스에 존재하지 않는 옷걸이 구조물을 임의로 추가하여 지정된 공간 배치를 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/b1cefa45-c59c-4402-bcf2-afff78af62a5/episodes/45cce53b-1dbe-4248-b920-c8467186e5dc/images/background_chain/L01B01.png",
    "asset_id": "083a0d35-5206-434f-a06b-6fbb4f35eac8",
    "role": "location_plate"
   },
   {
    "label": "CHARACTER REFERENCE — 태욱: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:796111>",
    "asset_id": "1be132ff-248c-4c27-a41e-af0a9e7b52b9",
    "role": "character_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a9aa83-a02c-740d-bc4f-11248db9d8b3",
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S1sh2__bgfirst_bg.png",
   "bg_asset_id": "1bb542c1-29d7-44c4-8d77-bcd5b1051760",
   "bg_record_key": "S1sh2::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S1sh2::cine": {
  "applied": false,
  "attempted_at": "2026-09-04T11:03:17.991624+00:00",
  "fingerprint": "bf517e5b8affafc12b64d7751779117a199efbb97bab293ca21e1305c0638497",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S1sh2_sel.png",
  "source_sha256": "acc186c587d78df65db0194850025e2848ebe73390f54c9526a3623cf55450c9",
  "error": "ReveTerminalError: reve 500 downstream_service_error: Downstream service error"
 },
 "S1sh5::signage": {
  "fp": "fa366d1ba258c9f4",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "S1sh5": {
  "input_fingerprint": "6c6c196aaeb4c03b",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 정순이 내민 낡은 노란 우산 손잡이를 향해 뻗어진 태욱의 손끝이 표면에 막 닿은 찰나.\n\nLOCATION (lock): Inside the laundry shop at the service counter, where the umbrella is passed across at close range. The worn fluorescent light flickers nearby. 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 counter height and within hand's reach, a slight high angle isolates the narrow gap between 정순's hand entering from frame left and 태욱's fingertips reaching from frame right. The dolly-in ends exactly as his fingertips touch the worn yellow umbrella handle at center, while the counter edge remains along the lower frame as a scale and spatial reference.\n- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 낡은 노란 우산 (extended between 정순 and 태욱) — The handle crosses diagonally through the central band, with 정순 holding one end and 태욱's fingertips meeting it; used as Primary contact point and emotional focus of the insert; 카운터 (positioned beneath the exchange) — Its top edge passes beneath the hands at a shallow diagonal; used as Anchors both hands in the same physical space without competing with the contact.\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: Flickering fluorescent light keeps the worn yellow of the umbrella subdued against the shop's muted, restrained palette.\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 about the place — its fixed features, wear and lighting — persists. The clothing, hair and overall look of 태욱 — who appear both in that photo and in this shot — are LOCKED to that photo. Anyone else visible in that photo is NOT in this shot: never carry their face, body or clothing onto anyone here. 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; it governs place and objects only, never who is in this shot or how the camera sees them): Carry over only the worn laundry interior surfaces, aged fixtures, flickering fluorescent illumination, and muted rainy-day color visible in the reference. Exclude the doorway-entry action and do not copy any person-specific details or transient rain-wet traces from the prior moment.\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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The laundry door remains open while the fluorescent light continues to flicker and steam lingers above the ironing board. The old yellow umbrella has been taken out from beneath the counter. 정순: She still has the shirt she was ironing and holds the old yellow umbrella out in front of her. 태욱: He still wears his backpack and has a rain-wet shoulder; his outstretched fingertips have just touched the umbrella handle.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일); 태욱 (한국인 남성, 20세의 젊은 얼굴, 자연스러운 얼굴선, 짙은색 짧은 머리카락, 꾸밈없는 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 정순이 내민 낡은 노란 우산 손잡이를 향해 뻗어진 태욱의 손끝이 표면에 막 닿은 찰나.\n\nLOCATION (lock): Inside the laundry shop at the service counter, where the umbrella is passed across at close range. The worn fluorescent light flickers nearby. 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 counter height and within hand's reach, a slight high angle isolates the narrow gap between 정순's hand entering from frame left and 태욱's fingertips reaching from frame right. The dolly-in ends exactly as his fingertips touch the worn yellow umbrella handle at center, while the counter edge remains along the lower frame as a scale and spatial reference.\n- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 낡은 노란 우산 (extended between 정순 and 태욱) — The handle crosses diagonally through the central band, with 정순 holding one end and 태욱's fingertips meeting it; used as Primary contact point and emotional focus of the insert; 카운터 (positioned beneath the exchange) — Its top edge passes beneath the hands at a shallow diagonal; used as Anchors both hands in the same physical space without competing with the contact.\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: Flickering fluorescent light keeps the worn yellow of the umbrella subdued against the shop's muted, restrained palette.\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 about the place — its fixed features, wear and lighting — persists. The clothing, hair and overall look of 태욱 — who appear both in that photo and in this shot — are LOCKED to that photo. Anyone else visible in that photo is NOT in this shot: never carry their face, body or clothing onto anyone here. 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; it governs place and objects only, never who is in this shot or how the camera sees them): Carry over only the worn laundry interior surfaces, aged fixtures, flickering fluorescent illumination, and muted rainy-day color visible in the reference. Exclude the doorway-entry action and do not copy any person-specific details or transient rain-wet traces from the prior moment.\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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The laundry door remains open while the fluorescent light continues to flicker and steam lingers above the ironing board. The old yellow umbrella has been taken out from beneath the counter. 정순: She still has the shirt she was ironing and holds the old yellow umbrella out in front of her. 태욱: He still wears his backpack and has a rain-wet shoulder; his outstretched fingertips have just touched the umbrella handle.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일); 태욱 (한국인 남성, 20세의 젊은 얼굴, 자연스러운 얼굴선, 짙은색 짧은 머리카락, 꾸밈없는 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 정순이 내민 낡은 노란 우산 손잡이를 향해 뻗어진 태욱의 손끝이 표면에 막 닿은 찰나.\n\nLOCATION (lock): Inside the laundry shop at the service counter, where the umbrella is passed across at close range. The worn fluorescent light flickers nearby. 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 counter height and within hand's reach, a slight high angle isolates the narrow gap between 정순's hand entering from frame left and 태욱's fingertips reaching from frame right. The dolly-in ends exactly as his fingertips touch the worn yellow umbrella handle at center, while the counter edge remains along the lower frame as a scale and spatial reference.\n- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 낡은 노란 우산 (extended between 정순 and 태욱) — The handle crosses diagonally through the central band, with 정순 holding one end and 태욱's fingertips meeting it; used as Primary contact point and emotional focus of the insert; 카운터 (positioned beneath the exchange) — Its top edge passes beneath the hands at a shallow diagonal; used as Anchors both hands in the same physical space without competing with the contact.\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: Flickering fluorescent light keeps the worn yellow of the umbrella subdued against the shop's muted, restrained palette.\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 about the place — its fixed features, wear and lighting — persists. The clothing, hair and overall look of 태욱 — who appear both in that photo and in this shot — are LOCKED to that photo. Anyone else visible in that photo is NOT in this shot: never carry their face, body or clothing onto anyone here. 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; it governs place and objects only, never who is in this shot or how the camera sees them): Carry over only the worn laundry interior surfaces, aged fixtures, flickering fluorescent illumination, and muted rainy-day color visible in the reference. Exclude the doorway-entry action and do not copy any person-specific details or transient rain-wet traces from the prior moment.\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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The laundry door remains open while the fluorescent light continues to flicker and steam lingers above the ironing board. The old yellow umbrella has been taken out from beneath the counter. 정순: She still has the shirt she was ironing and holds the old yellow umbrella out in front of her. 태욱: He still wears his backpack and has a rain-wet shoulder; his outstretched fingertips have just touched the umbrella handle.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일); 태욱 (한국인 남성, 20세의 젊은 얼굴, 자연스러운 얼굴선, 짙은색 짧은 머리카락, 꾸밈없는 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, and keep from the CAMERA & FRAME contract its subject scale, screen placement and key background elements exactly. For this candidate only, its camera angle and height are deliberately left open: choose a camera position distinctly different from the obvious one, as a film director picking a second setup on the same blocking. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "A",
    "direction": "태욱의 오른손 검지손가락이 우산의 나무 손잡이를 향해 뻗어 있으며, 표면에 정확히 닿아 있습니다.",
    "built_space": "화면 하단에 카운터가 위치하며, 배경에는 세탁기, 다리미판, 텀블러, 피어오르는 김 등 레퍼런스 이미지의 세탁소 내부 요소들이 정확한 위치에 배치되어 있습니다.",
    "entities": "프레임 왼쪽에서 들어온 정순의 손(나이 든 피부)이 낡은 노란 우산의 몸통을 쥐고 있고, 오른쪽에서 뻗어온 태욱의 손(물에 젖은 점퍼 소매)이 우산 손잡이에 닿아 있습니다.",
    "hard_violations": [],
    "physics": "우산은 정순의 왼손에 의해 단단히 쥐어져 공중에 지탱되고 있으며, 태욱의 손끝이 가볍게 맞닿은 물리적 상태가 자연스럽습니다."
   },
   {
    "label": "B",
    "direction": "정순이 우산 손잡이를 잡고 있으며, 태욱의 팔은 우산의 천 부분을 향해 뻗어 쥐고 있습니다.",
    "built_space": "카운터와 배경의 세탁기 등은 기준과 유사하나, 요구된 카메라 구도(손과 손잡이 중심의 인서트)를 무시하고 인물들의 상반신과 얼굴 하단이 노출될 정도로 프레임이 넓어졌습니다.",
    "entities": "정순과 태욱의 몸통 일부가 보이며, 노란 우산의 형태가 다소 왜곡되어 있습니다.",
    "hard_violations": [
     "[gemini-pro] 지정된 인서트 클로즈업 구도를 무시하고 미디엄 샷으로 변경됨",
     "[gemini-pro] 우측 하단에 우산 끝부분을 쥐고 있는 태욱의 또 다른 손(세 번째 손)이 등장함",
     "[gemini-pro] 우산 천 부분에서 정체불명의 노란색 다리 구조물이 자라나 카운터 바닥을 짚고 있음",
     "[gpt] 참고 장소의 고정된 서비스 카운터를 큰 개구부와 여러 수도꼭지·배관이 있는 세척대로 바꾸어, 존재하지 않던 고정 설비를 발명했다."
    ],
    "physics": "우산 천에서 뻗어 나온 불가능한 다리 형태의 구조물이 카운터 표면에 닿아 우산을 비정상적으로 지탱하고 있습니다."
   }
  ],
  "cross_model_order": {
   "policy": "cross_model_order_v2_slot0_fwd_slot1_rev_parallel2",
   "slots": [
    {
     "model": "gemini-pro",
     "order": "forward",
     "display_to_canonical": {
      "A": "A",
      "B": "B"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 10,
        "verdict_ko": "프롬프트가 요구한 정확한 인서트 클로즈업 구도를 완벽하게 따랐으며, 두 사람의 손과 우산 손잡이가 맞닿는 찰나의 순간과 이전 샷의 배경 디테일을 훌륭하게 구현했습니다."
       },
       {
        "label": "B",
        "score": 0,
        "verdict_ko": "인서트 클로즈업 지시를 무시하고 인물의 상반신을 노출했으며, 화면 우측 하단에 태욱의 세 번째 손이 등장하는 치명적인 오류가 있습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "태욱의 오른손 검지손가락이 우산의 나무 손잡이를 향해 뻗어 있으며, 표면에 정확히 닿아 있습니다.",
        "built_space": "화면 하단에 카운터가 위치하며, 배경에는 세탁기, 다리미판, 텀블러, 피어오르는 김 등 레퍼런스 이미지의 세탁소 내부 요소들이 정확한 위치에 배치되어 있습니다.",
        "entities": "프레임 왼쪽에서 들어온 정순의 손(나이 든 피부)이 낡은 노란 우산의 몸통을 쥐고 있고, 오른쪽에서 뻗어온 태욱의 손(물에 젖은 점퍼 소매)이 우산 손잡이에 닿아 있습니다.",
        "hard_violations": [],
        "physics": "우산은 정순의 왼손에 의해 단단히 쥐어져 공중에 지탱되고 있으며, 태욱의 손끝이 가볍게 맞닿은 물리적 상태가 자연스럽습니다."
       },
       {
        "label": "B",
        "direction": "정순이 우산 손잡이를 잡고 있으며, 태욱의 팔은 우산의 천 부분을 향해 뻗어 쥐고 있습니다.",
        "built_space": "카운터와 배경의 세탁기 등은 기준과 유사하나, 요구된 카메라 구도(손과 손잡이 중심의 인서트)를 무시하고 인물들의 상반신과 얼굴 하단이 노출될 정도로 프레임이 넓어졌습니다.",
        "entities": "정순과 태욱의 몸통 일부가 보이며, 노란 우산의 형태가 다소 왜곡되어 있습니다.",
        "hard_violations": [
         "지정된 인서트 클로즈업 구도를 무시하고 미디엄 샷으로 변경됨",
         "우측 하단에 우산 끝부분을 쥐고 있는 태욱의 또 다른 손(세 번째 손)이 등장함",
         "우산 천 부분에서 정체불명의 노란색 다리 구조물이 자라나 카운터 바닥을 짚고 있음"
        ],
        "physics": "우산 천에서 뻗어 나온 불가능한 다리 형태의 구조물이 카운터 표면에 닿아 우산을 비정상적으로 지탱하고 있습니다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 10,
        "verdict_ko": "프롬프트가 요구한 정확한 인서트 클로즈업 구도를 완벽하게 따랐으며, 두 사람의 손과 우산 손잡이가 맞닿는 찰나의 순간과 이전 샷의 배경 디테일을 훌륭하게 구현했습니다."
       },
       {
        "label": "B",
        "score": 0,
        "verdict_ko": "인서트 클로즈업 지시를 무시하고 인물의 상반신을 노출했으며, 화면 우측 하단에 태욱의 세 번째 손이 등장하는 치명적인 오류가 있습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "태욱의 오른손 검지손가락이 우산의 나무 손잡이를 향해 뻗어 있으며, 표면에 정확히 닿아 있습니다.",
        "built_space": "화면 하단에 카운터가 위치하며, 배경에는 세탁기, 다리미판, 텀블러, 피어오르는 김 등 레퍼런스 이미지의 세탁소 내부 요소들이 정확한 위치에 배치되어 있습니다.",
        "entities": "프레임 왼쪽에서 들어온 정순의 손(나이 든 피부)이 낡은 노란 우산의 몸통을 쥐고 있고, 오른쪽에서 뻗어온 태욱의 손(물에 젖은 점퍼 소매)이 우산 손잡이에 닿아 있습니다.",
        "hard_violations": [],
        "physics": "우산은 정순의 왼손에 의해 단단히 쥐어져 공중에 지탱되고 있으며, 태욱의 손끝이 가볍게 맞닿은 물리적 상태가 자연스럽습니다."
       },
       {
        "label": "B",
        "direction": "정순이 우산 손잡이를 잡고 있으며, 태욱의 팔은 우산의 천 부분을 향해 뻗어 쥐고 있습니다.",
        "built_space": "카운터와 배경의 세탁기 등은 기준과 유사하나, 요구된 카메라 구도(손과 손잡이 중심의 인서트)를 무시하고 인물들의 상반신과 얼굴 하단이 노출될 정도로 프레임이 넓어졌습니다.",
        "entities": "정순과 태욱의 몸통 일부가 보이며, 노란 우산의 형태가 다소 왜곡되어 있습니다.",
        "hard_violations": [
         "지정된 인서트 클로즈업 구도를 무시하고 미디엄 샷으로 변경됨",
         "우측 하단에 우산 끝부분을 쥐고 있는 태욱의 또 다른 손(세 번째 손)이 등장함",
         "우산 천 부분에서 정체불명의 노란색 다리 구조물이 자라나 카운터 바닥을 짚고 있음"
        ],
        "physics": "우산 천에서 뻗어 나온 불가능한 다리 형태의 구조물이 카운터 표면에 닿아 우산을 비정상적으로 지탱하고 있습니다."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 3,
        "verdict_ko": "태욱의 손끝이 손잡이에 막 닿는 순간이 아니라 양손으로 우산 천과 끝부분을 이미 붙잡은 장면이며, 삽입 클로즈업보다 훨씬 넓고 기존 카운터에 없던 수도 설비까지 추가됐다."
       },
       {
        "label": "B",
        "score": 9,
        "verdict_ko": "카운터 높이의 약한 하이앵글 삽입 클로즈업으로, 왼쪽 정순의 손이 내민 낡은 노란 우산 손잡이에 오른쪽 태욱의 손끝이 중앙에서 막 닿는 결정적 순간을 정확히 포착했다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정순의 왼손은 화면 왼쪽에서 우산의 굽은 손잡이를 잡고 오른쪽으로 내민다. 그러나 태욱의 왼손은 손잡이가 아니라 접힌 우산 천과 살대 부근을 움켜쥐고 있고, 오른손은 우산 반대쪽 끝을 잡는다. 따라서 태욱의 뻗은 손끝이 정순이 내민 손잡이에 닿는다는 지정 목표 관계가 보이지 않는다. 얼굴과 시선은 대부분 잘려 있어 확인할 수 없다.",
        "built_space": "세탁소 내부와 형광등, 뒤쪽 세탁기, 낡은 작업면은 보이지만 프레임이 양쪽 인물의 상반신까지 포함해 지정된 디테일 삽입보다 넓다. 아래에는 카운터 가장자리가 있으나, 기존의 단단한 서비스 카운터 대신 중앙에 큰 개구부와 여러 수도꼭지·배관이 달린 세척대 형태가 추가되어 참고 장소의 고정 구조와 맞지 않는다. 정순은 왼쪽, 태욱은 오른쪽에 서 있지만 우산 위로 손이 총 세 개 들어와 지정된 단순한 양손 교환 구도가 흐려진다.",
        "entities": "정순은 한국인 중년 여성로 보이는 일부 얼굴과 팔, 회색 상의와 갈색 앞치마로 나타나 참고 인물의 연령·성별·복장에 대체로 맞는다. 태욱은 젊은 한국인 남성의 체격, 짙은 녹색 재킷과 배낭으로 보이지만 얼굴은 제외되어 정확한 정체성 확인이 어렵다. 우산은 낡고 누렇게 바랜 접이식 장우산이지만, 태욱이 접촉해야 할 손잡이가 정순의 손 안에만 있고 접촉점이 잘못됐다.",
        "hard_violations": [
         "참고 장소의 고정된 서비스 카운터를 큰 개구부와 여러 수도꼭지·배관이 있는 세척대로 바꾸어, 존재하지 않던 고정 설비를 발명했다."
        ],
        "physics": "우산은 정순이 손잡이를 잡고 태욱이 접힌 천 부분과 반대쪽 끝을 각각 잡아 세 손으로 지지하므로 공중에 무지지 상태로 뜨지는 않는다. 각 손의 접촉 자체는 물리적으로 가능하지만, 이 자세는 손끝이 막 닿는 순간이 아니라 태욱이 이미 양손으로 우산을 확보한 뒤의 상태다."
       },
       {
        "label": "B",
        "direction": "정순의 손은 화면 왼쪽에서 접힌 우산 몸통을 잡아 굽은 손잡이를 오른쪽으로 내민다. 태욱의 손은 화면 오른쪽에서 왼쪽으로 뻗으며 검지 끝이 화면 중앙의 노란 손잡이 윗면에 막 닿는다. 지정된 출발 방향, 접촉 목표와 접촉 순간이 모두 분명하다. 얼굴과 시선은 삽입 프레이밍상 보이지 않는다.",
        "built_space": "카메라는 서비스 카운터 높이에서 약간 내려다보며, 낡은 카운터 윗면과 앞쪽 가장자리가 하단을 얕은 사선으로 지난다. 정순의 팔은 왼쪽, 태욱의 팔은 오른쪽에서 들어오고 손 사이의 좁은 간격과 중앙 접촉점이 격리된다. 배경에는 흐릿한 세탁기, 형광등, 다리미와 증기가 실제 거리와 크기에 맞게 작게 남아 있으며, 참고 세탁소의 노후한 재료와 조명을 유지한다. 반사나 구조적으로 불가능한 배치는 보이지 않는다.",
        "entities": "왼쪽 손과 팔은 주름과 체격이 자연스러운 50대 여성의 것으로 보이고 갈색 앞치마·회색 소매가 정순의 참고 복장과 맞는다. 오른쪽 손과 팔은 젊은 남성의 것으로 보이며 태욱의 짙은 녹색 방수 재킷과 빗물이 묻은 소매가 이어진다. 우산은 얼룩과 마모가 뚜렷한 낡은 노란 우산이며 굽은 손잡이의 형태도 명확하다. 프레이밍에 맞게 얼굴, 배낭과 셔츠는 보이지 않으며 다른 인물이나 읽을 수 있는 글자는 없다.",
        "hard_violations": [],
        "physics": "우산은 정순의 왼손이 접힌 천과 축 부근을 단단히 감싸 지지하고 있으며 일부는 카운터 가까이에 놓여 있다. 태욱의 검지는 손잡이에 가볍게 닿기만 하므로 ‘막 닿은 찰나’에 맞고, 손과 손목의 뻗는 각도도 자연스럽다. 떠 있거나 지지되지 않은 물체와 불가능한 관절 자세는 보이지 않는다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 3,
        "verdict_ko": "태욱의 손끝이 손잡이에 막 닿는 순간이 아니라 양손으로 우산 천과 끝부분을 이미 붙잡은 장면이며, 삽입 클로즈업보다 훨씬 넓고 기존 카운터에 없던 수도 설비까지 추가됐다."
       },
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "카운터 높이의 약한 하이앵글 삽입 클로즈업으로, 왼쪽 정순의 손이 내민 낡은 노란 우산 손잡이에 오른쪽 태욱의 손끝이 중앙에서 막 닿는 결정적 순간을 정확히 포착했다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "B",
        "direction": "정순의 왼손은 화면 왼쪽에서 우산의 굽은 손잡이를 잡고 오른쪽으로 내민다. 그러나 태욱의 왼손은 손잡이가 아니라 접힌 우산 천과 살대 부근을 움켜쥐고 있고, 오른손은 우산 반대쪽 끝을 잡는다. 따라서 태욱의 뻗은 손끝이 정순이 내민 손잡이에 닿는다는 지정 목표 관계가 보이지 않는다. 얼굴과 시선은 대부분 잘려 있어 확인할 수 없다.",
        "built_space": "세탁소 내부와 형광등, 뒤쪽 세탁기, 낡은 작업면은 보이지만 프레임이 양쪽 인물의 상반신까지 포함해 지정된 디테일 삽입보다 넓다. 아래에는 카운터 가장자리가 있으나, 기존의 단단한 서비스 카운터 대신 중앙에 큰 개구부와 여러 수도꼭지·배관이 달린 세척대 형태가 추가되어 참고 장소의 고정 구조와 맞지 않는다. 정순은 왼쪽, 태욱은 오른쪽에 서 있지만 우산 위로 손이 총 세 개 들어와 지정된 단순한 양손 교환 구도가 흐려진다.",
        "entities": "정순은 한국인 중년 여성로 보이는 일부 얼굴과 팔, 회색 상의와 갈색 앞치마로 나타나 참고 인물의 연령·성별·복장에 대체로 맞는다. 태욱은 젊은 한국인 남성의 체격, 짙은 녹색 재킷과 배낭으로 보이지만 얼굴은 제외되어 정확한 정체성 확인이 어렵다. 우산은 낡고 누렇게 바랜 접이식 장우산이지만, 태욱이 접촉해야 할 손잡이가 정순의 손 안에만 있고 접촉점이 잘못됐다.",
        "hard_violations": [
         "참고 장소의 고정된 서비스 카운터를 큰 개구부와 여러 수도꼭지·배관이 있는 세척대로 바꾸어, 존재하지 않던 고정 설비를 발명했다."
        ],
        "physics": "우산은 정순이 손잡이를 잡고 태욱이 접힌 천 부분과 반대쪽 끝을 각각 잡아 세 손으로 지지하므로 공중에 무지지 상태로 뜨지는 않는다. 각 손의 접촉 자체는 물리적으로 가능하지만, 이 자세는 손끝이 막 닿는 순간이 아니라 태욱이 이미 양손으로 우산을 확보한 뒤의 상태다."
       },
       {
        "label": "A",
        "direction": "정순의 손은 화면 왼쪽에서 접힌 우산 몸통을 잡아 굽은 손잡이를 오른쪽으로 내민다. 태욱의 손은 화면 오른쪽에서 왼쪽으로 뻗으며 검지 끝이 화면 중앙의 노란 손잡이 윗면에 막 닿는다. 지정된 출발 방향, 접촉 목표와 접촉 순간이 모두 분명하다. 얼굴과 시선은 삽입 프레이밍상 보이지 않는다.",
        "built_space": "카메라는 서비스 카운터 높이에서 약간 내려다보며, 낡은 카운터 윗면과 앞쪽 가장자리가 하단을 얕은 사선으로 지난다. 정순의 팔은 왼쪽, 태욱의 팔은 오른쪽에서 들어오고 손 사이의 좁은 간격과 중앙 접촉점이 격리된다. 배경에는 흐릿한 세탁기, 형광등, 다리미와 증기가 실제 거리와 크기에 맞게 작게 남아 있으며, 참고 세탁소의 노후한 재료와 조명을 유지한다. 반사나 구조적으로 불가능한 배치는 보이지 않는다.",
        "entities": "왼쪽 손과 팔은 주름과 체격이 자연스러운 50대 여성의 것으로 보이고 갈색 앞치마·회색 소매가 정순의 참고 복장과 맞는다. 오른쪽 손과 팔은 젊은 남성의 것으로 보이며 태욱의 짙은 녹색 방수 재킷과 빗물이 묻은 소매가 이어진다. 우산은 얼룩과 마모가 뚜렷한 낡은 노란 우산이며 굽은 손잡이의 형태도 명확하다. 프레이밍에 맞게 얼굴, 배낭과 셔츠는 보이지 않으며 다른 인물이나 읽을 수 있는 글자는 없다.",
        "hard_violations": [],
        "physics": "우산은 정순의 왼손이 접힌 천과 축 부근을 단단히 감싸 지지하고 있으며 일부는 카운터 가까이에 놓여 있다. 태욱의 검지는 손잡이에 가볍게 닿기만 하므로 ‘막 닿은 찰나’에 맞고, 손과 손목의 뻗는 각도도 자연스럽다. 떠 있거나 지지되지 않은 물체와 불가능한 관절 자세는 보이지 않는다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": true,
   "slot_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "route": "cross_slot_agree"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 2.0,
    "B": 0.333
   },
   "adjusted": {
    "A": 2.0,
    "B": 0.083
   },
   "violations": {
    "B": [
     "[gemini-pro] 지정된 인서트 클로즈업 구도를 무시하고 미디엄 샷으로 변경됨",
     "[gemini-pro] 우측 하단에 우산 끝부분을 쥐고 있는 태욱의 또 다른 손(세 번째 손)이 등장함",
     "[gemini-pro] 우산 천 부분에서 정체불명의 노란색 다리 구조물이 자라나 카운터 바닥을 짚고 있음",
     "[gpt] 참고 장소의 고정된 서비스 카운터를 큰 개구부와 여러 수도꼭지·배관이 있는 세척대로 바꾸어, 존재하지 않던 고정 설비를 발명했다."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "agreed": true
  },
  "totals": {
   "A": 2000,
   "B": 83
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 2000,
    "verdict_ko": "프롬프트가 요구한 정확한 인서트 클로즈업 구도를 완벽하게 따랐으며, 두 사람의 손과 우산 손잡이가 맞닿는 찰나의 순간과 이전 샷의 배경 디테일을 훌륭하게 구현했습니다."
   },
   {
    "label": "B",
    "score": 83,
    "verdict_ko": "인서트 클로즈업 지시를 무시하고 인물의 상반신을 노출했으며, 화면 우측 하단에 태욱의 세 번째 손이 등장하는 치명적인 오류가 있습니다.  ★위반: [gemini-pro] 지정된 인서트 클로즈업 구도를 무시하고 미디엄 샷으로 변경됨 / [gemini-pro] 우측 하단에 우산 끝부분을 쥐고 있는 태욱의 또 다른 손(세 번째 손)이 등장함 / [gemini-pro] 우산 천 부분에서 정체불명의 노란색 다리 구조물이 자라나 카운터 바닥을 짚고 있음 / [gpt] 참고 장소의 고정된 서비스 카운터를 큰 개구부와 여러 수도꼭지·배관이 있는 세척대로 바꾸어, 존재하지 않던 고정 설비를 발명했다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features and lighting mood are LOCKED to this photo; never copy its camera framing. The clothing, hair and overall look of 태욱 — who appear both in that photo and in this shot — are LOCKED to that photo. Anyone else visible in that photo is NOT in this shot: never carry their face, body or clothing onto anyone here. 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/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S1sh2_sel.png",
    "asset_id": "5c796d49-fc50-4bf0-b78a-da21ef26f9c0",
    "role": "prev_still"
   },
   {
    "label": "CHARACTER REFERENCE — 정순: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:721445>",
    "asset_id": "c709a52a-fbe5-4309-a3de-85e738c118c7",
    "role": "character_ref"
   },
   {
    "label": "CHARACTER REFERENCE — 태욱: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:796111>",
    "asset_id": "1be132ff-248c-4c27-a41e-af0a9e7b52b9",
    "role": "character_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a9aa8f-a3ee-70e3-aa0c-214f791149c4",
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S1sh2"
  }
 },
 "S1sh5::cine": {
  "applied": false,
  "attempted_at": "2026-09-04T11:04:51.036588+00:00",
  "fingerprint": "769f20c3a8fef442e62aa1acf9ce8ba94d124387c71a2d8df752a625272219dd",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S1sh5_sel.png",
  "source_sha256": "98ca641054ed410404de4089aef18b11bd6e64f714517f003e02ecb9b3b8dac7",
  "error": "ReveTerminalError: reve 500 downstream_service_error: Downstream service error"
 },
 "S2sh2::signage": {
  "fp": "c5db234ae8a5d1ec",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "era_assess::7b6e176edd2a6c1d": {
  "subjects": [
   {
    "subject_native": "현대 한국 주택가의 골목길 세탁소 앞 담벼락",
    "search_terms_native": [
     "동네 세탁소 골목",
     "골목길 세탁소 외관",
     "주택가 골목 담벼락"
    ],
    "language_lock_native": "반드시 한국어로만 검색해야 하며 다른 언어로 번역하거나 추가하지 마십시오.",
    "reason_ko": "한국 특유의 주택가 골목 구조와 동네 세탁소의 외관, 낡은 시멘트 담벼락 질감을 구현하기 위해 참조가 필요함."
   }
  ],
  "subject_text": "세탁소 앞 물때 낀 담벼락 사이의 좁은 골목\n물때가 번진 담벼락 사이의 좁은 세탁소 앞 골목. 낮의 흐린 자연광과 빗물에 젖은 바닥이 보이는 비좁은 통로.",
  "identity": "canonical",
  "scope_id": "L02",
  "scope_role": "location_exterior",
  "scope_sha": "f9cc17441032f20b"
 },
 "era_fail::d2ad3e0de6ec0b1c": {
  "stage": "research",
  "subject": "현대 한국 주택가의 골목길 세탁소 앞 담벼락",
  "terms": [
   "동네 세탁소 골목",
   "골목길 세탁소 외관",
   "주택가 골목 담벼락"
  ],
  "status": "no_usable",
  "queries": [
   [
    "동네 세탁소 골목 골목길 세탁소 외관 주택가 골목 담벼락",
    "현대 한국 주택가 골목길 세탁소 앞 담벼락"
   ]
  ],
  "candidate_urls": [
   "https://image.cine21.com/resize/cine21/movie/2025/1002/15_43_58__68de1f2e094d2%5BH800-%5D.jpg",
   "https://img4.daumcdn.net/thumb/R658x0.q70/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fnews%2F202506%2F15%2Fned%2F20250615100015757pthn.jpg",
   "https://cdn.imweb.me/upload/S202001067fd13fb34339a/b56f0820b1007.jpg",
   "https://map2.daumcdn.net/map_roadview/2023/08/9134735/2_101775_9134735_20230802023324_cube/back_1200.jpg"
  ],
  "coarse": {
   "eligible": [],
   "chosen_index": 0,
   "reason": "종류·보임·기준을 다 만족하는 후보가 없다 — 이 라운드에선 안 고른다",
   "single_judge": true,
   "rejected_judges": {}
  },
  "verdicts": [
   {
    "index": 1,
    "object_type_match": "no",
    "visible": false,
    "criteria_match": "unsure",
    "similarity": 60
   },
   {
    "index": 2,
    "object_type_match": "no",
    "visible": false,
    "criteria_match": "unsure",
    "similarity": 50
   },
   {
    "index": 3,
    "object_type_match": "no",
    "visible": false,
    "criteria_match": "unsure",
    "similarity": 50
   },
   {
    "index": 4,
    "object_type_match": "no",
    "visible": false,
    "criteria_match": "unsure",
    "similarity": 50
   }
  ],
  "chosen_reason_ko": "",
  "attempts": 2
 },
 "S2sh2::bgfirst_bg": {
  "input_fingerprint": "d276a7338e37fe77",
  "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): 세탁소 문가에 멈춰 선 정순이 한 손에 다림질하던 천을 늘어뜨려 쥔 채 골목 쪽을 가만히 바라보는 상체 클로즈업.\n\nLOCATION (lock): Outside at the laundry shop doorway, where the entrance opens directly onto the narrow, rain-puddled alley between stained walls.\n\nTIME OF DAY (lock): day, rainy.\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: Just outside the doorway at chest height, the completed push-in holds 정순 in an oblique three-quarter view on the right side of frame, with the alley direction retained as negative space to her left. Her upper body occupies most of the frame, but the composition drops low enough to preserve the hand loosely gripping the hanging shirt as she remains absorbed in watching 태욱's departure.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 정순 in the middle-right of the frame, foreground, looks toward alley exit at frame left; alley direction in the middle-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입구 (정순 paused at the threshold) — The entrance edge runs behind 정순 while opening the composition toward the alley on frame left; used as Creates a boundary between her stillness and 태욱's direction of departure; 다림질하던 셔츠 (hanging from one hand); used as Retained in the lower frame as evidence that 정순 stopped working to watch him leave; 좁은 골목 (rainwater pooled along it) — The visible stretch recedes away from the doorway into the left-side gaze space; used as Provides the off-screen direction of 정순's attention; 담벼락 (water marks risen along its surface) — Its alley-facing plane recedes through the background beside 정순's gaze line; used as Adds tactile depth to the narrow alley without distracting from her expression.\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: Daytime ambient light is subdued by the rainy setting, maintaining muted colors, restrained brightness, and gentle-to-moderate 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 readable writing anywhere: surfaces that would carry writing may be present, but stage any wording out of legibility — an oblique angle, distance, shallow focus. No captions, watermarks or overlay text.",
  "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): 세탁소 문가에 멈춰 선 정순이 한 손에 다림질하던 천을 늘어뜨려 쥔 채 골목 쪽을 가만히 바라보는 상체 클로즈업.\n\nLOCATION (lock): Outside at the laundry shop doorway, where the entrance opens directly onto the narrow, rain-puddled alley between stained walls.\n\nTIME OF DAY (lock): day, rainy.\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: Just outside the doorway at chest height, the completed push-in holds 정순 in an oblique three-quarter view on the right side of frame, with the alley direction retained as negative space to her left. Her upper body occupies most of the frame, but the composition drops low enough to preserve the hand loosely gripping the hanging shirt as she remains absorbed in watching 태욱's departure.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 정순 in the middle-right of the frame, foreground, looks toward alley exit at frame left; alley direction in the middle-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입구 (정순 paused at the threshold) — The entrance edge runs behind 정순 while opening the composition toward the alley on frame left; used as Creates a boundary between her stillness and 태욱's direction of departure; 다림질하던 셔츠 (hanging from one hand); used as Retained in the lower frame as evidence that 정순 stopped working to watch him leave; 좁은 골목 (rainwater pooled along it) — The visible stretch recedes away from the doorway into the left-side gaze space; used as Provides the off-screen direction of 정순's attention; 담벼락 (water marks risen along its surface) — Its alley-facing plane recedes through the background beside 정순's gaze line; used as Adds tactile depth to the narrow alley without distracting from her expression.\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: Daytime ambient light is subdued by the rainy setting, maintaining muted colors, restrained brightness, and gentle-to-moderate 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 readable writing anywhere: surfaces that would carry writing may be present, but stage any wording out of legibility — an oblique angle, distance, shallow focus. No captions, watermarks or overlay text.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S2sh2__bgfirst_bg.png",
  "asset_id": "7b13a4d1-cd23-4050-85e0-fe0e3bac9556",
  "input_asset_ids": [
   "3f1d6ec5-ed97-44c1-a0e2-072f73de16d6",
   "ebcf0501-39d0-4ae5-a60b-1ba6335060fc"
  ]
 },
 "S2sh2": {
  "input_fingerprint": "7112795de7411a96",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 세탁소 문가에 멈춰 선 정순이 한 손에 다림질하던 천을 늘어뜨려 쥔 채 골목 쪽을 가만히 바라보는 상체 클로즈업.\n\nLOCATION (lock): Outside at the laundry shop doorway, where the entrance opens directly onto the narrow, rain-puddled alley between stained walls. 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: Just outside the doorway at chest height, the completed push-in holds 정순 in an oblique three-quarter view on the right side of frame, with the alley direction retained as negative space to her left. Her upper body occupies most of the frame, but the composition drops low enough to preserve the hand loosely gripping the hanging shirt as she remains absorbed in watching 태욱's departure.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 정순 in the middle-right of the frame, foreground, looks toward alley exit at frame left; alley direction in the middle-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입구 (정순 paused at the threshold) — The entrance edge runs behind 정순 while opening the composition toward the alley on frame left; used as Creates a boundary between her stillness and 태욱's direction of departure; 다림질하던 셔츠 (hanging from one hand); used as Retained in the lower frame as evidence that 정순 stopped working to watch him leave; 좁은 골목 (rainwater pooled along it) — The visible stretch recedes away from the doorway into the left-side gaze space; used as Provides the off-screen direction of 정순's attention; 담벼락 (water marks risen along its surface) — Its alley-facing plane recedes through the background beside 정순's gaze line; used as Adds tactile depth to the narrow alley without distracting from her expression.\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: Daytime ambient light is subdued by the rainy setting, maintaining muted colors, restrained brightness, and gentle-to-moderate 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Rainwater is pooled in the narrow alley, water stains mark the wall, and the laundry doorway remains open. The student ID left behind by Taewook is already resting on the counter inside the laundry. 정순: She stands at the laundry doorway, gazing toward the alley while holding the same ironing shirt loosely in one hand.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
  "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 — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 세탁소 문가에 멈춰 선 정순이 한 손에 다림질하던 천을 늘어뜨려 쥔 채 골목 쪽을 가만히 바라보는 상체 클로즈업.\n\nLOCATION (lock): Outside at the laundry shop doorway, where the entrance opens directly onto the narrow, rain-puddled alley between stained walls. The shot takes place here — the FIRST attached image (SHOT BACKGROUND) is this exact place, already built: its ground, structures, horizon, materials and lighting are the finished truth of this location and must not be redesigned or replaced. No location photograph is attached — read the place from that image alone, and add no scenery, structure, vehicle or fixture that it does not already show. This lock governs the place only; the figures in the shot follow the staging and pose instructions.\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: Just outside the doorway at chest height, the completed push-in holds 정순 in an oblique three-quarter view on the right side of frame, with the alley direction retained as negative space to her left. Her upper body occupies most of the frame, but the composition drops low enough to preserve the hand loosely gripping the hanging shirt as she remains absorbed in watching 태욱's departure.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 정순 in the middle-right of the frame, foreground, looks toward alley exit at frame left; alley direction in the middle-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입구 (정순 paused at the threshold) — The entrance edge runs behind 정순 while opening the composition toward the alley on frame left; used as Creates a boundary between her stillness and 태욱's direction of departure; 다림질하던 셔츠 (hanging from one hand); used as Retained in the lower frame as evidence that 정순 stopped working to watch him leave; 좁은 골목 (rainwater pooled along it) — The visible stretch recedes away from the doorway into the left-side gaze space; used as Provides the off-screen direction of 정순's attention; 담벼락 (water marks risen along its surface) — Its alley-facing plane recedes through the background beside 정순's gaze line; used as Adds tactile depth to the narrow alley without distracting from her expression.\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: Daytime ambient light is subdued by the rainy setting, maintaining muted colors, restrained brightness, and gentle-to-moderate 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Rainwater is pooled in the narrow alley, water stains mark the wall, and the laundry doorway remains open. The student ID left behind by Taewook is already resting on the counter inside the laundry. 정순: She stands at the laundry doorway, gazing toward the alley while holding the same ironing shirt loosely in one hand.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): day, rainy.\n\nSHOT TEXT (authoritative, Korean): 세탁소 문가에 멈춰 선 정순이 한 손에 다림질하던 천을 늘어뜨려 쥔 채 골목 쪽을 가만히 바라보는 상체 클로즈업.\n\nLOCATION (lock): Outside at the laundry shop doorway, where the entrance opens directly onto the narrow, rain-puddled alley between stained walls. 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: Just outside the doorway at chest height, the completed push-in holds 정순 in an oblique three-quarter view on the right side of frame, with the alley direction retained as negative space to her left. Her upper body occupies most of the frame, but the composition drops low enough to preserve the hand loosely gripping the hanging shirt as she remains absorbed in watching 태욱's departure.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 정순 in the middle-right of the frame, foreground, looks toward alley exit at frame left; alley direction in the middle-left of the frame, background.\n- KEY BACKGROUND ELEMENTS: 세탁소 출입구 (정순 paused at the threshold) — The entrance edge runs behind 정순 while opening the composition toward the alley on frame left; used as Creates a boundary between her stillness and 태욱's direction of departure; 다림질하던 셔츠 (hanging from one hand); used as Retained in the lower frame as evidence that 정순 stopped working to watch him leave; 좁은 골목 (rainwater pooled along it) — The visible stretch recedes away from the doorway into the left-side gaze space; used as Provides the off-screen direction of 정순's attention; 담벼락 (water marks risen along its surface) — Its alley-facing plane recedes through the background beside 정순's gaze line; used as Adds tactile depth to the narrow alley without distracting from her expression.\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: Daytime ambient light is subdued by the rainy setting, maintaining muted colors, restrained brightness, and gentle-to-moderate 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Rainwater is pooled in the narrow alley, water stains mark the wall, and the laundry doorway remains open. The student ID left behind by Taewook is already resting on the counter inside the laundry. 정순: She stands at the laundry doorway, gazing toward the alley while holding the same ironing shirt loosely in one hand.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, and keep from the CAMERA & FRAME contract its subject scale, screen placement and key background elements exactly. For this candidate only, its camera angle and height are deliberately left open: choose a camera position distinctly different from the obvious one, as a film director picking a second setup on the same blocking. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "roll_refs": {
   "A": [
    {
     "label": "SHOT BACKGROUND",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S2sh2__bgfirst_bg.png",
     "asset_id": "7b13a4d1-cd23-4050-85e0-fe0e3bac9556",
     "role": "bgfirst_bg"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/conti/conti_S2sh2.png",
     "asset_id": "3f1d6ec5-ed97-44c1-a0e2-072f73de16d6",
     "role": "conti_light"
    },
    {
     "label": "CHARACTER REFERENCE — 정순: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:721445>",
     "asset_id": "c709a52a-fbe5-4309-a3de-85e738c118c7",
     "role": "character_ref"
    }
   ],
   "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/b1cefa45-c59c-4402-bcf2-afff78af62a5/episodes/45cce53b-1dbe-4248-b920-c8467186e5dc/images/background_chain/L02B01.png",
     "asset_id": "ebcf0501-39d0-4ae5-a60b-1ba6335060fc",
     "role": "location_plate"
    },
    {
     "label": "CHARACTER REFERENCE — 정순: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:721445>",
     "asset_id": "c709a52a-fbe5-4309-a3de-85e738c118c7",
     "role": "character_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "judge_flip_skipped": "judge_fn_owns_order",
  "readings": [
   {
    "label": "A",
    "direction": "정순은 프레임 왼쪽의 비 오는 좁은 골목을 향해 시선을 고정하고 있습니다.",
    "built_space": "오른쪽에 세탁소 출입구와 내부 공간(카운터 위에 놓인 학생증 포함)이 보이며, 왼쪽으로는 빗물이 고인 골목과 담벼락이 깊이감 있게 배치되어 카메라 구도와 완벽하게 부합합니다.",
    "entities": "인물 레퍼런스와 외모가 일치하는 정순이 한 손(오른손)에 파란색 셔츠를 쥐고 있습니다.",
    "hard_violations": [],
    "physics": "정순은 바닥에 안정적으로 서 있으며, 오른손으로 셔츠를 단단히 쥐어 중력에 의해 자연스럽게 늘어지도록 지탱하고 있습니다."
   },
   {
    "label": "B",
    "direction": "정순은 프레임 왼쪽의 빗물이 고인 골목을 바라보고 있습니다.",
    "built_space": "오른쪽에 세탁소 출입구가 있고 왼쪽으로 비 오는 골목과 담벼락이 배치되어 공간적 배경은 지시문과 일치합니다.",
    "entities": "인물 레퍼런스와 외모가 일치하는 정순이 등장하나, 지시문(한 손)과 달리 두 손으로 셔츠를 잡고 있습니다.",
    "hard_violations": [
     "[gemini-pro] physically impossible anatomy (정순의 자세와 어깨 각도상 자연스럽게 연결될 수 없는 위치에서 등장한 기형적인 오른손)"
    ],
    "physics": "셔츠를 잡고 있는 두 손 중 프레임 왼쪽의 손(오른손)은 몸의 관절이나 팔과 물리적으로 연결될 수 없는 각도에 있어 구조적으로 불가능한 형태입니다."
   }
  ],
  "cross_model_order": {
   "policy": "cross_model_order_v2_slot0_fwd_slot1_rev_parallel2",
   "slots": [
    {
     "model": "gemini-pro",
     "order": "forward",
     "display_to_canonical": {
      "A": "A",
      "B": "B"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 8,
        "verdict_ko": "요구된 상체 클로즈업보다 다소 넓은 화각(미디엄 샷)이지만, 한 손에 셔츠를 늘어뜨려 쥔 동작과 비 오는 골목 및 세탁소 내부(카운터 위 학생증 포함)의 환경을 해부학적 오류 없이 정확하게 구현했습니다."
       },
       {
        "label": "B",
        "score": 2,
        "verdict_ko": "상체 클로즈업 화각에는 더 근접했으나, '한 손' 지시와 달리 두 손으로 셔츠를 잡고 있으며 그 중 하나의 손이 해부학적으로 불가능하게 배치되는 치명적인 오류가 있습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정순은 프레임 왼쪽의 비 오는 좁은 골목을 향해 시선을 고정하고 있습니다.",
        "built_space": "오른쪽에 세탁소 출입구와 내부 공간(카운터 위에 놓인 학생증 포함)이 보이며, 왼쪽으로는 빗물이 고인 골목과 담벼락이 깊이감 있게 배치되어 카메라 구도와 완벽하게 부합합니다.",
        "entities": "인물 레퍼런스와 외모가 일치하는 정순이 한 손(오른손)에 파란색 셔츠를 쥐고 있습니다.",
        "hard_violations": [],
        "physics": "정순은 바닥에 안정적으로 서 있으며, 오른손으로 셔츠를 단단히 쥐어 중력에 의해 자연스럽게 늘어지도록 지탱하고 있습니다."
       },
       {
        "label": "B",
        "direction": "정순은 프레임 왼쪽의 빗물이 고인 골목을 바라보고 있습니다.",
        "built_space": "오른쪽에 세탁소 출입구가 있고 왼쪽으로 비 오는 골목과 담벼락이 배치되어 공간적 배경은 지시문과 일치합니다.",
        "entities": "인물 레퍼런스와 외모가 일치하는 정순이 등장하나, 지시문(한 손)과 달리 두 손으로 셔츠를 잡고 있습니다.",
        "hard_violations": [
         "physically impossible anatomy (정순의 자세와 어깨 각도상 자연스럽게 연결될 수 없는 위치에서 등장한 기형적인 오른손)"
        ],
        "physics": "셔츠를 잡고 있는 두 손 중 프레임 왼쪽의 손(오른손)은 몸의 관절이나 팔과 물리적으로 연결될 수 없는 각도에 있어 구조적으로 불가능한 형태입니다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 8,
        "verdict_ko": "요구된 상체 클로즈업보다 다소 넓은 화각(미디엄 샷)이지만, 한 손에 셔츠를 늘어뜨려 쥔 동작과 비 오는 골목 및 세탁소 내부(카운터 위 학생증 포함)의 환경을 해부학적 오류 없이 정확하게 구현했습니다."
       },
       {
        "label": "B",
        "score": 2,
        "verdict_ko": "상체 클로즈업 화각에는 더 근접했으나, '한 손' 지시와 달리 두 손으로 셔츠를 잡고 있으며 그 중 하나의 손이 해부학적으로 불가능하게 배치되는 치명적인 오류가 있습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "정순은 프레임 왼쪽의 비 오는 좁은 골목을 향해 시선을 고정하고 있습니다.",
        "built_space": "오른쪽에 세탁소 출입구와 내부 공간(카운터 위에 놓인 학생증 포함)이 보이며, 왼쪽으로는 빗물이 고인 골목과 담벼락이 깊이감 있게 배치되어 카메라 구도와 완벽하게 부합합니다.",
        "entities": "인물 레퍼런스와 외모가 일치하는 정순이 한 손(오른손)에 파란색 셔츠를 쥐고 있습니다.",
        "hard_violations": [],
        "physics": "정순은 바닥에 안정적으로 서 있으며, 오른손으로 셔츠를 단단히 쥐어 중력에 의해 자연스럽게 늘어지도록 지탱하고 있습니다."
       },
       {
        "label": "B",
        "direction": "정순은 프레임 왼쪽의 빗물이 고인 골목을 바라보고 있습니다.",
        "built_space": "오른쪽에 세탁소 출입구가 있고 왼쪽으로 비 오는 골목과 담벼락이 배치되어 공간적 배경은 지시문과 일치합니다.",
        "entities": "인물 레퍼런스와 외모가 일치하는 정순이 등장하나, 지시문(한 손)과 달리 두 손으로 셔츠를 잡고 있습니다.",
        "hard_violations": [
         "physically impossible anatomy (정순의 자세와 어깨 각도상 자연스럽게 연결될 수 없는 위치에서 등장한 기형적인 오른손)"
        ],
        "physics": "셔츠를 잡고 있는 두 손 중 프레임 왼쪽의 손(오른손)은 몸의 관절이나 팔과 물리적으로 연결될 수 없는 각도에 있어 구조적으로 불가능한 형태입니다."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "요구된 상체 클로즈업, 화면 오른쪽의 정순, 왼쪽 골목을 향한 시선과 여백을 가장 정확히 구현했으나 셔츠를 한 손이 아니라 두 손으로 잡고 있다."
       },
       {
        "label": "B",
        "score": 7,
        "verdict_ko": "한 손으로 셔츠를 들고 골목 왼쪽을 바라보며 문턱과 학생증까지 정확히 배치했지만, 허벅지까지 드러나는 미디엄 숏이라 권위적인 클로즈업 구도를 크게 벗어난다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정순의 얼굴과 두 눈은 화면 왼쪽으로 뻗은 골목의 먼 출구 방향을 향한다. 시선은 카메라가 아니라 태욱이 떠난 것으로 설정된 골목 방향에 자연스럽게 닿으며, 무기나 별도의 지시 물체는 없다.",
        "built_space": "화면 오른쪽에 열린 세탁소 출입구 한 곳과 그 문틀·유리문 가장자리, 정순 뒤의 문턱이 보인다. 왼쪽에는 물웅덩이가 이어진 좁은 골목 한 줄과 물때 낀 담벼락 한 면이 원근에 맞게 후퇴한다. 정순은 출입구 경계에 서 있고 카메라는 문밖 가슴 높이에서 비스듬히 본 구도다. 실내 카운터와 학생증은 좁은 프레이밍 밖이며, 유리의 약한 반사는 이 카메라 위치에서 가능하다.",
        "entities": "등장 인물은 한 명뿐이며 한국인 중년 여성으로 보이고, 정순 참고 이미지의 짙은 단정한 단발머리·중년 얼굴선·회색 티셔츠·갈색 앞치마와 잘 맞는다. 하단에는 실제 천의 주름과 무게를 가진 연한 셔츠가 보인다. 다만 한 손에 늘어뜨린다는 지시와 달리 양손이 셔츠의 서로 다른 부분을 잡고 있다. 다른 사람이나 읽을 수 있는 문구는 보이지 않는다.",
        "hard_violations": [],
        "physics": "정순의 발은 클로즈업 밖이지만 몸통은 문턱에서 수직으로 이어지는 정상적인 입식 자세이며 공중에 뜬 징후가 없다. 셔츠는 양손이 직접 잡아 지지하고 아래로 자연스럽게 처지므로 부유하지 않는다."
       },
       {
        "label": "B",
        "direction": "정순의 머리와 눈은 화면 왼쪽의 골목 끝을 향하고 있어 떠나는 태욱의 오프스크린 방향과 일치한다. 시선은 카메라나 실내가 아닌 골목 출구 쪽에 놓인다.",
        "built_space": "열린 세탁소 출입구 한 곳이 화면 오른쪽에 있고, 정순은 그 문턱에 서 있다. 실내에는 카운터 한 개, 그 위 학생증 한 장, 뒤쪽에 세탁 장비들이 보인다. 왼쪽에는 빗물이 고인 좁은 골목과 물때 낀 담벼락이 후퇴한다. 공간 관계와 출입구 방향은 참고 장소에 부합하고 불가능한 반사는 없지만, 카메라가 넓게 빠져 정순의 허벅지 부근까지 담으면서 출입구 내부가 과도하게 많이 노출된다.",
        "entities": "사람은 정순 한 명뿐이며 한국인 중년 여성의 얼굴, 짙은 단발머리, 회색 티셔츠와 갈색 앞치마가 참고 인물과 대체로 일치한다. 연한 셔츠는 한 손에 잡혀 있고 카운터 위에는 태욱의 학생증으로 읽히는 카드 한 장이 놓여 있다. 카드의 세부 글자는 판독되지 않으며 추가 인물은 없다.",
        "hard_violations": [],
        "physics": "정순의 발은 프레임 아래에 있지만 서 있는 몸의 중심과 문턱 접지가 자연스럽게 이어지고 공중 부유는 없다. 셔츠는 왼손이 확실히 움켜쥐어 지지하며 천이 아래로 처진다. 다만 느슨하게 늘어뜨렸다기보다 팔에 다소 뭉쳐 들린 상태다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 9,
        "verdict_ko": "요구된 상체 클로즈업, 화면 오른쪽의 정순, 왼쪽 골목을 향한 시선과 여백을 가장 정확히 구현했으나 셔츠를 한 손이 아니라 두 손으로 잡고 있다."
       },
       {
        "label": "A",
        "score": 7,
        "verdict_ko": "한 손으로 셔츠를 들고 골목 왼쪽을 바라보며 문턱과 학생증까지 정확히 배치했지만, 허벅지까지 드러나는 미디엄 숏이라 권위적인 클로즈업 구도를 크게 벗어난다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "B",
        "direction": "정순의 얼굴과 두 눈은 화면 왼쪽으로 뻗은 골목의 먼 출구 방향을 향한다. 시선은 카메라가 아니라 태욱이 떠난 것으로 설정된 골목 방향에 자연스럽게 닿으며, 무기나 별도의 지시 물체는 없다.",
        "built_space": "화면 오른쪽에 열린 세탁소 출입구 한 곳과 그 문틀·유리문 가장자리, 정순 뒤의 문턱이 보인다. 왼쪽에는 물웅덩이가 이어진 좁은 골목 한 줄과 물때 낀 담벼락 한 면이 원근에 맞게 후퇴한다. 정순은 출입구 경계에 서 있고 카메라는 문밖 가슴 높이에서 비스듬히 본 구도다. 실내 카운터와 학생증은 좁은 프레이밍 밖이며, 유리의 약한 반사는 이 카메라 위치에서 가능하다.",
        "entities": "등장 인물은 한 명뿐이며 한국인 중년 여성으로 보이고, 정순 참고 이미지의 짙은 단정한 단발머리·중년 얼굴선·회색 티셔츠·갈색 앞치마와 잘 맞는다. 하단에는 실제 천의 주름과 무게를 가진 연한 셔츠가 보인다. 다만 한 손에 늘어뜨린다는 지시와 달리 양손이 셔츠의 서로 다른 부분을 잡고 있다. 다른 사람이나 읽을 수 있는 문구는 보이지 않는다.",
        "hard_violations": [],
        "physics": "정순의 발은 클로즈업 밖이지만 몸통은 문턱에서 수직으로 이어지는 정상적인 입식 자세이며 공중에 뜬 징후가 없다. 셔츠는 양손이 직접 잡아 지지하고 아래로 자연스럽게 처지므로 부유하지 않는다."
       },
       {
        "label": "A",
        "direction": "정순의 머리와 눈은 화면 왼쪽의 골목 끝을 향하고 있어 떠나는 태욱의 오프스크린 방향과 일치한다. 시선은 카메라나 실내가 아닌 골목 출구 쪽에 놓인다.",
        "built_space": "열린 세탁소 출입구 한 곳이 화면 오른쪽에 있고, 정순은 그 문턱에 서 있다. 실내에는 카운터 한 개, 그 위 학생증 한 장, 뒤쪽에 세탁 장비들이 보인다. 왼쪽에는 빗물이 고인 좁은 골목과 물때 낀 담벼락이 후퇴한다. 공간 관계와 출입구 방향은 참고 장소에 부합하고 불가능한 반사는 없지만, 카메라가 넓게 빠져 정순의 허벅지 부근까지 담으면서 출입구 내부가 과도하게 많이 노출된다.",
        "entities": "사람은 정순 한 명뿐이며 한국인 중년 여성의 얼굴, 짙은 단발머리, 회색 티셔츠와 갈색 앞치마가 참고 인물과 대체로 일치한다. 연한 셔츠는 한 손에 잡혀 있고 카운터 위에는 태욱의 학생증으로 읽히는 카드 한 장이 놓여 있다. 카드의 세부 글자는 판독되지 않으며 추가 인물은 없다.",
        "hard_violations": [],
        "physics": "정순의 발은 프레임 아래에 있지만 서 있는 몸의 중심과 문턱 접지가 자연스럽게 이어지고 공중 부유는 없다. 셔츠는 왼손이 확실히 움켜쥐어 지지하며 천이 아래로 처진다. 다만 느슨하게 늘어뜨렸다기보다 팔에 다소 뭉쳐 들린 상태다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": false,
   "slot_winner": {
    "gemini-pro": "A",
    "gpt": "B"
   },
   "route": "cross_slot_combined"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 1.778,
    "B": 1.25
   },
   "adjusted": {
    "A": 1.778,
    "B": 1.0
   },
   "violations": {
    "B": [
     "[gemini-pro] physically impossible anatomy (정순의 자세와 어깨 각도상 자연스럽게 연결될 수 없는 위치에서 등장한 기형적인 오른손)"
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "B"
   },
   "agreed": false
  },
  "totals": {
   "A": 1778,
   "B": 1000
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1778,
    "verdict_ko": "요구된 상체 클로즈업보다 다소 넓은 화각(미디엄 샷)이지만, 한 손에 셔츠를 늘어뜨려 쥔 동작과 비 오는 골목 및 세탁소 내부(카운터 위 학생증 포함)의 환경을 해부학적 오류 없이 정확하게 구현했습니다."
   },
   {
    "label": "B",
    "score": 1000,
    "verdict_ko": "상체 클로즈업 화각에는 더 근접했으나, '한 손' 지시와 달리 두 손으로 셔츠를 잡고 있으며 그 중 하나의 손이 해부학적으로 불가능하게 배치되는 치명적인 오류가 있습니다.  ★위반: [gemini-pro] physically impossible anatomy (정순의 자세와 어깨 각도상 자연스럽게 연결될 수 없는 위치에서 등장한 기형적인 오른손)"
   }
  ],
  "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/b1cefa45-c59c-4402-bcf2-afff78af62a5/episodes/45cce53b-1dbe-4248-b920-c8467186e5dc/images/background_chain/L02B01.png",
    "asset_id": "ebcf0501-39d0-4ae5-a60b-1ba6335060fc",
    "role": "location_plate"
   },
   {
    "label": "CHARACTER REFERENCE — 정순: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:721445>",
    "asset_id": "c709a52a-fbe5-4309-a3de-85e738c118c7",
    "role": "character_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a9aa93-fb16-7d3b-acd4-e785038518f8",
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S2sh2__bgfirst_bg.png",
   "bg_asset_id": "7b13a4d1-cd23-4050-85e0-fe0e3bac9556",
   "bg_record_key": "S2sh2::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S2sh2::cine": {
  "applied": false,
  "attempted_at": "2026-09-04T11:09:17.061702+00:00",
  "fingerprint": "822ade0497560cd4c5379648c049374ed080b8258cc10507cf73548e4294f7a5",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S2sh2_sel.png",
  "source_sha256": "b0a16a13c2c1ee33efd370bef7893f1e09f6fffbf264cf3ed79737afe3c1c42f",
  "error": "ReveTerminalError: reve 500 downstream_service_error: Downstream service error"
 },
 "S3sh2::signage": {
  "fp": "00ae12eed09205e2",
  "inscriptions": [],
  "cues": [
   {
    "text_native": "",
    "source": "scene_text_implied",
    "source_quote": "학생증"
   }
  ],
  "dropped": []
 },
 "S3sh2": {
  "input_fingerprint": "c7126627f0290545",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 세탁소 카운터 표면 위로 인물 사진이 인쇄된 학생증 하나가 덩그러니 놓여 있는 클로즈업.\n\nLOCATION (lock): The top of the service counter inside the worn laundry shop. The interior is unevenly lit because one fluorescent light is out. 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: Just above the counter and offset to one side, the camera looks shallowly across the surface toward the lone student ID near center-left. The dolly-in settles with the portrait-bearing face legible while enough counter remains around the card to preserve its small, abandoned presence rather than enlarging it unrealistically.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: 학생증 (left alone on the counter) — The portrait-bearing face is turned toward the camera, showing the printed photograph and identification content; used as Primary focal object, isolated by surrounding counter space; 카운터 (holding the lone student ID in the visible area) — Its upper plane crosses the frame at a shallow angle beneath the card; used as Provides scale and emphasizes that the ID has been left behind.\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: With one fluorescent light out in the evening interior, the counter sits under subdued, uneven illumination with muted colors and 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\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): One fluorescent light is out, and a ladder stands beneath the light. A single student ID bearing a printed portrait lies on the counter.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person appears in frame, and no face, and no figure standing, sitting or passing through. Do not invent someone to occupy the space.\n\nOne thing this rule does NOT forbid: the hand or arm of whoever is doing what the shot text describes. If the moment is an object being held, carried, operated, read, aimed, worn or handed over, the hand doing it belongs in frame, resting on or gripping that object exactly as its real use requires. An object photographed floating unsupported, with the hand that must be holding it erased, is a worse failure than the person you were avoiding — render the hand.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 세탁소 카운터 표면 위로 인물 사진이 인쇄된 학생증 하나가 덩그러니 놓여 있는 클로즈업.\n\nLOCATION (lock): The top of the service counter inside the worn laundry shop. The interior is unevenly lit because one fluorescent light is out. 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: Just above the counter and offset to one side, the camera looks shallowly across the surface toward the lone student ID near center-left. The dolly-in settles with the portrait-bearing face legible while enough counter remains around the card to preserve its small, abandoned presence rather than enlarging it unrealistically.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: 학생증 (left alone on the counter) — The portrait-bearing face is turned toward the camera, showing the printed photograph and identification content; used as Primary focal object, isolated by surrounding counter space; 카운터 (holding the lone student ID in the visible area) — Its upper plane crosses the frame at a shallow angle beneath the card; used as Provides scale and emphasizes that the ID has been left behind.\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: With one fluorescent light out in the evening interior, the counter sits under subdued, uneven illumination with muted colors and 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\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): One fluorescent light is out, and a ladder stands beneath the light. A single student ID bearing a printed portrait lies on the counter.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person appears in frame, and no face, and no figure standing, sitting or passing through. Do not invent someone to occupy the space.\n\nOne thing this rule does NOT forbid: the hand or arm of whoever is doing what the shot text describes. If the moment is an object being held, carried, operated, read, aimed, worn or handed over, the hand doing it belongs in frame, resting on or gripping that object exactly as its real use requires. An object photographed floating unsupported, with the hand that must be holding it erased, is a worse failure than the person you were avoiding — render the hand.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 세탁소 카운터 표면 위로 인물 사진이 인쇄된 학생증 하나가 덩그러니 놓여 있는 클로즈업.\n\nLOCATION (lock): The top of the service counter inside the worn laundry shop. The interior is unevenly lit because one fluorescent light is out. 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: Just above the counter and offset to one side, the camera looks shallowly across the surface toward the lone student ID near center-left. The dolly-in settles with the portrait-bearing face legible while enough counter remains around the card to preserve its small, abandoned presence rather than enlarging it unrealistically.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: 학생증 (left alone on the counter) — The portrait-bearing face is turned toward the camera, showing the printed photograph and identification content; used as Primary focal object, isolated by surrounding counter space; 카운터 (holding the lone student ID in the visible area) — Its upper plane crosses the frame at a shallow angle beneath the card; used as Provides scale and emphasizes that the ID has been left behind.\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: With one fluorescent light out in the evening interior, the counter sits under subdued, uneven illumination with muted colors and 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\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): One fluorescent light is out, and a ladder stands beneath the light. A single student ID bearing a printed portrait lies on the counter.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person appears in frame, and no face, and no figure standing, sitting or passing through. Do not invent someone to occupy the space.\n\nOne thing this rule does NOT forbid: the hand or arm of whoever is doing what the shot text describes. If the moment is an object being held, carried, operated, read, aimed, worn or handed over, the hand doing it belongs in frame, resting on or gripping that object exactly as its real use requires. An object photographed floating unsupported, with the hand that must be holding it erased, is a worse failure than the person you were avoiding — render the hand.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, and keep from the CAMERA & FRAME contract its subject scale, screen placement and key background elements exactly. For this candidate only, its camera angle and height are deliberately left open: choose a camera position distinctly different from the obvious one, as a film director picking a second setup on the same blocking. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "A",
    "direction": "카메라는 카운터 표면을 가로질러 중앙 좌측에 홀로 놓인 학생증을 비스듬히 얕은 각도로 바라보고 있다.",
    "built_space": "화면은 낡은 나무 카운터 표면에 집중되어 있으며, 얕은 심도 너머로 펜 꽂이, 접시, 세탁기 등 레퍼런스의 공간 구성 요소들이 흐릿하게 배치되어 있다.",
    "entities": "여성의 사진이 인쇄된 학생증이 카운터 위에 덩그러니 놓여 있으며, 텍스트는 의도대로 읽을 수 없게 처리되었다.",
    "hard_violations": [
     "[gpt] 학생증 전면의 한글과 숫자 일부가 읽을 수 있을 정도로 선명해 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
    ],
    "physics": "학생증은 카운터의 평평한 표면 위에 물리적으로 자연스럽게 밀착되어 놓여 있다."
   },
   {
    "label": "B",
    "direction": "카메라는 카운터 위에 위치해 세탁소 내부 안쪽을 정면으로 깊게 바라보고 있다.",
    "built_space": "전경의 카운터부터 배경의 세탁기, 벽에 걸린 달력, 건조 중인 옷가지들, 그리고 중앙 뒤편의 사다리까지 공간 전체가 넓은 화각으로 뚜렷하게 보인다.",
    "entities": "카운터 위에는 사진이 인쇄된 학생증이 있고, 배경에는 명시된 사다리가 존재한다.",
    "hard_violations": [
     "[gpt] 학생증의 한글 및 숫자 정보와 왼쪽 달력의 문자·숫자가 읽을 수 있는 형태로 노출되어 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
    ],
    "physics": "학생증은 카운터 표면 위에 안정적으로 놓여 있다."
   }
  ],
  "cross_model_order": {
   "policy": "cross_model_order_v2_slot0_fwd_slot1_rev_parallel2",
   "slots": [
    {
     "model": "gemini-pro",
     "order": "forward",
     "display_to_canonical": {
      "A": "A",
      "B": "B"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "지시된 클로즈업 프레이밍과 얕은 카메라 앵글을 완벽하게 구현하였으며, 올바른 프레이밍에 따른 얕은 심도로 인해 배경의 사다리 등 일부 요소가 생략된 것은 평가 기준에 완벽히 부합합니다."
       },
       {
        "label": "B",
        "score": 3,
        "verdict_ko": "가장 높은 우선순위인 '클로즈업' 샷 크기 지시를 완전히 무시하고 세탁소 내부 전체를 보여주는 와이드 샷을 생성하여 연출 의도를 크게 훼손했습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "카메라는 카운터 표면을 가로질러 중앙 좌측에 홀로 놓인 학생증을 비스듬히 얕은 각도로 바라보고 있다.",
        "built_space": "화면은 낡은 나무 카운터 표면에 집중되어 있으며, 얕은 심도 너머로 펜 꽂이, 접시, 세탁기 등 레퍼런스의 공간 구성 요소들이 흐릿하게 배치되어 있다.",
        "entities": "여성의 사진이 인쇄된 학생증이 카운터 위에 덩그러니 놓여 있으며, 텍스트는 의도대로 읽을 수 없게 처리되었다.",
        "hard_violations": [],
        "physics": "학생증은 카운터의 평평한 표면 위에 물리적으로 자연스럽게 밀착되어 놓여 있다."
       },
       {
        "label": "B",
        "direction": "카메라는 카운터 위에 위치해 세탁소 내부 안쪽을 정면으로 깊게 바라보고 있다.",
        "built_space": "전경의 카운터부터 배경의 세탁기, 벽에 걸린 달력, 건조 중인 옷가지들, 그리고 중앙 뒤편의 사다리까지 공간 전체가 넓은 화각으로 뚜렷하게 보인다.",
        "entities": "카운터 위에는 사진이 인쇄된 학생증이 있고, 배경에는 명시된 사다리가 존재한다.",
        "hard_violations": [],
        "physics": "학생증은 카운터 표면 위에 안정적으로 놓여 있다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "지시된 클로즈업 프레이밍과 얕은 카메라 앵글을 완벽하게 구현하였으며, 올바른 프레이밍에 따른 얕은 심도로 인해 배경의 사다리 등 일부 요소가 생략된 것은 평가 기준에 완벽히 부합합니다."
       },
       {
        "label": "B",
        "score": 3,
        "verdict_ko": "가장 높은 우선순위인 '클로즈업' 샷 크기 지시를 완전히 무시하고 세탁소 내부 전체를 보여주는 와이드 샷을 생성하여 연출 의도를 크게 훼손했습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "카메라는 카운터 표면을 가로질러 중앙 좌측에 홀로 놓인 학생증을 비스듬히 얕은 각도로 바라보고 있다.",
        "built_space": "화면은 낡은 나무 카운터 표면에 집중되어 있으며, 얕은 심도 너머로 펜 꽂이, 접시, 세탁기 등 레퍼런스의 공간 구성 요소들이 흐릿하게 배치되어 있다.",
        "entities": "여성의 사진이 인쇄된 학생증이 카운터 위에 덩그러니 놓여 있으며, 텍스트는 의도대로 읽을 수 없게 처리되었다.",
        "hard_violations": [],
        "physics": "학생증은 카운터의 평평한 표면 위에 물리적으로 자연스럽게 밀착되어 놓여 있다."
       },
       {
        "label": "B",
        "direction": "카메라는 카운터 위에 위치해 세탁소 내부 안쪽을 정면으로 깊게 바라보고 있다.",
        "built_space": "전경의 카운터부터 배경의 세탁기, 벽에 걸린 달력, 건조 중인 옷가지들, 그리고 중앙 뒤편의 사다리까지 공간 전체가 넓은 화각으로 뚜렷하게 보인다.",
        "entities": "카운터 위에는 사진이 인쇄된 학생증이 있고, 배경에는 명시된 사다리가 존재한다.",
        "hard_violations": [],
        "physics": "학생증은 카운터 표면 위에 안정적으로 놓여 있다."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "카운터 바로 위의 얕은 측면 클로즈업, 중앙 왼쪽의 작은 단독 학생증, 충분한 여백을 가장 정확히 구현했지만 카드의 한글과 숫자가 읽혀 ‘읽을 수 있는 글자 금지’를 위반했다."
       },
       {
        "label": "A",
        "score": 3,
        "verdict_ko": "학생증의 방향과 세탁소 위치, 꺼진 조명 아래 사다리 상태는 맞지만 화면이 지나치게 넓어 권위 있는 클로즈업 구도를 놓쳤고 카드와 배경에 읽을 수 있는 글자가 보인다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "시선이나 무기, 이동체는 없다. 학생증의 인물 사진과 정보가 인쇄된 앞면은 위쪽 및 카메라 쪽을 향해 있어 카메라가 그 앞면을 볼 수 있다.",
        "built_space": "낡은 서비스 카운터가 전경 전체를 가로지르고 학생증은 화면 하단 왼쪽에 놓여 있다. 뒤에는 흰색 전면식 기계 2대, 어두운 후면 기계들, 오른쪽 의류 걸이와 벽걸이 선풍기, 시계, 중앙 사다리 1개가 보인다. 사다리는 천장의 어두운 조명 구간 아래에 서 있어 지속 상태와 대체로 맞지만, 카운터 너머 공간을 너무 많이 포함해 지정된 클로즈업보다 훨씬 넓다.",
        "entities": "카운터 위에는 인물 사진이 인쇄된 물리적인 학생증 1개만 있으며 살아 있는 사람이나 신체는 없다. 학생증은 실제 카드 크기로 보이고 사진은 한국인 여성으로 보인다. 카운터의 마모와 세탁소 기계·의류 걸이는 위치 사진의 장소와 부합한다.",
        "hard_violations": [
         "학생증의 한글 및 숫자 정보와 왼쪽 달력의 문자·숫자가 읽을 수 있는 형태로 노출되어 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
        ],
        "physics": "학생증은 카운터 상판에 평평하게 닿아 지지되며 뜨거나 움직이지 않는다. 사다리는 바닥에 양쪽 다리를 두고 자립하고 있고, 그 밖에 공중에 떠 있거나 지지되지 않은 물체는 없다."
       },
       {
        "label": "B",
        "direction": "시선이나 무기, 이동체는 없다. 학생증의 인물 사진이 있는 앞면이 위쪽과 카메라 쪽을 향하며, 얕은 카메라 시선에서 사진과 카드 전면을 확인할 수 있다.",
        "built_space": "카메라는 낡은 서비스 카운터 바로 위에서 한쪽으로 치우쳐 상판을 얕게 가로질러 본다. 학생증은 중앙 왼쪽에 있고 주변에 넓은 빈 상판이 남는다. 왼쪽 뒤에는 위치 사진과 대응하는 펜꽂이, 접시, 탁상 서류와 계산기류가 각각 보이며, 오른쪽 위에는 흰색 전면식 기계 2대의 일부가 흐리게 보인다. 조명과 사다리는 정확한 클로즈업 밖이라 보이지 않으며, 반사 오류나 중복된 고정 설비는 없다.",
        "entities": "인물 사진이 인쇄된 학생증 1개가 카운터에 홀로 놓여 있고 살아 있는 사람이나 손, 추가 카드가 없다. 사진은 한국인 여성으로 보이며 카드의 두께·광택·크기는 실제 학생증처럼 표현됐다. 마모된 목재성 카운터와 뒤쪽 비품은 위치 사진의 해당 지점과 잘 맞는다.",
        "hard_violations": [
         "학생증 전면의 한글과 숫자 일부가 읽을 수 있을 정도로 선명해 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
        ],
        "physics": "학생증 전체가 카운터 상판에 직접 놓여 지지되며 들리거나 부유하지 않는다. 다른 움직이는 몸이나 공중 물체는 없고 물리적으로 불가능한 지지는 보이지 않는다."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 5,
        "verdict_ko": "카운터 바로 위의 얕은 측면 클로즈업, 중앙 왼쪽의 작은 단독 학생증, 충분한 여백을 가장 정확히 구현했지만 카드의 한글과 숫자가 읽혀 ‘읽을 수 있는 글자 금지’를 위반했다."
       },
       {
        "label": "B",
        "score": 3,
        "verdict_ko": "학생증의 방향과 세탁소 위치, 꺼진 조명 아래 사다리 상태는 맞지만 화면이 지나치게 넓어 권위 있는 클로즈업 구도를 놓쳤고 카드와 배경에 읽을 수 있는 글자가 보인다."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "B",
        "direction": "시선이나 무기, 이동체는 없다. 학생증의 인물 사진과 정보가 인쇄된 앞면은 위쪽 및 카메라 쪽을 향해 있어 카메라가 그 앞면을 볼 수 있다.",
        "built_space": "낡은 서비스 카운터가 전경 전체를 가로지르고 학생증은 화면 하단 왼쪽에 놓여 있다. 뒤에는 흰색 전면식 기계 2대, 어두운 후면 기계들, 오른쪽 의류 걸이와 벽걸이 선풍기, 시계, 중앙 사다리 1개가 보인다. 사다리는 천장의 어두운 조명 구간 아래에 서 있어 지속 상태와 대체로 맞지만, 카운터 너머 공간을 너무 많이 포함해 지정된 클로즈업보다 훨씬 넓다.",
        "entities": "카운터 위에는 인물 사진이 인쇄된 물리적인 학생증 1개만 있으며 살아 있는 사람이나 신체는 없다. 학생증은 실제 카드 크기로 보이고 사진은 한국인 여성으로 보인다. 카운터의 마모와 세탁소 기계·의류 걸이는 위치 사진의 장소와 부합한다.",
        "hard_violations": [
         "학생증의 한글 및 숫자 정보와 왼쪽 달력의 문자·숫자가 읽을 수 있는 형태로 노출되어 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
        ],
        "physics": "학생증은 카운터 상판에 평평하게 닿아 지지되며 뜨거나 움직이지 않는다. 사다리는 바닥에 양쪽 다리를 두고 자립하고 있고, 그 밖에 공중에 떠 있거나 지지되지 않은 물체는 없다."
       },
       {
        "label": "A",
        "direction": "시선이나 무기, 이동체는 없다. 학생증의 인물 사진이 있는 앞면이 위쪽과 카메라 쪽을 향하며, 얕은 카메라 시선에서 사진과 카드 전면을 확인할 수 있다.",
        "built_space": "카메라는 낡은 서비스 카운터 바로 위에서 한쪽으로 치우쳐 상판을 얕게 가로질러 본다. 학생증은 중앙 왼쪽에 있고 주변에 넓은 빈 상판이 남는다. 왼쪽 뒤에는 위치 사진과 대응하는 펜꽂이, 접시, 탁상 서류와 계산기류가 각각 보이며, 오른쪽 위에는 흰색 전면식 기계 2대의 일부가 흐리게 보인다. 조명과 사다리는 정확한 클로즈업 밖이라 보이지 않으며, 반사 오류나 중복된 고정 설비는 없다.",
        "entities": "인물 사진이 인쇄된 학생증 1개가 카운터에 홀로 놓여 있고 살아 있는 사람이나 손, 추가 카드가 없다. 사진은 한국인 여성으로 보이며 카드의 두께·광택·크기는 실제 학생증처럼 표현됐다. 마모된 목재성 카운터와 뒤쪽 비품은 위치 사진의 해당 지점과 잘 맞는다.",
        "hard_violations": [
         "학생증 전면의 한글과 숫자 일부가 읽을 수 있을 정도로 선명해 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
        ],
        "physics": "학생증 전체가 카운터 상판에 직접 놓여 지지되며 들리거나 부유하지 않는다. 다른 움직이는 몸이나 공중 물체는 없고 물리적으로 불가능한 지지는 보이지 않는다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": true,
   "slot_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "route": "cross_slot_agree"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 2.0,
    "B": 0.933
   },
   "adjusted": {
    "A": 1.75,
    "B": 0.683
   },
   "violations": {
    "B": [
     "[gpt] 학생증의 한글 및 숫자 정보와 왼쪽 달력의 문자·숫자가 읽을 수 있는 형태로 노출되어 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
    ],
    "A": [
     "[gpt] 학생증 전면의 한글과 숫자 일부가 읽을 수 있을 정도로 선명해 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "agreed": true
  },
  "totals": {
   "A": 1750,
   "B": 683
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1750,
    "verdict_ko": "지시된 클로즈업 프레이밍과 얕은 카메라 앵글을 완벽하게 구현하였으며, 올바른 프레이밍에 따른 얕은 심도로 인해 배경의 사다리 등 일부 요소가 생략된 것은 평가 기준에 완벽히 부합합니다.  ★위반: [gpt] 학생증 전면의 한글과 숫자 일부가 읽을 수 있을 정도로 선명해 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
   },
   {
    "label": "B",
    "score": 683,
    "verdict_ko": "가장 높은 우선순위인 '클로즈업' 샷 크기 지시를 완전히 무시하고 세탁소 내부 전체를 보여주는 와이드 샷을 생성하여 연출 의도를 크게 훼손했습니다.  ★위반: [gpt] 학생증의 한글 및 숫자 정보와 왼쪽 달력의 문자·숫자가 읽을 수 있는 형태로 노출되어 ‘읽을 수 있는 글자 없음’ 지시를 위반한다."
   }
  ],
  "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/b1cefa45-c59c-4402-bcf2-afff78af62a5/episodes/45cce53b-1dbe-4248-b920-c8467186e5dc/images/background_chain/L01B02.png",
    "asset_id": "a8c9e5bc-93e1-4087-81da-527ada65bd94",
    "role": "location_plate"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a9aaa1-4b9a-7f6c-b15f-58265a94a209",
  "ref_mode": "플레이트만 (배경 전용)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S3sh2::cine": {
  "applied": false,
  "attempted_at": "2026-09-04T11:11:11.213604+00:00",
  "fingerprint": "2b3a734745294c12e56ff4f32d9a4a523704a147c3c84126ac4e2e018795fc22",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S3sh2_sel.png",
  "source_sha256": "6f2bfe314e802f22b6b84eea598985b1b6ad8913f00f7d535e7ee1785efc1595",
  "error": "ReveTerminalError: reve 500 downstream_service_error: Downstream service error"
 },
 "S3sh4::signage": {
  "fp": "3cd10e0b9dee2631",
  "inscriptions": [],
  "cues": [
   {
    "text_native": "",
    "source": "scene_text_implied",
    "source_quote": "학생증"
   }
  ],
  "dropped": []
 },
 "S3sh4": {
  "input_fingerprint": "8bd2425d602d9f69",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 학생증을 단단히 쥔 정순의 손이 자신의 복부 쪽 주머니 입구 부근에 닿아 있는 순간.\n\nLOCATION (lock): Inside the laundry shop beside the service counter, where the found ID is being moved toward the wearer’s apron pocket. One fluorescent light is out, leaving the interior unevenly lit. 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 정순's abdomen height on her side three-quarter axis, the static camera isolates her gripping hand at center and the nearby apron-pocket opening just below it. Her fingers close firmly around the student ID as the hand reaches the pocket, with only a narrow portion of her torso retained to establish the action physically.\n- FRAMING SCALE: insert close-up on a detail\n- FRAME LAYOUT: apron pocket opening in the lower-center of the frame, foreground; 정순's gripping hand and abdomen in the middle-center of the frame, foreground, reaches for apron pocket opening.\n- KEY BACKGROUND ELEMENTS: 학생증 (held firmly in 정순's hand) — Its portrait-bearing face is angled partly toward the camera but is mostly obscured by her closing fingers; used as Forms the central detail between her grip and the pocket opening; 앞치마 주머니 (opening positioned beside the gripping hand) — The pocket opening faces upward near the lower center of frame; used as Provides the immediate destination of the student ID and clarifies the interrupted insertion.\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 evening interior remains subdued and unevenly illuminated because one fluorescent light is out, preserving muted color and 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\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established about the place — its fixed features, wear and lighting — persists. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): One fluorescent light remains out, and the ladder remains beneath it. The student ID is no longer lying on the counter. 정순: She wears an apron and firmly holds the student ID at the opening of its abdominal pocket.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 학생증을 단단히 쥔 정순의 손이 자신의 복부 쪽 주머니 입구 부근에 닿아 있는 순간.\n\nLOCATION (lock): Inside the laundry shop beside the service counter, where the found ID is being moved toward the wearer’s apron pocket. One fluorescent light is out, leaving the interior unevenly lit. 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 정순's abdomen height on her side three-quarter axis, the static camera isolates her gripping hand at center and the nearby apron-pocket opening just below it. Her fingers close firmly around the student ID as the hand reaches the pocket, with only a narrow portion of her torso retained to establish the action physically.\n- FRAMING SCALE: insert close-up on a detail\n- FRAME LAYOUT: apron pocket opening in the lower-center of the frame, foreground; 정순's gripping hand and abdomen in the middle-center of the frame, foreground, reaches for apron pocket opening.\n- KEY BACKGROUND ELEMENTS: 학생증 (held firmly in 정순's hand) — Its portrait-bearing face is angled partly toward the camera but is mostly obscured by her closing fingers; used as Forms the central detail between her grip and the pocket opening; 앞치마 주머니 (opening positioned beside the gripping hand) — The pocket opening faces upward near the lower center of frame; used as Provides the immediate destination of the student ID and clarifies the interrupted insertion.\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 evening interior remains subdued and unevenly illuminated because one fluorescent light is out, preserving muted color and 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\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established about the place — its fixed features, wear and lighting — persists. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): One fluorescent light remains out, and the ladder remains beneath it. The student ID is no longer lying on the counter. 정순: She wears an apron and firmly holds the student ID at the opening of its abdominal pocket.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary Korean-speaking setting; all people are Korean-speaking unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 학생증을 단단히 쥔 정순의 손이 자신의 복부 쪽 주머니 입구 부근에 닿아 있는 순간.\n\nLOCATION (lock): Inside the laundry shop beside the service counter, where the found ID is being moved toward the wearer’s apron pocket. One fluorescent light is out, leaving the interior unevenly lit. 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 정순's abdomen height on her side three-quarter axis, the static camera isolates her gripping hand at center and the nearby apron-pocket opening just below it. Her fingers close firmly around the student ID as the hand reaches the pocket, with only a narrow portion of her torso retained to establish the action physically.\n- FRAMING SCALE: insert close-up on a detail\n- FRAME LAYOUT: apron pocket opening in the lower-center of the frame, foreground; 정순's gripping hand and abdomen in the middle-center of the frame, foreground, reaches for apron pocket opening.\n- KEY BACKGROUND ELEMENTS: 학생증 (held firmly in 정순's hand) — Its portrait-bearing face is angled partly toward the camera but is mostly obscured by her closing fingers; used as Forms the central detail between her grip and the pocket opening; 앞치마 주머니 (opening positioned beside the gripping hand) — The pocket opening faces upward near the lower center of frame; used as Provides the immediate destination of the student ID and clarifies the interrupted insertion.\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 evening interior remains subdued and unevenly illuminated because one fluorescent light is out, preserving muted color and 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\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established about the place — its fixed features, wear and lighting — persists. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. 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 within whatever this brief already fixes,\nnot literal fantasy imagery.\n\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 it puts in\nthis shot, imagine that description and still show them as a\nconcrete, fully-formed human being:\nrender the human form as fully as the framing shows — never reduce\na person to a shape, blob, 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 naturally positioned for the action\nthe SHOT TEXT already gives them, gaze on the target the SHOT TEXT\nimplies; avoid a stiff attention stance (feet together, arms hanging\nstraight down) and a blank stare into the lens. If the SHOT TEXT or\nany POSE/IMMOBILE contract above stages stillness, death, sleep,\nunconsciousness, restraint, drill or an explicit direct-to-camera\nlook, follow THAT exactly — this clause never overrides 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\nBODY & SUPPORT (default only — every explicit direction above wins): a body relates to what holds it. A seated person sits the way the seat is built to be used — hips on the seat, back toward the backrest, legs falling naturally toward the floor; people sharing adjacent seats each occupy their own seat, side by side. Hands, hips and feet keep believable contact with whatever they rest on. When the text stages someone frozen, stunned or holding still, the body stops mid-action exactly where the moment caught it — weight already committed to one side, hands where the interrupted movement left them — rather than resetting into a symmetric at-attention stance. If the SHOT TEXT or any contract above stages a specific arrangement, follow THAT exactly.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): One fluorescent light remains out, and the ladder remains beneath it. The student ID is no longer lying on the counter. 정순: She wears an apron and firmly holds the student ID at the opening of its abdominal pocket.\n\nTHE HAND THAT IS DOING THIS: the object at the centre of this shot is being held, operated, read, aimed or handed over by 정순 right now, so 정순's hand — and as much of the wrist and forearm as the framing reaches — is in the frame, gripping or resting on that object exactly the way its real use requires. Match that hand to 정순: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정순 (한국인 여성, 50대의 중년 얼굴, 자연스러운 중년 얼굴선, 짙은색 머리카락, 단정한 기본 헤어스타일) — 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\nCHARACTER REFERENCE ROLE (follow exactly): the attached CHARACTER REFERENCE images establish identity only — face, hair, build and clothing. The pose, gaze direction, camera angle and framing inside those reference images belong to the reference photos, not to this shot; never copy them. Pose and gaze in this shot follow only the CAMERA and action text of this prompt. When this prompt stages a character's face as covered or hidden — by a costume head, mask, helmet, hood, or a body turned away — that staging wins: keep the covering exactly as described and never pull the reference face into view from under it; the reference then guides only what stays visible, such as build and clothing.\n\nNo readable writing anywhere in this image. Surfaces that would carry writing may be present, but nothing a viewer could read: stage every one of them out of legibility — a hand across, an oblique angle, shallow focus, or simply turned away. No caption, subtitle, watermark, logo or overlay.\n\nCINEMATIC FINISH: this must read as a key frame from a film, not a posed photograph — motivated practical light with soft falloff, depth layering of foreground, midground and background, subtle air and atmosphere, restrained film grain. Whatever this brief and its reference images already fix — the framing, the layout, where each thing stands — stays exactly as fixed; make only the choices they leave open, and make those the more cinematic ones.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, and keep from the CAMERA & FRAME contract its subject scale, screen placement and key background elements exactly. For this candidate only, its camera angle and height are deliberately left open: choose a camera position distinctly different from the obvious one, as a film director picking a second setup on the same blocking. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "shot_run_spend_attempt_count": 1,
  "initial_roll_all_fail": true,
  "readings": [
   {
    "label": "A",
    "direction": "학생증을 쥔 손이 앞치마 주머니 입구를 향해 정확히 위치함.",
    "built_space": "배경의 세탁소 카운터, 달력, 펜꽂이, 작은 접시가 올바른 위치와 원근감으로 자연스럽게 배치됨.",
    "entities": "회색 셔츠와 갈색 앞치마는 레퍼런스와 일치하나, 학생증이 가로형에서 세로형 디자인으로 변형되었으며 'STUDENT ID' 텍스트가 선명하게 노출됨.",
    "hard_violations": [
     "[gemini-pro] invented objects (가로형 학생증 소품의 레이아웃을 세로형으로 완전히 변형함)",
     "[gpt] 학생증의 ‘STUDENT ID’와 일부 영문·숫자가 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
    ],
    "physics": "손이 카드를 물리적으로 단단히 쥐고 있으며 중력에 맞게 지탱됨."
   },
   {
    "label": "B",
    "direction": "학생증이 앞치마 주머니 틈을 향해 올바르게 방향을 잡고 있음.",
    "built_space": "세탁소 카운터와 소품들이 배경에 적절한 거리감으로 배치됨.",
    "entities": "의상은 레퍼런스와 일치하나, 학생증 레이아웃이 변형되었고 글자가 선명함.",
    "hard_violations": [
     "[gemini-pro] invented objects (학생증 소품 레이아웃 임의 변형)",
     "[gemini-pro] physically impossible anatomy (가운데 손가락의 길이와 관절 구조가 기형적으로 왜곡됨)",
     "[gpt] 학생증의 ‘STUDENT ID’ 등 영문이 명확히 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
    ],
    "physics": "카드를 잡고 있으나 비정상적인 손가락 구조로 인해 물리적 파지 상태가 부자연스러움."
   }
  ],
  "cross_model_order": {
   "policy": "cross_model_order_v2_slot0_fwd_slot1_rev_parallel2",
   "slots": [
    {
     "model": "gemini-pro",
     "order": "forward",
     "display_to_canonical": {
      "A": "A",
      "B": "B"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 3,
        "verdict_ko": "지정된 프레이밍과 조명, 배경은 잘 구현했으나, 가로형 학생증 소품을 세로형으로 임의 변형하고 텍스트 판독 불가 지시를 위반함."
       },
       {
        "label": "B",
        "score": 1,
        "verdict_ko": "학생증 소품의 임의 변형 및 텍스트 지시 위반에 더해, 카드를 쥔 손가락의 해부학적 구조가 심각하게 왜곡됨."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "학생증을 쥔 손이 앞치마 주머니 입구를 향해 정확히 위치함.",
        "built_space": "배경의 세탁소 카운터, 달력, 펜꽂이, 작은 접시가 올바른 위치와 원근감으로 자연스럽게 배치됨.",
        "entities": "회색 셔츠와 갈색 앞치마는 레퍼런스와 일치하나, 학생증이 가로형에서 세로형 디자인으로 변형되었으며 'STUDENT ID' 텍스트가 선명하게 노출됨.",
        "hard_violations": [
         "invented objects (가로형 학생증 소품의 레이아웃을 세로형으로 완전히 변형함)"
        ],
        "physics": "손이 카드를 물리적으로 단단히 쥐고 있으며 중력에 맞게 지탱됨."
       },
       {
        "label": "B",
        "direction": "학생증이 앞치마 주머니 틈을 향해 올바르게 방향을 잡고 있음.",
        "built_space": "세탁소 카운터와 소품들이 배경에 적절한 거리감으로 배치됨.",
        "entities": "의상은 레퍼런스와 일치하나, 학생증 레이아웃이 변형되었고 글자가 선명함.",
        "hard_violations": [
         "invented objects (학생증 소품 레이아웃 임의 변형)",
         "physically impossible anatomy (가운데 손가락의 길이와 관절 구조가 기형적으로 왜곡됨)"
        ],
        "physics": "카드를 잡고 있으나 비정상적인 손가락 구조로 인해 물리적 파지 상태가 부자연스러움."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 3,
        "verdict_ko": "지정된 프레이밍과 조명, 배경은 잘 구현했으나, 가로형 학생증 소품을 세로형으로 임의 변형하고 텍스트 판독 불가 지시를 위반함."
       },
       {
        "label": "B",
        "score": 1,
        "verdict_ko": "학생증 소품의 임의 변형 및 텍스트 지시 위반에 더해, 카드를 쥔 손가락의 해부학적 구조가 심각하게 왜곡됨."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "A",
        "direction": "학생증을 쥔 손이 앞치마 주머니 입구를 향해 정확히 위치함.",
        "built_space": "배경의 세탁소 카운터, 달력, 펜꽂이, 작은 접시가 올바른 위치와 원근감으로 자연스럽게 배치됨.",
        "entities": "회색 셔츠와 갈색 앞치마는 레퍼런스와 일치하나, 학생증이 가로형에서 세로형 디자인으로 변형되었으며 'STUDENT ID' 텍스트가 선명하게 노출됨.",
        "hard_violations": [
         "invented objects (가로형 학생증 소품의 레이아웃을 세로형으로 완전히 변형함)"
        ],
        "physics": "손이 카드를 물리적으로 단단히 쥐고 있으며 중력에 맞게 지탱됨."
       },
       {
        "label": "B",
        "direction": "학생증이 앞치마 주머니 틈을 향해 올바르게 방향을 잡고 있음.",
        "built_space": "세탁소 카운터와 소품들이 배경에 적절한 거리감으로 배치됨.",
        "entities": "의상은 레퍼런스와 일치하나, 학생증 레이아웃이 변형되었고 글자가 선명함.",
        "hard_violations": [
         "invented objects (학생증 소품 레이아웃 임의 변형)",
         "physically impossible anatomy (가운데 손가락의 길이와 관절 구조가 기형적으로 왜곡됨)"
        ],
        "physics": "카드를 잡고 있으나 비정상적인 손가락 구조로 인해 물리적 파지 상태가 부자연스러움."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 6,
        "verdict_ko": "손·학생증·주머니 입구가 요구된 중앙 삽입 클로즈업에 가장 정확하고 사진 면도 손가락에 대부분 가려졌지만, 카드의 영문이 읽히고 꺼진 형광등에 따른 불균일한 조명이 충분히 드러나지 않는다."
       },
       {
        "label": "A",
        "score": 5,
        "verdict_ko": "학생증을 주머니로 넣는 동작과 세탁소 카운터의 연속성은 좋지만, 사진 면이 지나치게 많이 노출되고 주머니가 오른쪽으로 치우쳤으며 카드 영문이 선명히 읽힌다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정순의 손은 학생증을 아래쪽 앞치마 주머니 입구 안으로 향하게 하고 있으며, 카드의 아래쪽은 이미 주머니에 들어가 있어 목표가 명확히 주머니에 닿는다. 시선이나 무기 등 다른 방향 요소는 없다.",
        "built_space": "복부 높이의 측면 3/4축 근접 화면으로, 앞치마 주머니 1개가 하단 중앙에서 다소 오른쪽에 있고 정순의 손과 복부가 중앙에 있다. 배경에는 서비스 카운터 1개, 세탁기 원형 문 1개와 형광등 일부가 보이며 이전 장면의 낡은 카운터 재질과 세탁소 공간을 비교적 잘 잇는다. 사다리는 이 삽입 구도 밖이라 보이지 않으며 반사는 문제되지 않는다.",
        "entities": "보이는 인물 부분은 회색 반소매와 갈색 앞치마를 착용한 중년 여성의 팔·손·복부로 정순 설정에 부합한다. 학생증은 젊은 한국인 남성 사진이 있는 낡은 플라스틱 카드이지만 기준 카드와 세부 배치가 다르고, 인물 사진과 카드 면이 요구보다 많이 노출되어 있다. 다른 사람이나 별도 소품은 없다.",
        "hard_violations": [
         "학생증의 ‘STUDENT ID’ 등 영문이 명확히 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
        ],
        "physics": "학생증은 정순의 손가락과 엄지로 확실히 잡혀 있고 아래쪽은 주머니 안과 입구에 지지된다. 손은 보이는 팔에 자연스럽게 연결되며, 카드가 주머니로 내려가는 동작과 접촉 관계에 물리적 모순이 없다."
       },
       {
        "label": "B",
        "direction": "정순의 손과 학생증은 중앙에서 바로 아래의 앞치마 주머니 입구를 향하며 카드 하단이 입구 안에 걸쳐 있어 목적지가 정확히 주머니로 읽힌다. 다른 시선·무기·이동체는 없다.",
        "built_space": "앞치마 주머니 1개가 하단 중앙, 손과 복부가 중간 중앙에 놓여 권위 구도와 가장 가깝다. 배경에는 서비스 카운터 1개, 흐릿한 세탁 설비와 천장 형광등이 보인다. 다만 화면에 보이는 형광등들이 밝게 켜져 있어 한 등이 꺼진 불균일한 조명 상태가 명확하지 않다. 사다리는 올바른 삽입 크롭 밖이며 불가능한 반사는 없다.",
        "entities": "회색 반소매, 낡은 갈색 앞치마, 중년 여성으로 보이는 팔과 손이 정순의 의상·연령 단서에 부합한다. 학생증은 낡은 플라스틱 카드이고 사진이 손가락에 대부분 가려져 있어 지시와 잘 맞지만, 기준 카드와 세부 인쇄 배치는 완전히 같지 않다. 추가 인물이나 중복 신체는 없다.",
        "hard_violations": [
         "학생증의 ‘STUDENT ID’와 일부 영문·숫자가 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
        ],
        "physics": "학생증은 엄지와 굽힌 손가락 사이에 잡혀 있으며 하단은 주머니 입구에 들어가 지지된다. 손과 팔의 연결, 카드의 중량감, 주머니 천의 접촉 모두 삽입 도중의 실제 동작으로 성립하며 떠 있는 요소가 없다."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 6,
        "verdict_ko": "손·학생증·주머니 입구가 요구된 중앙 삽입 클로즈업에 가장 정확하고 사진 면도 손가락에 대부분 가려졌지만, 카드의 영문이 읽히고 꺼진 형광등에 따른 불균일한 조명이 충분히 드러나지 않는다."
       },
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "학생증을 주머니로 넣는 동작과 세탁소 카운터의 연속성은 좋지만, 사진 면이 지나치게 많이 노출되고 주머니가 오른쪽으로 치우쳤으며 카드 영문이 선명히 읽힌다."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "B",
        "direction": "정순의 손은 학생증을 아래쪽 앞치마 주머니 입구 안으로 향하게 하고 있으며, 카드의 아래쪽은 이미 주머니에 들어가 있어 목표가 명확히 주머니에 닿는다. 시선이나 무기 등 다른 방향 요소는 없다.",
        "built_space": "복부 높이의 측면 3/4축 근접 화면으로, 앞치마 주머니 1개가 하단 중앙에서 다소 오른쪽에 있고 정순의 손과 복부가 중앙에 있다. 배경에는 서비스 카운터 1개, 세탁기 원형 문 1개와 형광등 일부가 보이며 이전 장면의 낡은 카운터 재질과 세탁소 공간을 비교적 잘 잇는다. 사다리는 이 삽입 구도 밖이라 보이지 않으며 반사는 문제되지 않는다.",
        "entities": "보이는 인물 부분은 회색 반소매와 갈색 앞치마를 착용한 중년 여성의 팔·손·복부로 정순 설정에 부합한다. 학생증은 젊은 한국인 남성 사진이 있는 낡은 플라스틱 카드이지만 기준 카드와 세부 배치가 다르고, 인물 사진과 카드 면이 요구보다 많이 노출되어 있다. 다른 사람이나 별도 소품은 없다.",
        "hard_violations": [
         "학생증의 ‘STUDENT ID’ 등 영문이 명확히 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
        ],
        "physics": "학생증은 정순의 손가락과 엄지로 확실히 잡혀 있고 아래쪽은 주머니 안과 입구에 지지된다. 손은 보이는 팔에 자연스럽게 연결되며, 카드가 주머니로 내려가는 동작과 접촉 관계에 물리적 모순이 없다."
       },
       {
        "label": "A",
        "direction": "정순의 손과 학생증은 중앙에서 바로 아래의 앞치마 주머니 입구를 향하며 카드 하단이 입구 안에 걸쳐 있어 목적지가 정확히 주머니로 읽힌다. 다른 시선·무기·이동체는 없다.",
        "built_space": "앞치마 주머니 1개가 하단 중앙, 손과 복부가 중간 중앙에 놓여 권위 구도와 가장 가깝다. 배경에는 서비스 카운터 1개, 흐릿한 세탁 설비와 천장 형광등이 보인다. 다만 화면에 보이는 형광등들이 밝게 켜져 있어 한 등이 꺼진 불균일한 조명 상태가 명확하지 않다. 사다리는 올바른 삽입 크롭 밖이며 불가능한 반사는 없다.",
        "entities": "회색 반소매, 낡은 갈색 앞치마, 중년 여성으로 보이는 팔과 손이 정순의 의상·연령 단서에 부합한다. 학생증은 낡은 플라스틱 카드이고 사진이 손가락에 대부분 가려져 있어 지시와 잘 맞지만, 기준 카드와 세부 인쇄 배치는 완전히 같지 않다. 추가 인물이나 중복 신체는 없다.",
        "hard_violations": [
         "학생증의 ‘STUDENT ID’와 일부 영문·숫자가 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
        ],
        "physics": "학생증은 엄지와 굽힌 손가락 사이에 잡혀 있으며 하단은 주머니 입구에 들어가 지지된다. 손과 팔의 연결, 카드의 중량감, 주머니 천의 접촉 모두 삽입 도중의 실제 동작으로 성립하며 떠 있는 요소가 없다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": true,
   "slot_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "route": "cross_slot_agree"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 2.0,
    "B": 1.167
   },
   "adjusted": {
    "A": 1.75,
    "B": 0.917
   },
   "violations": {
    "A": [
     "[gemini-pro] invented objects (가로형 학생증 소품의 레이아웃을 세로형으로 완전히 변형함)",
     "[gpt] 학생증의 ‘STUDENT ID’와 일부 영문·숫자가 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
    ],
    "B": [
     "[gemini-pro] invented objects (학생증 소품 레이아웃 임의 변형)",
     "[gemini-pro] physically impossible anatomy (가운데 손가락의 길이와 관절 구조가 기형적으로 왜곡됨)",
     "[gpt] 학생증의 ‘STUDENT ID’ 등 영문이 명확히 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "A"
   },
   "agreed": true
  },
  "totals": {
   "A": 1750,
   "B": 917
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1750,
    "verdict_ko": "지정된 프레이밍과 조명, 배경은 잘 구현했으나, 가로형 학생증 소품을 세로형으로 임의 변형하고 텍스트 판독 불가 지시를 위반함.  ★위반: [gemini-pro] invented objects (가로형 학생증 소품의 레이아웃을 세로형으로 완전히 변형함) / [gpt] 학생증의 ‘STUDENT ID’와 일부 영문·숫자가 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
   },
   {
    "label": "B",
    "score": 917,
    "verdict_ko": "학생증 소품의 임의 변형 및 텍스트 지시 위반에 더해, 카드를 쥔 손가락의 해부학적 구조가 심각하게 왜곡됨.  ★위반: [gemini-pro] invented objects (학생증 소품 레이아웃 임의 변형) / [gemini-pro] physically impossible anatomy (가운데 손가락의 길이와 관절 구조가 기형적으로 왜곡됨) / [gpt] 학생증의 ‘STUDENT ID’ 등 영문이 명확히 읽혀 ‘화면 어디에도 읽을 수 있는 글자가 없어야 한다’는 지시를 위반한다."
   }
  ],
  "refs": [
   {
    "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features and lighting mood are LOCKED to this photo; never copy its camera framing. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. 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/b1cefa45-c59c-4402-bcf2-afff78af62a5/images/45cce53b-1dbe-4248-b920-c8467186e5dc/scene/recipe/S3sh2_sel.png",
    "asset_id": "03f6a5ad-b165-4692-b050-0543a5dd9482",
    "role": "prev_still"
   },
   {
    "label": "CHARACTER REFERENCE — 정순: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:721445>",
    "asset_id": "c709a52a-fbe5-4309-a3de-85e738c118c7",
    "role": "character_ref"
   },
   {
    "label": "PROP REFERENCE — 인물 사진이 인쇄된 학생증: the exact object appearing in this shot; match its look, material and wear exactly.",
    "path": "<bytes:771554>",
    "asset_id": "60e7a5eb-e0b9-49c8-b1cd-f9da348e70f9",
    "role": "prop_ref"
   }
  ],
  "critique_skipped": true,
  "needs_reshoot": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a9aaa5-6cbd-7ba0-b95a-5b87224fe816",
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S3sh2"
  }
 },
 "S3sh4::cine": {
  "applied": false,
  "attempted_at": "2026-09-04T11:13:18.195295+00:00",
  "fingerprint": "a90c92ce98308aa8937afd61cf2c1c4dfacf48229fad98d9e32c25b4ba34b74f",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S3sh4_sel.png",
  "source_sha256": "bc5943de57bf22d79addcc20cc3f0939823e13f1ffa96f42ef0d9d702d896459",
  "error": "ReveTerminalError: reve 500 downstream_service_error: Downstream service error"
 }
}