{
 "S1sh4::confined_fp_apt": {
  "applies": true,
  "reason_ko": "이 숏은 버스 운전석 및 요금 수납 구역이라는 제어 장치가 집중된 밀폐된 공간 내부에서 진행되며, 계기판, 기둥, 요금통, 그리고 인물의 위치 관계가 시각적으로 정확히 묘사되어야 하므로 평면도 배치가 필요합니다.",
  "input_fingerprint": "af6160e68baee7bd"
 },
 "S1sh4::signage": {
  "fp": "d048605823df3f17",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "confinedfp::aed44c973857": {
  "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/confinedfp_base_aed44c973857.png",
  "place_text": "Inside the cramped front fare-collection area of the first intercity bus, directly beside the driver’s instrument panel and the chained fare box.",
  "input_fingerprint": "c0b8a645efd0244d"
 },
 "S1sh4::confined_fp": {
  "reads": {
   "controls": "Driver's station on the left contains a steering wheel and a panel with three analog gauges. A chained farebox is attached to a pole adjacent to this station.",
   "mirrors": "No mirrors are visible.",
   "camera": "Positioned low in the front aisle, facing slightly left toward the driver's area.",
   "occupants": "A woman crouches in the aisle next to the farebox. All visible passenger seats are empty."
  },
  "mismatches": [],
  "scene_description_en": "The camera, positioned low in the bus aisle, looks slightly down and leftward. In the center-right, Jeong-im crouches in a navy uniform and cap, gazing downward. On the left, a metal farebox is chained to a pole. Above it, the driver's station displays three analog gauges and a steering wheel segment. A rubber-matted aisle extends backward, flanked by empty passenger seats with worn blue and white vinyl covers.",
  "fixed": true,
  "input_fingerprint": "e4cfe9ea63fa5e51"
 },
 "era_assess::e58d5118728e22e3": {
  "subjects": [
   {
    "subject_native": "1980년대 한국 시외버스 내부 및 요금통",
    "search_terms_native": [
     "1980년대 시외버스 내부",
     "옛날 버스 요금통",
     "80년대 버스 내부 좌석",
     "과거 시내버스 시외버스 운전석"
    ],
    "language_lock_native": "반드시 한국어로만 검색하고 다른 언어로 번역하거나 추가하지 마십시오.",
    "reason_ko": "1980년대 한국 시외버스의 비닐 시트, 바닥 고무 매트, 쇠사슬로 묶인 구형 요금통 등 특유의 내장재와 설비 구조는 현대 및 해외 버스와 확연히 다릅니다."
   }
  ],
  "subject_text": "낡은 비닐 좌석 커버와 고무 통로 매트, 쇠사슬로 고정된 요금통이 있는 시외버스 1호차 내부\n낡은 비닐 좌석 커버와 고무 통로 매트가 있는 1980년대 시외버스 내부. 전면에는 아날로그 계기와 쇠사슬 고정 요금통이 있다.",
  "identity": "canonical",
  "scope_id": "L01",
  "scope_role": "location_interior",
  "scope_sha": "28bf5edf5f724ac1"
 },
 "era_ref::8f484ee77a47c9d6": {
  "subject": "1980년대 한국 시외버스 내부 및 요금통",
  "terms": [
   "1980년대 시외버스 내부",
   "옛날 버스 요금통",
   "80년대 버스 내부 좌석",
   "과거 시내버스 시외버스 운전석"
  ],
  "queries": [
   [
    "1980년대 한국 시외버스 내부 좌석 운전석",
    "옛날 한국 버스 요금통 1980년대 시내버스 내부"
   ]
  ],
  "candidates": 4,
  "picked_index": 1,
  "picked_url": "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
  "picked_reason_ko": "요금통은 보이지 않으나, 메인 구조인 1980년대 버스 내부의 뼈대와 비례, 운전석 주위의 형태가 방해물 없이 가장 뚜렷하게 잘 보여 세트 제작을 위한 최적의 레퍼런스입니다.",
  "verdicts": [
   {
    "index": 1,
    "usable": true,
    "score": 85,
    "reason_ko": "버스 내부의 전체적인 공간 구조와 운전석, 엔진룸, 바닥 및 창문의 형태가 매우 선명하게 드러납니다."
   },
   {
    "index": 2,
    "usable": true,
    "score": 55,
    "reason_ko": "해당 시대의 실제 버스 내부 모습이나, 승객들로 인해 구조와 디테일을 명확히 파악하기 어렵습니다."
   },
   {
    "index": 3,
    "usable": true,
    "score": 25,
    "reason_ko": "요금통 단일 객체만 근접 촬영되어 메인 구조인 버스 내부의 형태를 전혀 알 수 없습니다."
   },
   {
    "index": 4,
    "usable": false,
    "score": 0,
    "reason_ko": "교통카드 관련 안내문이 붙어 있어 1980년대라는 시대적 배경과 일치하지 않습니다."
   }
  ],
  "candidate_urls": [
   "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
   "https://i3.ruliweb.com/img/19/09/13/16d2a5a205c1dd773.jpg",
   "https://silversbh.daup7.com/v1/tmdckghlttnxhd3.jpg",
   "https://silversbh.daup7.com/v1/qjtmejsxjxjq1.jpg"
  ],
  "sha256": "587de5353dace6b5a009871a0c0885a7670336392d63862d50b3f47e7d8fe625",
  "file": "eraref_8f484ee77a47c9d6.png"
 },
 "S1sh4": {
  "input_fingerprint": "d247d5c6193c1a03",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 쇠사슬로 둥근 아날로그 계기판 아래 기둥에 묶인 요금통 바로 앞에 쪼그려 앉아 시선을 아래로 고정한 정임.\n\nLOCATION (lock): Inside the cramped front fare-collection area of the first intercity bus, directly beside the driver’s instrument panel and the chained fare box. 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- FRAMING SCALE: medium shot\n- KEY BACKGROUND ELEMENTS: 쇠사슬에 묶인 요금통 (기둥에 묶인 채 정임 바로 앞에 놓여 있음) — Its upper opening angles partly toward the camera while its chained side faces the aisle; used as Lower-left foreground anchor leading the eye beneath 정임's lowered eyeline; 둥근 아날로그 계기 세 개 (운전석 옆 계기판에 나란히 붙어 있음) — The three gauge faces are visible above the fare box; their markings face into the bus interior; used as Upper background marker connecting the fare box to the driver's area; 고무 매트가 깔린 통로 (통로 바닥에 깔려 있음); used as Receding background line behind 정임 that reinforces the low camera position; 낡은 비닐 좌석 커버 (각 좌석 등받이에 씌워져 있음) — Seat backs turn in alternating oblique planes along both sides of the aisle; used as Soft peripheral enclosure around the central search area.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained evening ambient light with muted navy, aged beige, subdued neutrals, moderate contrast, and subtle film grain.\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): Bus 1 remains parked with worn vinyl seat covers, a rubber aisle mat, three round analog gauges, and the farebox chained to the pillar beneath them. Several 100-won coins and two folded 1,000-won bills remain at the bottom of the farebox. 정임: She wears a navy conductor's uniform and a billed cap, crouching directly in front of the farebox with her gaze lowered.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
  "roll_prompts": {
   "A": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThe camera, positioned low in the bus aisle, looks slightly down and leftward. In the center-right, Jeong-im crouches in a navy uniform and cap, gazing downward. On the left, a metal farebox is chained to a pole. Above it, the driver's station displays three analog gauges and a steering wheel segment. A rubber-matted aisle extends backward, flanked by empty passenger seats with worn blue and white vinyl covers.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 쇠사슬로 둥근 아날로그 계기판 아래 기둥에 묶인 요금통 바로 앞에 쪼그려 앉아 시선을 아래로 고정한 정임.\n\nLOCATION (lock): Inside the cramped front fare-collection area of the first intercity bus, directly beside the driver’s instrument panel and the chained fare box. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained evening ambient light with muted navy, aged beige, subdued neutrals, moderate contrast, and subtle film grain.\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): Bus 1 remains parked with worn vinyl seat covers, a rubber aisle mat, three round analog gauges, and the farebox chained to the pillar beneath them. Several 100-won coins and two folded 1,000-won bills remain at the bottom of the farebox. 정임: She wears a navy conductor's uniform and a billed cap, crouching directly in front of the farebox with her gaze lowered.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
   "B": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThe camera, positioned low in the bus aisle, looks slightly down and leftward. In the center-right, Jeong-im crouches in a navy uniform and cap, gazing downward. On the left, a metal farebox is chained to a pole. Above it, the driver's station displays three analog gauges and a steering wheel segment. A rubber-matted aisle extends backward, flanked by empty passenger seats with worn blue and white vinyl covers.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 쇠사슬로 둥근 아날로그 계기판 아래 기둥에 묶인 요금통 바로 앞에 쪼그려 앉아 시선을 아래로 고정한 정임.\n\nLOCATION (lock): Inside the cramped front fare-collection area of the first intercity bus, directly beside the driver’s instrument panel and the chained fare box. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained evening ambient light with muted navy, aged beige, subdued neutrals, moderate contrast, and subtle film grain.\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): Bus 1 remains parked with worn vinyl seat covers, a rubber aisle mat, three round analog gauges, and the farebox chained to the pillar beneath them. Several 100-won coins and two folded 1,000-won bills remain at the bottom of the farebox. 정임: She wears a navy conductor's uniform and a billed cap, crouching directly in front of the farebox with her gaze lowered.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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\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": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S1sh4_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "쇠사슬로 묶인 기계식 요금통",
     "path": "<bytes:1341694>",
     "asset_id": "75e7f3d4-fcee-4096-94ba-9c87d856bb4a",
     "role": "prop_ref"
    }
   ],
   "B": [
    {
     "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S1sh4_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "쇠사슬로 묶인 기계식 요금통",
     "path": "<bytes:1341694>",
     "asset_id": "75e7f3d4-fcee-4096-94ba-9c87d856bb4a",
     "role": "prop_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "A",
    "direction": "정임은 요금통 바로 앞에 쪼그려 앉아 바닥과 요금통의 하단 쪽으로 시선을 고정하고 있음.",
    "built_space": "버스 내부. 운전대와 3개의 계기판이 있는 조종석, 기둥에 묶인 녹색 요금통, 고무 매트가 깔린 통로, 낡은 비닐 좌석 커버가 모두 프롬프트가 요구한 위치와 비례에 맞게 올바르게 배치됨.",
    "entities": "정임(20대 여성, 단정한 검은 머리, 승무원 유니폼 및 모자 착용)의 묘사가 지시문과 일치함. 요금통은 지정된 시대 레퍼런스의 형태와 색상(녹색)을 정확히 따랐으나, 금지된 텍스트('시외버스 요금통', '기계식')가 선명하게 렌더링됨. 요금통 내부 창을 통해 지폐와 동전이 보임.",
    "hard_violations": [
     "[gpt] 요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 등 여러 한글 표기가 선명하게 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
    ],
    "physics": "정임은 두 발로 바닥을 디디고 자연스럽게 쪼그려 앉아 체중을 지탱하고 있으며, 요금통은 바닥에 놓인 채 쇠사슬로 기둥에 단단히 고정되어 있음."
   },
   {
    "label": "B",
    "direction": "정임은 쪼그려 앉아 바닥 쪽으로 시선을 고정하고 있음.",
    "built_space": "버스 내부. 운전석, 계기판, 기둥, 통로, 좌석 등이 평면도 레퍼런스(단순 레이아웃 참고용)의 구도를 그대로 복사하여 배치됨.",
    "entities": "정임의 얼굴과 복장은 캐릭터 레퍼런스와 일치하나, 요금통이 지정된 시대 레퍼런스(녹색 기계식 요금통)를 완전히 무시하고 평면도 레퍼런스에 있는 임시 형태(어두운 회색)를 그대로 가져옴. 하단에 지폐와 동전이 보임.",
    "hard_violations": [
     "[gpt] 아래 수납칸 지폐의 ‘1000’ 숫자가 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
    ],
    "physics": "정임은 바닥을 디디고 쪼그려 앉아 있으며, 요금통은 바닥에 놓이고 쇠사슬로 연결되어 지탱됨."
   }
  ],
  "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": 7,
        "verdict_ko": "지정된 시대 레퍼런스의 녹색 요금통을 훌륭하게 합성하고 전체적인 구도와 조명을 완벽히 구현했으나, 프롬프트에서 명시적으로 금지한 '읽을 수 있는 텍스트(시외버스 요금통)'가 그대로 노출되어 감점되었습니다."
       },
       {
        "label": "B",
        "score": 3,
        "verdict_ko": "평면도 레퍼런스의 구도와 조명을 단순 복사하는 데 그쳤으며, 반드시 형태와 재질을 따르라고 지시된 시대 레퍼런스의 녹색 요금통을 전혀 반영하지 않아 지시 사항을 심각하게 위반했습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임은 요금통 바로 앞에 쪼그려 앉아 바닥과 요금통의 하단 쪽으로 시선을 고정하고 있음.",
        "built_space": "버스 내부. 운전대와 3개의 계기판이 있는 조종석, 기둥에 묶인 녹색 요금통, 고무 매트가 깔린 통로, 낡은 비닐 좌석 커버가 모두 프롬프트가 요구한 위치와 비례에 맞게 올바르게 배치됨.",
        "entities": "정임(20대 여성, 단정한 검은 머리, 승무원 유니폼 및 모자 착용)의 묘사가 지시문과 일치함. 요금통은 지정된 시대 레퍼런스의 형태와 색상(녹색)을 정확히 따랐으나, 금지된 텍스트('시외버스 요금통', '기계식')가 선명하게 렌더링됨. 요금통 내부 창을 통해 지폐와 동전이 보임.",
        "hard_violations": [],
        "physics": "정임은 두 발로 바닥을 디디고 자연스럽게 쪼그려 앉아 체중을 지탱하고 있으며, 요금통은 바닥에 놓인 채 쇠사슬로 기둥에 단단히 고정되어 있음."
       },
       {
        "label": "B",
        "direction": "정임은 쪼그려 앉아 바닥 쪽으로 시선을 고정하고 있음.",
        "built_space": "버스 내부. 운전석, 계기판, 기둥, 통로, 좌석 등이 평면도 레퍼런스(단순 레이아웃 참고용)의 구도를 그대로 복사하여 배치됨.",
        "entities": "정임의 얼굴과 복장은 캐릭터 레퍼런스와 일치하나, 요금통이 지정된 시대 레퍼런스(녹색 기계식 요금통)를 완전히 무시하고 평면도 레퍼런스에 있는 임시 형태(어두운 회색)를 그대로 가져옴. 하단에 지폐와 동전이 보임.",
        "hard_violations": [],
        "physics": "정임은 바닥을 디디고 쪼그려 앉아 있으며, 요금통은 바닥에 놓이고 쇠사슬로 연결되어 지탱됨."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 7,
        "verdict_ko": "지정된 시대 레퍼런스의 녹색 요금통을 훌륭하게 합성하고 전체적인 구도와 조명을 완벽히 구현했으나, 프롬프트에서 명시적으로 금지한 '읽을 수 있는 텍스트(시외버스 요금통)'가 그대로 노출되어 감점되었습니다."
       },
       {
        "label": "B",
        "score": 3,
        "verdict_ko": "평면도 레퍼런스의 구도와 조명을 단순 복사하는 데 그쳤으며, 반드시 형태와 재질을 따르라고 지시된 시대 레퍼런스의 녹색 요금통을 전혀 반영하지 않아 지시 사항을 심각하게 위반했습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "정임은 요금통 바로 앞에 쪼그려 앉아 바닥과 요금통의 하단 쪽으로 시선을 고정하고 있음.",
        "built_space": "버스 내부. 운전대와 3개의 계기판이 있는 조종석, 기둥에 묶인 녹색 요금통, 고무 매트가 깔린 통로, 낡은 비닐 좌석 커버가 모두 프롬프트가 요구한 위치와 비례에 맞게 올바르게 배치됨.",
        "entities": "정임(20대 여성, 단정한 검은 머리, 승무원 유니폼 및 모자 착용)의 묘사가 지시문과 일치함. 요금통은 지정된 시대 레퍼런스의 형태와 색상(녹색)을 정확히 따랐으나, 금지된 텍스트('시외버스 요금통', '기계식')가 선명하게 렌더링됨. 요금통 내부 창을 통해 지폐와 동전이 보임.",
        "hard_violations": [],
        "physics": "정임은 두 발로 바닥을 디디고 자연스럽게 쪼그려 앉아 체중을 지탱하고 있으며, 요금통은 바닥에 놓인 채 쇠사슬로 기둥에 단단히 고정되어 있음."
       },
       {
        "label": "B",
        "direction": "정임은 쪼그려 앉아 바닥 쪽으로 시선을 고정하고 있음.",
        "built_space": "버스 내부. 운전석, 계기판, 기둥, 통로, 좌석 등이 평면도 레퍼런스(단순 레이아웃 참고용)의 구도를 그대로 복사하여 배치됨.",
        "entities": "정임의 얼굴과 복장은 캐릭터 레퍼런스와 일치하나, 요금통이 지정된 시대 레퍼런스(녹색 기계식 요금통)를 완전히 무시하고 평면도 레퍼런스에 있는 임시 형태(어두운 회색)를 그대로 가져옴. 하단에 지폐와 동전이 보임.",
        "hard_violations": [],
        "physics": "정임은 바닥을 디디고 쪼그려 앉아 있으며, 요금통은 바닥에 놓이고 쇠사슬로 연결되어 지탱됨."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 7,
        "verdict_ko": "낮은 시선의 웅크린 정임, 좌하단 요금통, 그 위의 계기 세 개와 중간 숏 구도가 가장 정확하지만 지폐의 ‘1000’이 읽히고 요금통 내부 금액도 지정 수량보다 많아 보인다."
       },
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "인물의 동작과 버스 구조는 대체로 맞지만 더 넓은 프레이밍이며 요금통의 한글 표기가 선명하게 노출되어 ‘읽을 수 있는 글자 금지’를 크게 위반한다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임은 카메라를 보지 않고 고개와 두 눈을 아래로 내려, 요금통 바로 옆 바닥과 자신의 손·발 부근을 향해 시선을 고정한다. 무기나 이동체는 없다.",
        "built_space": "낡은 버스 앞 요금 수납 구역이며, 좌하단에 기둥과 쇠사슬로 연결된 요금통 1개, 그 위 운전석 옆 패널에 둥근 계기 3개가 보인다. 정임은 요금통 바로 오른쪽 통로에 쪼그려 앉아 있다. 고무 통로 매트가 뒤로 이어지고 양쪽의 낡은 비닐 좌석 등받이가 비스듬히 통로를 감싼다. 반사상 오류는 보이지 않는다.",
        "entities": "정임은 한국인으로 보이는 20대 여성 1명이며 중간 길이 검은 머리, 남색 승무원 제복과 챙 달린 모자를 착용했다. 녹색 기계식 요금통은 쇠사슬로 기둥에 묶여 있고 동전과 천 원권으로 보이는 지폐가 보인다. 다만 위 투시창과 아래 칸 양쪽에 돈이 있어 두 장보다 많은 지폐와 과도한 수납물이 있는 듯하며, 요금통 형식도 참고물보다 하단 개방 칸이 추가된 형태다.",
        "hard_violations": [
         "아래 수납칸 지폐의 ‘1000’ 숫자가 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
        ],
        "physics": "정임의 체중은 바닥에 닿은 두 구두와 깊게 굽힌 다리로 지지되고, 팔과 손은 정강이·발목 부근에 자연스럽게 놓여 있다. 실제로 가능한 쪼그려 앉은 자세다. 요금통은 바닥에 놓여 있고 쇠사슬은 기둥과 요금통에 연결되어 있어 떠 있는 물체가 없다."
       },
       {
        "label": "B",
        "direction": "정임은 고개를 숙이고 시선을 아래 왼쪽의 요금통 상부와 그 앞 바닥 쪽으로 향한다. 카메라를 보지 않으며 무기나 이동체는 없다.",
        "built_space": "버스 전면부에 요금통 1개가 좌하단, 기둥 1개와 운전석 계기판이 그 뒤에 있으며 둥근 계기 3개가 나란히 보인다. 운전대와 운전석 구조가 왼쪽에 크게 포함되고, 정임은 요금통 오른쪽 통로에 쪼그려 앉았다. 고무 매트 통로와 양쪽 낡은 비닐 좌석이 뒤로 이어진다. 구조적 중복이나 불가능한 반사는 보이지 않지만 A보다 시야가 넓어 정임과 요금통의 화면 점유율이 작다.",
        "entities": "한국인으로 보이는 20대 여성 정임 1명만 있으며 검은 중간 길이 머리, 남색 승무원 제복과 챙 달린 모자가 맞는다. 참고물에 가까운 녹색 기계식 요금통, 쇠사슬, 자물쇠, 금속 컵, 동전과 지폐가 보인다. 그러나 요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 및 작은 안내문이 실제로 읽히며, 내부 지폐도 정확히 두 장인지 불분명하다.",
        "hard_violations": [
         "요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 등 여러 한글 표기가 선명하게 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
        ],
        "physics": "정임은 바닥에 닿은 구두와 깊게 접은 다리로 체중을 지탱하며 손은 정강이 부근에 놓여 있어 가능한 웅크린 자세다. 요금통은 바닥에 안정적으로 서 있고 쇠사슬이 기둥 쪽으로 이어진다. 지지 없이 떠 있는 사람이나 물체는 없다."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 7,
        "verdict_ko": "낮은 시선의 웅크린 정임, 좌하단 요금통, 그 위의 계기 세 개와 중간 숏 구도가 가장 정확하지만 지폐의 ‘1000’이 읽히고 요금통 내부 금액도 지정 수량보다 많아 보인다."
       },
       {
        "label": "A",
        "score": 5,
        "verdict_ko": "인물의 동작과 버스 구조는 대체로 맞지만 더 넓은 프레이밍이며 요금통의 한글 표기가 선명하게 노출되어 ‘읽을 수 있는 글자 금지’를 크게 위반한다."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "B",
        "direction": "정임은 카메라를 보지 않고 고개와 두 눈을 아래로 내려, 요금통 바로 옆 바닥과 자신의 손·발 부근을 향해 시선을 고정한다. 무기나 이동체는 없다.",
        "built_space": "낡은 버스 앞 요금 수납 구역이며, 좌하단에 기둥과 쇠사슬로 연결된 요금통 1개, 그 위 운전석 옆 패널에 둥근 계기 3개가 보인다. 정임은 요금통 바로 오른쪽 통로에 쪼그려 앉아 있다. 고무 통로 매트가 뒤로 이어지고 양쪽의 낡은 비닐 좌석 등받이가 비스듬히 통로를 감싼다. 반사상 오류는 보이지 않는다.",
        "entities": "정임은 한국인으로 보이는 20대 여성 1명이며 중간 길이 검은 머리, 남색 승무원 제복과 챙 달린 모자를 착용했다. 녹색 기계식 요금통은 쇠사슬로 기둥에 묶여 있고 동전과 천 원권으로 보이는 지폐가 보인다. 다만 위 투시창과 아래 칸 양쪽에 돈이 있어 두 장보다 많은 지폐와 과도한 수납물이 있는 듯하며, 요금통 형식도 참고물보다 하단 개방 칸이 추가된 형태다.",
        "hard_violations": [
         "아래 수납칸 지폐의 ‘1000’ 숫자가 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
        ],
        "physics": "정임의 체중은 바닥에 닿은 두 구두와 깊게 굽힌 다리로 지지되고, 팔과 손은 정강이·발목 부근에 자연스럽게 놓여 있다. 실제로 가능한 쪼그려 앉은 자세다. 요금통은 바닥에 놓여 있고 쇠사슬은 기둥과 요금통에 연결되어 있어 떠 있는 물체가 없다."
       },
       {
        "label": "A",
        "direction": "정임은 고개를 숙이고 시선을 아래 왼쪽의 요금통 상부와 그 앞 바닥 쪽으로 향한다. 카메라를 보지 않으며 무기나 이동체는 없다.",
        "built_space": "버스 전면부에 요금통 1개가 좌하단, 기둥 1개와 운전석 계기판이 그 뒤에 있으며 둥근 계기 3개가 나란히 보인다. 운전대와 운전석 구조가 왼쪽에 크게 포함되고, 정임은 요금통 오른쪽 통로에 쪼그려 앉았다. 고무 매트 통로와 양쪽 낡은 비닐 좌석이 뒤로 이어진다. 구조적 중복이나 불가능한 반사는 보이지 않지만 A보다 시야가 넓어 정임과 요금통의 화면 점유율이 작다.",
        "entities": "한국인으로 보이는 20대 여성 정임 1명만 있으며 검은 중간 길이 머리, 남색 승무원 제복과 챙 달린 모자가 맞는다. 참고물에 가까운 녹색 기계식 요금통, 쇠사슬, 자물쇠, 금속 컵, 동전과 지폐가 보인다. 그러나 요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 및 작은 안내문이 실제로 읽히며, 내부 지폐도 정확히 두 장인지 불분명하다.",
        "hard_violations": [
         "요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 등 여러 한글 표기가 선명하게 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
        ],
        "physics": "정임은 바닥에 닿은 구두와 깊게 접은 다리로 체중을 지탱하며 손은 정강이 부근에 놓여 있어 가능한 웅크린 자세다. 요금통은 바닥에 안정적으로 서 있고 쇠사슬이 기둥 쪽으로 이어진다. 지지 없이 떠 있는 사람이나 물체는 없다."
       }
      ]
     },
     "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": 1.714,
    "B": 1.429
   },
   "adjusted": {
    "A": 1.464,
    "B": 1.179
   },
   "violations": {
    "B": [
     "[gpt] 아래 수납칸 지폐의 ‘1000’ 숫자가 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
    ],
    "A": [
     "[gpt] 요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 등 여러 한글 표기가 선명하게 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "B"
   },
   "agreed": false
  },
  "totals": {
   "A": 1464,
   "B": 1179
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1464,
    "verdict_ko": "지정된 시대 레퍼런스의 녹색 요금통을 훌륭하게 합성하고 전체적인 구도와 조명을 완벽히 구현했으나, 프롬프트에서 명시적으로 금지한 '읽을 수 있는 텍스트(시외버스 요금통)'가 그대로 노출되어 감점되었습니다.  ★위반: [gpt] 요금통 전면의 ‘시외버스 요금통’, ‘기계식’ 등 여러 한글 표기가 선명하게 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
   },
   {
    "label": "B",
    "score": 1179,
    "verdict_ko": "평면도 레퍼런스의 구도와 조명을 단순 복사하는 데 그쳤으며, 반드시 형태와 재질을 따르라고 지시된 시대 레퍼런스의 녹색 요금통을 전혀 반영하지 않아 지시 사항을 심각하게 위반했습니다.  ★위반: [gpt] 아래 수납칸 지폐의 ‘1000’ 숫자가 판독되어 ‘읽을 수 있는 글자 없음’ 조건을 위반한다."
   }
  ],
  "refs": [
   {
    "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S1sh4_confinedfp.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "정임",
    "path": "<bytes:812895>",
    "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
    "role": "character_ref"
   },
   {
    "label": "쇠사슬로 묶인 기계식 요금통",
    "path": "<bytes:1341694>",
    "asset_id": "75e7f3d4-fcee-4096-94ba-9c87d856bb4a",
    "role": "prop_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a92a50-4f14-76f5-b98e-6b2b59c0cd3c",
  "confined_fp": {
   "base_key": "confinedfp::aed44c973857",
   "apt_reason": "이 숏은 버스 운전석 및 요금 수납 구역이라는 제어 장치가 집중된 밀폐된 공간 내부에서 진행되며, 계기판, 기둥, 요금통, 그리고 인물의 위치 관계가 시각적으로 정확히 묘사되어야 하므로 평면도 배치가 필요합니다.",
   "fixed": true,
   "mismatches": [],
   "era_research": {
    "subject": "1980년대 한국 시외버스 내부 및 요금통",
    "queries": [
     [
      "1980년대 한국 시외버스 내부 좌석 운전석",
      "옛날 한국 버스 요금통 1980년대 시내버스 내부"
     ]
    ],
    "picked_url": "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
    "sha256": "587de5353dace6b5a009871a0c0885a7670336392d63862d50b3f47e7d8fe625",
    "file": "eraref_8f484ee77a47c9d6.png"
   }
  },
  "ref_mode": "confined_fp: 도면+장면설명+엔티티",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S1sh4::cine": {
  "applied": true,
  "attempted_at": "2026-08-29T09:31:19.510769+00:00",
  "fingerprint": "f83b4be512b356393bccc32ca3271e097a830e02de2af9faa471484810d6bb2f",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S1sh4_sel.png",
  "source_sha256": "9062406ac003974a4ccca2392cbb9d0cf9455261f420cd496b96a5ec9d8f8267",
  "file": "S1sh4_cine.png",
  "staged_sha256": "38cdeb9cb11b6c5a8c51861bef3cfdfc079993f8a4b4a4fc3939b78a0d7909a4",
  "latency_ms": 92287,
  "request_id": "01a04cdc-02f9-7d32-8331-c66649c28c63",
  "status_url": "https://queue.fal.run/reve/2.1/requests/01a04cdc-02f9-7d32-8331-c66649c28c63/status",
  "response_url": "https://queue.fal.run/reve/2.1/requests/01a04cdc-02f9-7d32-8331-c66649c28c63",
  "submitted_at": "2026-08-29T09:31:22.299332+00:00"
 },
 "S1sh7::confined_fp_apt": {
  "applies": true,
  "reason_ko": "이 숏은 버스 내부의 운전석, 앞문, 그리고 요금통이라는 특정한 공간적 배치가 매우 중요한 장면입니다. 안내양 정임이 요금통 앞에 앉아 앞문의 운전기사를 올려다보는 시선과 위치 관계가 정확하게 표현되어야 하며, 이 공간 배치가 어긋날 경우 숏의 논리가 완전히 깨지므로 평면도 레이아웃 가이드가 필요합니다.",
  "input_fingerprint": "46db8dd1c1f59756"
 },
 "S1sh7::signage": {
  "fp": "4a07774a2b267bfc",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "confinedfp::891ad52ba205": {
  "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/confinedfp_base_891ad52ba205.png",
  "place_text": "Inside the first intercity bus at the front fare box, immediately beside the open entryway and within sight of the front-door steps.",
  "input_fingerprint": "a249155691aadac4"
 },
 "S1sh7::confined_fp": {
  "reads": {
   "controls": "Driver seat with a steering wheel and gear shift located in the upper left. Front fare box located centrally, adjacent to the passenger seats and entryway.",
   "mirrors": "No mirrors are shown on the diagram.",
   "camera": "Positioned in the bottom-left area, pointing diagonally up and to the right (northeast) toward the fare box and open entryway.",
   "occupants": "정임 is marked at the front fare box. No other occupants are indicated."
  },
  "mismatches": [
   "The text states 기사 최씨 (Driver Choi) is standing at the front doorway and steps, but he is entirely missing from the diagram.",
   "The text mentions three round analog gauges, which are not depicted on the diagram."
  ],
  "scene_description_en": "This wide shot captures the front interior of an intercity bus, viewed from a camera positioned in the near lower-left and pointing diagonally forward to the right. In the near foreground to the center-left, the front passenger seats stretch into the frame. Just beyond them in the center midground stands the front fare box, where 정임 is positioned. In the far upper-left background, the unoccupied driver's seat, steering wheel, and gear shift are visible. To the far right in the background, the empty open entryway and front-door steps descend outward. No mirrors or other individuals are present in this framing.",
  "fixed": true,
  "input_fingerprint": "68652de299558f32"
 },
 "S1sh7": {
  "input_fingerprint": "1af64de821f422c1",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 요금통 앞에 쪼그려 앉은 채 어깨를 틀어 앞문의 기사 최씨를 올려다보는 정임.\n\nLOCATION (lock): Inside the first intercity bus at the front fare box, immediately beside the open entryway and within sight of the front-door steps. 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- FRAMING SCALE: wide shot\n- FRAME LAYOUT: 정임 in the lower-left of the frame, foreground, looks toward 기사 최씨 at the front doorway; 기사 최씨 in the middle-right of the frame, background, looks toward 정임 beside the fare box; front doorway and steps in the middle-right of the frame, background.\n- KEY BACKGROUND ELEMENTS: 앞문과 계단 (기사 최씨가 계단에 한 발을 올리고 문틀을 짚고 있음) — The doorway is seen diagonally from inside, with its stepped approach receding behind 기사 최씨; used as Background elevation anchor that establishes the vertical separation between the two characters; 쇠사슬에 묶인 요금통 (기둥에 묶인 채 정임 앞에 있음) — Its chained side and part of its upper opening are visible between the characters; used as Lower-frame relational anchor linking 정임's interrupted search to the exchange; 둥근 아날로그 계기 세 개 (운전석 옆 계기판에 나란히 붙어 있음) — The gauge faces remain partly visible above the fare box, facing the bus interior; used as Compact background context for the front section of the bus.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained evening ambient light maintains muted navy and subdued neutral tones with moderate contrast and tactile film grain.\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 정임 — 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 the worn vinyl seat covers, rubber aisle mat, analog gauges, and the chained fare box with its existing contents. Do not add the newer seat covers, bare aisle floor, or empty fare-box pedestal from the other bus, and do not duplicate any person from the reference.\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): Bus 1 retains its worn vinyl seat covers, rubber aisle mat, three round analog gauges, and chained farebox. The coins and two folded bills remain inside the farebox. 정임: She remains in her navy conductor's uniform and billed cap, crouched by the farebox with her shoulders turned and her gaze raised toward the front doorway.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
  "roll_prompts": {
   "A": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThis wide shot captures the front interior of an intercity bus, viewed from a camera positioned in the near lower-left and pointing diagonally forward to the right. In the near foreground to the center-left, the front passenger seats stretch into the frame. Just beyond them in the center midground stands the front fare box, where 정임 is positioned. In the far upper-left background, the unoccupied driver's seat, steering wheel, and gear shift are visible. To the far right in the background, the empty open entryway and front-door steps descend outward. No mirrors or other individuals are present in this framing.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 요금통 앞에 쪼그려 앉은 채 어깨를 틀어 앞문의 기사 최씨를 올려다보는 정임.\n\nLOCATION (lock): Inside the first intercity bus at the front fare box, immediately beside the open entryway and within sight of the front-door steps. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained evening ambient light maintains muted navy and subdued neutral tones with moderate contrast and tactile film grain.\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): Bus 1 retains its worn vinyl seat covers, rubber aisle mat, three round analog gauges, and chained farebox. The coins and two folded bills remain inside the farebox. 정임: She remains in her navy conductor's uniform and billed cap, crouched by the farebox with her shoulders turned and her gaze raised toward the front doorway.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
   "B": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThis wide shot captures the front interior of an intercity bus, viewed from a camera positioned in the near lower-left and pointing diagonally forward to the right. In the near foreground to the center-left, the front passenger seats stretch into the frame. Just beyond them in the center midground stands the front fare box, where 정임 is positioned. In the far upper-left background, the unoccupied driver's seat, steering wheel, and gear shift are visible. To the far right in the background, the empty open entryway and front-door steps descend outward. No mirrors or other individuals are present in this framing.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 요금통 앞에 쪼그려 앉은 채 어깨를 틀어 앞문의 기사 최씨를 올려다보는 정임.\n\nLOCATION (lock): Inside the first intercity bus at the front fare box, immediately beside the open entryway and within sight of the front-door steps. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Restrained evening ambient light maintains muted navy and subdued neutral tones with moderate contrast and tactile film grain.\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): Bus 1 retains its worn vinyl seat covers, rubber aisle mat, three round analog gauges, and chained farebox. The coins and two folded bills remain inside the farebox. 정임: She remains in her navy conductor's uniform and billed cap, crouched by the farebox with her shoulders turned and her gaze raised toward the front doorway.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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\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": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S1sh7_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "쇠사슬로 묶인 기계식 요금통",
     "path": "<bytes:1341694>",
     "asset_id": "75e7f3d4-fcee-4096-94ba-9c87d856bb4a",
     "role": "prop_ref"
    }
   ],
   "B": [
    {
     "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S1sh7_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "쇠사슬로 묶인 기계식 요금통",
     "path": "<bytes:1341694>",
     "asset_id": "75e7f3d4-fcee-4096-94ba-9c87d856bb4a",
     "role": "prop_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "initial_roll_all_fail": true,
  "readings": [
   {
    "label": "A",
    "direction": "정임은 앞문 쪽을 바라보지만, 시선의 대상인 기사 최씨가 없어 허공을 향함.",
    "built_space": "버스 내부로 붉은색 좌석과 요금통이 있으나, 지시된 프레임 구도(좌측 하단 정임, 우측 기사)가 구현되지 않음.",
    "entities": "정임과 요금통의 형태는 일치하지만 기사 최씨가 완전히 누락됨. 요금통 표면에 명확히 읽히는 한글이 존재함.",
    "hard_violations": [
     "[gemini-pro] 지정된 인물 누락 (기사 최씨 없음)",
     "[gemini-pro] 금지된 읽을 수 있는 텍스트 노출 (요금통 표면의 한글)",
     "[gpt] 요금통 전면에 읽을 수 있는 한국어 표제와 안내문이 선명하게 노출되어 금지된 텍스트가 프레임에 유출되었다."
    ],
    "physics": "정임이 바닥에 쪼그려 앉은 자세로 두 발에 체중이 지지되어 있음."
   },
   {
    "label": "B",
    "direction": "정임은 우측의 기사 최씨를 올려다보고, 기사 최씨는 정임을 내려다보며 명확히 시선이 교차함.",
    "built_space": "버스 내부로 요금통과 앞문이 보이나, 좌석이 레퍼런스와 다른 파란색이며 정임이 프레임 우측에 치우쳐 있음.",
    "entities": "정임, 기사 최씨, 요금통이 모두 등장하나 아날로그 계기판의 형태와 위치가 지시와 다소 다름.",
    "hard_violations": [
     "[gemini-pro] 인물 위치 위반 (기사 최씨가 계단에 발을 올리지 않고 버스 안쪽 평탄한 바닥에 서 있음)",
     "[gpt] 운전석 옆에 고정된 원형 아날로그 계기가 지정된 세 개보다 많이 보여 고정 설비가 중복되었다."
    ],
    "physics": "정임은 쪼그려 앉고 기사는 서 있는 상태로, 두 사람 모두 두 발이 바닥에 안정적으로 닿아 있음."
   }
  ],
  "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": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "정임과 기사 최씨의 상호작용은 표현되었으나, 기사의 위치와 자세가 지정된 바와 다르고 좌석 색상이 레퍼런스와 어긋남."
       },
       {
        "label": "A",
        "score": 3,
        "verdict_ko": "필수 인물인 기사 최씨가 화면에 전혀 등장하지 않으며, 요금통에 읽을 수 있는 텍스트가 명확히 노출되어 심각한 위반임."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임은 앞문 쪽을 바라보지만, 시선의 대상인 기사 최씨가 없어 허공을 향함.",
        "built_space": "버스 내부로 붉은색 좌석과 요금통이 있으나, 지시된 프레임 구도(좌측 하단 정임, 우측 기사)가 구현되지 않음.",
        "entities": "정임과 요금통의 형태는 일치하지만 기사 최씨가 완전히 누락됨. 요금통 표면에 명확히 읽히는 한글이 존재함.",
        "hard_violations": [
         "지정된 인물 누락 (기사 최씨 없음)",
         "금지된 읽을 수 있는 텍스트 노출 (요금통 표면의 한글)"
        ],
        "physics": "정임이 바닥에 쪼그려 앉은 자세로 두 발에 체중이 지지되어 있음."
       },
       {
        "label": "B",
        "direction": "정임은 우측의 기사 최씨를 올려다보고, 기사 최씨는 정임을 내려다보며 명확히 시선이 교차함.",
        "built_space": "버스 내부로 요금통과 앞문이 보이나, 좌석이 레퍼런스와 다른 파란색이며 정임이 프레임 우측에 치우쳐 있음.",
        "entities": "정임, 기사 최씨, 요금통이 모두 등장하나 아날로그 계기판의 형태와 위치가 지시와 다소 다름.",
        "hard_violations": [
         "인물 위치 위반 (기사 최씨가 계단에 발을 올리지 않고 버스 안쪽 평탄한 바닥에 서 있음)"
        ],
        "physics": "정임은 쪼그려 앉고 기사는 서 있는 상태로, 두 사람 모두 두 발이 바닥에 안정적으로 닿아 있음."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "정임과 기사 최씨의 상호작용은 표현되었으나, 기사의 위치와 자세가 지정된 바와 다르고 좌석 색상이 레퍼런스와 어긋남."
       },
       {
        "label": "A",
        "score": 3,
        "verdict_ko": "필수 인물인 기사 최씨가 화면에 전혀 등장하지 않으며, 요금통에 읽을 수 있는 텍스트가 명확히 노출되어 심각한 위반임."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "A",
        "direction": "정임은 앞문 쪽을 바라보지만, 시선의 대상인 기사 최씨가 없어 허공을 향함.",
        "built_space": "버스 내부로 붉은색 좌석과 요금통이 있으나, 지시된 프레임 구도(좌측 하단 정임, 우측 기사)가 구현되지 않음.",
        "entities": "정임과 요금통의 형태는 일치하지만 기사 최씨가 완전히 누락됨. 요금통 표면에 명확히 읽히는 한글이 존재함.",
        "hard_violations": [
         "지정된 인물 누락 (기사 최씨 없음)",
         "금지된 읽을 수 있는 텍스트 노출 (요금통 표면의 한글)"
        ],
        "physics": "정임이 바닥에 쪼그려 앉은 자세로 두 발에 체중이 지지되어 있음."
       },
       {
        "label": "B",
        "direction": "정임은 우측의 기사 최씨를 올려다보고, 기사 최씨는 정임을 내려다보며 명확히 시선이 교차함.",
        "built_space": "버스 내부로 요금통과 앞문이 보이나, 좌석이 레퍼런스와 다른 파란색이며 정임이 프레임 우측에 치우쳐 있음.",
        "entities": "정임, 기사 최씨, 요금통이 모두 등장하나 아날로그 계기판의 형태와 위치가 지시와 다소 다름.",
        "hard_violations": [
         "인물 위치 위반 (기사 최씨가 계단에 발을 올리지 않고 버스 안쪽 평탄한 바닥에 서 있음)"
        ],
        "physics": "정임은 쪼그려 앉고 기사는 서 있는 상태로, 두 사람 모두 두 발이 바닥에 안정적으로 닿아 있음."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 5,
        "verdict_ko": "정임과 최 기사의 상호 시선, 쪼그린 자세, 열린 앞문 배치는 가장 가깝지만 정임이 하단 좌측이 아닌 중앙에 가깝고 기사가 문틀을 짚지 않으며 계기 수가 초과된다."
       },
       {
        "label": "B",
        "score": 1,
        "verdict_ko": "정임과 시대형 요금통은 보이지만 핵심 인물인 최 기사와 두 사람의 시선 교환이 통째로 빠지고, 요금통의 글씨도 선명하게 노출된다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임은 몸과 어깨를 오른쪽으로 틀어 열린 앞문의 최 기사를 올려다보고, 최 기사도 아래쪽 정임을 내려다본다. 두 사람의 시선은 서로에게 정확히 닿는다. 무기나 이동 물체는 없다.",
        "built_space": "버스 내부 후방 쪽에서 전면을 보는 넓은 구도다. 열린 앞문은 오른쪽, 하강하는 출입 계단은 기사 아래에 있으며, 정임은 요금통 옆 통로 바닥에 쪼그려 앉아 있다. 최 기사는 문간에 서 있지만 요구된 문틀 접촉은 없고 팔을 아래로 늘어뜨렸다. 쇠사슬이 달린 요금통 하나가 두 사람 사이에 있고 고무 통로 매트와 낡은 좌석이 보인다. 운전석 옆에는 요구된 세 개보다 많은 원형 계기 얼굴이 보인다. 반사상은 특별히 불가능해 보이지 않는다.",
        "entities": "정임은 한국인으로 보이는 20대 여성이고 남색 제복과 챙모자, 중간 길이 검은 머리를 갖춰 참조 정체성과 대체로 맞는다. 최 기사는 한국인 중년 남성으로 보이며 앞문에 있다. 요금통은 녹색 금속제이고 쇠사슬이 있으나 참조의 투명창과 세부 구조가 빠진 단순한 상자형이라 형태 충실도가 낮고, 동전과 접힌 지폐는 확인되지 않는다. 낡은 좌석과 고무 매트는 존재한다.",
        "hard_violations": [
         "운전석 옆에 고정된 원형 아날로그 계기가 지정된 세 개보다 많이 보여 고정 설비가 중복되었다."
        ],
        "physics": "정임은 두 발을 바닥에 대고 무릎을 접은 안정적인 쪼그림 자세이며 떠 있지 않는다. 최 기사는 계단·문턱 쪽에 체중을 싣고 서 있어 지지가 가능하다. 요금통은 바닥에 놓이고 쇠사슬로 고정되어 있다. 다만 기사의 한 발이 어느 계단에 놓였는지는 부분적으로 가려져 명확하지 않다."
       },
       {
        "label": "B",
        "direction": "정임은 오른쪽 열린 출입문 바깥쪽을 올려다보지만 그 시선의 목표인 최 기사가 화면에 없다. 따라서 요구된 정임과 최 기사 사이의 상호 시선은 성립하지 않는다.",
        "built_space": "버스 내부를 길게 보는 넓은 구도이며 정임은 하단 좌측이 아니라 화면 중앙에 쪼그려 앉아 있다. 열린 앞문과 낮아지는 출입구 공간은 오른쪽에 있고, 요금통 하나가 문 옆에 있다. 쇠사슬은 보이지만 기둥을 감아 고정된 관계가 분명하지 않다. 낡은 좌석과 고무 통로 매트는 보이나, 최 기사가 서서 한 발을 올려야 할 계단 배치와 문틀 접촉은 구현되지 않았다. 지정된 세 개의 원형 계기도 보이지 않는다. 불가능한 반사는 보이지 않는다.",
        "entities": "정임은 한국인으로 보이는 20대 여성이고 남색 제복과 챙모자, 검은 머리를 갖췄다. 녹색 기계식 요금통은 참조 형태와 비교적 가깝고 창 안에 동전과 지폐가 보인다. 그러나 최 기사라는 핵심 인물이 완전히 누락됐다. 요금통 전면에는 한국어 표제와 안내문이 읽을 수 있을 정도로 노출되어 무문자 조건을 어긴다.",
        "hard_violations": [
         "요금통 전면에 읽을 수 있는 한국어 표제와 안내문이 선명하게 노출되어 금지된 텍스트가 프레임에 유출되었다."
        ],
        "physics": "정임은 두 발로 바닥을 딛고 깊게 쪼그려 앉아 있어 신체 지지가 자연스럽다. 요금통은 출입구 옆 단상에 놓여 있고 난간과 바닥이 이를 지지한다. 공중에 뜬 신체나 물체는 없다. 다만 최 기사가 없으므로 한 발을 계단에 올리고 문틀을 짚는 지정 동작 자체를 검증할 수 없다."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "정임과 최 기사의 상호 시선, 쪼그린 자세, 열린 앞문 배치는 가장 가깝지만 정임이 하단 좌측이 아닌 중앙에 가깝고 기사가 문틀을 짚지 않으며 계기 수가 초과된다."
       },
       {
        "label": "A",
        "score": 1,
        "verdict_ko": "정임과 시대형 요금통은 보이지만 핵심 인물인 최 기사와 두 사람의 시선 교환이 통째로 빠지고, 요금통의 글씨도 선명하게 노출된다."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "B",
        "direction": "정임은 몸과 어깨를 오른쪽으로 틀어 열린 앞문의 최 기사를 올려다보고, 최 기사도 아래쪽 정임을 내려다본다. 두 사람의 시선은 서로에게 정확히 닿는다. 무기나 이동 물체는 없다.",
        "built_space": "버스 내부 후방 쪽에서 전면을 보는 넓은 구도다. 열린 앞문은 오른쪽, 하강하는 출입 계단은 기사 아래에 있으며, 정임은 요금통 옆 통로 바닥에 쪼그려 앉아 있다. 최 기사는 문간에 서 있지만 요구된 문틀 접촉은 없고 팔을 아래로 늘어뜨렸다. 쇠사슬이 달린 요금통 하나가 두 사람 사이에 있고 고무 통로 매트와 낡은 좌석이 보인다. 운전석 옆에는 요구된 세 개보다 많은 원형 계기 얼굴이 보인다. 반사상은 특별히 불가능해 보이지 않는다.",
        "entities": "정임은 한국인으로 보이는 20대 여성이고 남색 제복과 챙모자, 중간 길이 검은 머리를 갖춰 참조 정체성과 대체로 맞는다. 최 기사는 한국인 중년 남성으로 보이며 앞문에 있다. 요금통은 녹색 금속제이고 쇠사슬이 있으나 참조의 투명창과 세부 구조가 빠진 단순한 상자형이라 형태 충실도가 낮고, 동전과 접힌 지폐는 확인되지 않는다. 낡은 좌석과 고무 매트는 존재한다.",
        "hard_violations": [
         "운전석 옆에 고정된 원형 아날로그 계기가 지정된 세 개보다 많이 보여 고정 설비가 중복되었다."
        ],
        "physics": "정임은 두 발을 바닥에 대고 무릎을 접은 안정적인 쪼그림 자세이며 떠 있지 않는다. 최 기사는 계단·문턱 쪽에 체중을 싣고 서 있어 지지가 가능하다. 요금통은 바닥에 놓이고 쇠사슬로 고정되어 있다. 다만 기사의 한 발이 어느 계단에 놓였는지는 부분적으로 가려져 명확하지 않다."
       },
       {
        "label": "A",
        "direction": "정임은 오른쪽 열린 출입문 바깥쪽을 올려다보지만 그 시선의 목표인 최 기사가 화면에 없다. 따라서 요구된 정임과 최 기사 사이의 상호 시선은 성립하지 않는다.",
        "built_space": "버스 내부를 길게 보는 넓은 구도이며 정임은 하단 좌측이 아니라 화면 중앙에 쪼그려 앉아 있다. 열린 앞문과 낮아지는 출입구 공간은 오른쪽에 있고, 요금통 하나가 문 옆에 있다. 쇠사슬은 보이지만 기둥을 감아 고정된 관계가 분명하지 않다. 낡은 좌석과 고무 통로 매트는 보이나, 최 기사가 서서 한 발을 올려야 할 계단 배치와 문틀 접촉은 구현되지 않았다. 지정된 세 개의 원형 계기도 보이지 않는다. 불가능한 반사는 보이지 않는다.",
        "entities": "정임은 한국인으로 보이는 20대 여성이고 남색 제복과 챙모자, 검은 머리를 갖췄다. 녹색 기계식 요금통은 참조 형태와 비교적 가깝고 창 안에 동전과 지폐가 보인다. 그러나 최 기사라는 핵심 인물이 완전히 누락됐다. 요금통 전면에는 한국어 표제와 안내문이 읽을 수 있을 정도로 노출되어 무문자 조건을 어긴다.",
        "hard_violations": [
         "요금통 전면에 읽을 수 있는 한국어 표제와 안내문이 선명하게 노출되어 금지된 텍스트가 프레임에 유출되었다."
        ],
        "physics": "정임은 두 발로 바닥을 딛고 깊게 쪼그려 앉아 있어 신체 지지가 자연스럽다. 요금통은 출입구 옆 단상에 놓여 있고 난간과 바닥이 이를 지지한다. 공중에 뜬 신체나 물체는 없다. 다만 최 기사가 없으므로 한 발을 계단에 올리고 문틀을 짚는 지정 동작 자체를 검증할 수 없다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": true,
   "slot_winner": {
    "gemini-pro": "B",
    "gpt": "B"
   },
   "route": "cross_slot_agree"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 0.8,
    "B": 2.0
   },
   "adjusted": {
    "A": 0.55,
    "B": 1.75
   },
   "violations": {
    "A": [
     "[gemini-pro] 지정된 인물 누락 (기사 최씨 없음)",
     "[gemini-pro] 금지된 읽을 수 있는 텍스트 노출 (요금통 표면의 한글)",
     "[gpt] 요금통 전면에 읽을 수 있는 한국어 표제와 안내문이 선명하게 노출되어 금지된 텍스트가 프레임에 유출되었다."
    ],
    "B": [
     "[gemini-pro] 인물 위치 위반 (기사 최씨가 계단에 발을 올리지 않고 버스 안쪽 평탄한 바닥에 서 있음)",
     "[gpt] 운전석 옆에 고정된 원형 아날로그 계기가 지정된 세 개보다 많이 보여 고정 설비가 중복되었다."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "B",
    "gpt": "B"
   },
   "agreed": true
  },
  "totals": {
   "B": 1750,
   "A": 550
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 1750,
    "verdict_ko": "정임과 기사 최씨의 상호작용은 표현되었으나, 기사의 위치와 자세가 지정된 바와 다르고 좌석 색상이 레퍼런스와 어긋남.  ★위반: [gemini-pro] 인물 위치 위반 (기사 최씨가 계단에 발을 올리지 않고 버스 안쪽 평탄한 바닥에 서 있음) / [gpt] 운전석 옆에 고정된 원형 아날로그 계기가 지정된 세 개보다 많이 보여 고정 설비가 중복되었다."
   },
   {
    "label": "A",
    "score": 550,
    "verdict_ko": "필수 인물인 기사 최씨가 화면에 전혀 등장하지 않으며, 요금통에 읽을 수 있는 텍스트가 명확히 노출되어 심각한 위반임.  ★위반: [gemini-pro] 지정된 인물 누락 (기사 최씨 없음) / [gemini-pro] 금지된 읽을 수 있는 텍스트 노출 (요금통 표면의 한글) / [gpt] 요금통 전면에 읽을 수 있는 한국어 표제와 안내문이 선명하게 노출되어 금지된 텍스트가 프레임에 유출되었다."
   }
  ],
  "refs": [
   {
    "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S1sh7_confinedfp.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "정임",
    "path": "<bytes:812895>",
    "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
    "role": "character_ref"
   },
   {
    "label": "쇠사슬로 묶인 기계식 요금통",
    "path": "<bytes:1341694>",
    "asset_id": "75e7f3d4-fcee-4096-94ba-9c87d856bb4a",
    "role": "prop_ref"
   }
  ],
  "critique_skipped": true,
  "needs_reshoot": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a92a74-3de9-7f7b-b9eb-4744f2cc0417",
  "confined_fp": {
   "base_key": "confinedfp::891ad52ba205",
   "apt_reason": "이 숏은 버스 내부의 운전석, 앞문, 그리고 요금통이라는 특정한 공간적 배치가 매우 중요한 장면입니다. 안내양 정임이 요금통 앞에 앉아 앞문의 운전기사를 올려다보는 시선과 위치 관계가 정확하게 표현되어야 하며, 이 공간 배치가 어긋날 경우 숏의 논리가 완전히 깨지므로 평면도 레이아웃 가이드가 필요합니다.",
   "fixed": true,
   "mismatches": [
    "The text states 기사 최씨 (Driver Choi) is standing at the front doorway and steps, but he is entirely missing from the diagram.",
    "The text mentions three round analog gauges, which are not depicted on the diagram."
   ],
   "era_research": {
    "subject": "1980년대 한국 시외버스 내부 및 요금통",
    "queries": [
     [
      "1980년대 한국 시외버스 내부 좌석 운전석",
      "옛날 한국 버스 요금통 1980년대 시내버스 내부"
     ]
    ],
    "picked_url": "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
    "sha256": "587de5353dace6b5a009871a0c0885a7670336392d63862d50b3f47e7d8fe625",
    "file": "eraref_8f484ee77a47c9d6.png"
   }
  },
  "ref_mode": "confined_fp: 도면+장면설명+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S1sh4"
  }
 },
 "S1sh7::cine": {
  "applied": true,
  "attempted_at": "2026-08-29T09:39:04.390644+00:00",
  "fingerprint": "2de0e0e1df2bbb86954fdb28ecd5fde7f85325af9851b1b9e4fb5f753073fe36",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S1sh7_sel.png",
  "source_sha256": "0b43cbf52206118554b96c0c79ac4fb440390e8d4115e624a4c28fd1c5386b48",
  "file": "S1sh7_cine.png",
  "staged_sha256": "88a07f87f8cb00520cba4f4672df2cf6ebe0bced1f1ab4207dbebd88f706cb9b",
  "latency_ms": 93290,
  "request_id": "01a04ce3-1b54-7280-8e87-34898b48ee70",
  "status_url": "https://queue.fal.run/reve/2.1/requests/01a04ce3-1b54-7280-8e87-34898b48ee70/status",
  "response_url": "https://queue.fal.run/reve/2.1/requests/01a04ce3-1b54-7280-8e87-34898b48ee70",
  "submitted_at": "2026-08-29T09:39:07.287891+00:00"
 },
 "S2sh1::confined_fp_apt": {
  "applies": true,
  "reason_ko": "샷이 버스 내부(통로 및 좌석)를 배경으로 하고 있으며, 차량 내부의 좌석 배치와 통로, 손잡이 등의 공간 구조가 정확하게 표현되어야 하므로 평면도 레이아웃 보조가 유용합니다.",
  "input_fingerprint": "ff3e3bc46f5092b3"
 },
 "S2sh1::signage": {
  "fp": "125b738e82160bd2",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "confinedfp::c27b91b795db": {
  "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/confinedfp_base_c27b91b795db.png",
  "place_text": "Inside the passenger aisle of the second intercity bus, between the newly covered seats on the bare, unmatted floor.",
  "input_fingerprint": "ee36318eac6494c7"
 },
 "S2sh1::confined_fp": {
  "reads": {
   "controls": "Driver station located at the front right, with an empty farebox pedestal beside it.",
   "mirrors": "No mirrors or reflective surfaces are indicated in the diagram.",
   "camera": "Positioned at the very rear of the central aisle, pointing straight forward toward the front of the bus.",
   "occupants": "Jeong-im is marked as occupying seat 2A (front left passenger row)."
  },
  "mismatches": [
   "The diagram places Jeong-im explicitly inside seat 2A, whereas the text describes her gripping a handhold and bending forward to scan the floor, which suggests she is standing or leaning deeply into the central aisle rather than sitting."
  ],
  "scene_description_en": "The camera is positioned low at the extreme rear of the bus, looking straight forward down the central receding line of the bare, unmatted aisle in a wide shot. Successive rows of passenger seats with new covers run along both the left and right sides of the screen, creating staggered side framing that recedes into the distance. In the distant front-left of the frame at seat 2A, Jeong-im is visible in her navy conductor's uniform and billed cap. She grips a handhold tightly and leans her upper body forward, scanning the exposed floor in the aisle. Far ahead on the right side of the screen sits the driver's station and an empty square farebox pedestal. The inner edges of the rear doorway subtly border the extreme left and right foreground corners, framing the entry point, while no mirrors or reflective surfaces are visible from this deep vantage point.",
  "fixed": true,
  "input_fingerprint": "ef3630b720ceaf30"
 },
 "era_assess::2fb58482debf6f50": {
  "subjects": [
   {
    "subject_native": "1980년대 한국 시외버스 내부",
    "search_terms_native": [
     "1980년대 시외버스 내부",
     "80년대 고속버스 내부",
     "옛날 시외버스 실내 사진",
     "과거 직행버스 내부"
    ],
    "language_lock_native": "반드시 한국어로 된 1980년대 시외버스 및 고속버스 내부 사진만 검색하고 다른 언어로 번역하거나 추가하지 마십시오.",
    "reason_ko": "1980년대 한국 버스 특유의 좌석 형태, 바닥 마감, 운전석 주변 받침대 구조가 현대 버스와 확연히 다름"
   }
  ],
  "subject_text": "새 좌석 커버와 매트 없는 통로 바닥, 요금통이 빠진 빈 받침대가 있는 시외버스 2호차 내부\n새 좌석 커버와 매트가 없는 통로 바닥을 갖춘 1980년대 시외버스 내부. 운전석 옆에는 요금통 없이 빈 받침대만 남아 있다.",
  "identity": "canonical",
  "scope_id": "L02",
  "scope_role": "location_interior",
  "scope_sha": "517de936fc7c22eb"
 },
 "era_ref::d0f7b4bcb5eda6d3": {
  "subject": "1980년대 한국 시외버스 내부",
  "terms": [
   "1980년대 시외버스 내부",
   "80년대 고속버스 내부",
   "옛날 시외버스 실내 사진",
   "과거 직행버스 내부"
  ],
  "queries": [
   [
    "1980년대 한국 시외버스 내부 실내 사진",
    "80년대 한국 고속버스 내부 직행버스 실내"
   ],
   [
    "1980년대 고속버스 안내양 내부 한국",
    "1980년대 시외버스 차내 한국 사진"
   ],
   [
    "\"고속버스 내부\" \"1988\" 한국",
    "\"시외버스 내부\" \"1980년대\"",
    "\"고속버스 실내\" \"1980년대\" 한국",
    "\"직행버스 내부\" 옛날 사진"
   ]
  ],
  "candidates": 4,
  "picked_index": 1,
  "picked_url": "https://www.kwnews.co.kr/photos/2024/06/12/2024061211260506422_x.jpg",
  "picked_reason_ko": "당시 시외버스의 전형적인 특징인 전방향 좌석, 흰색 머리받침 덮개, 천장 선반 등의 내부 구조가 일상적인 모습 속에서 뚜렷하게 확인되어 가장 적합한 참고 자료입니다.",
  "verdicts": [
   {
    "index": 1,
    "usable": true,
    "score": 85,
    "reason_ko": "승객과 안내양이 탑승한 80년대 시외버스의 좌석 배치와 선반 등 내부 구조가 명확하게 보입니다."
   },
   {
    "index": 2,
    "usable": false,
    "score": 0,
    "reason_ko": "시외버스가 아닌 시내버스의 내부 구조이며, 복원되어 전시용으로 관리되는 차량입니다."
   },
   {
    "index": 3,
    "usable": false,
    "score": 0,
    "reason_ko": "측면을 향한 좌석이 있는 시내버스 내부이며, 구조를 파악하기에는 시야가 좁습니다."
   },
   {
    "index": 4,
    "usable": false,
    "score": 0,
    "reason_ko": "현대 시내버스를 배경으로 한 드라마나 영화의 연출된 장면으로 시대와 차종이 맞지 않습니다."
   }
  ],
  "candidate_urls": [
   "https://www.kwnews.co.kr/photos/2024/06/12/2024061211260506422_x.jpg",
   "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
   "https://p1-tt.byteimg.com/origin/pgc-image/838657bcf83a4368a08b391f0c3de828.jpg",
   "https://static.wixstatic.com/media/d1604a_5851c1f1471743f2b93c0eb959788f79~mv2.png/v1/fill/w_810%2Ch_636%2Cal_c%2Cq_90%2Cenc_avif%2Cquality_auto/d1604a_5851c1f1471743f2b93c0eb959788f79~mv2.png"
  ],
  "sha256": "ee4a8afeeb2143bab2e2a46aaef4fa51c49964533662a5164851cd3723e17ae0",
  "file": "eraref_d0f7b4bcb5eda6d3.png"
 },
 "S2sh1": {
  "input_fingerprint": "a3bbe32a24cd6d42",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 고무 매트 없이 맨바닥이 드러난 통로와 새 좌석 커버가 씌워진 시외버스 2호차 안, 손잡이를 단단히 쥐고 상체를 숙여 바닥을 응시하는 정임.\n\nLOCATION (lock): Inside the passenger aisle of the second intercity bus, between the newly covered seats on the bare, unmatted floor. The shot takes place here — the LOCATION text above is the only authority for this place — no location photograph is attached. Build the place strictly from that text and the shot text, inventing nothing beyond them.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: 맨바닥이 드러난 통로 (통로 매트가 깔리지 않아 바닥이 드러나 있음); used as Central receding line that makes the empty search area legible from the low rear position; 새 좌석 커버 (좌석마다 새것으로 교체되어 있음) — Successive seat backs show alternating side and rear planes along the oblique aisle view; used as Staggered side framing around 정임 and visual contrast with the first bus; 뒷문 입구 (정임이 올라온 뒤 카메라가 바로 안쪽에 위치함) — The inner edges of the rear doorway sit near the frame margins behind the camera position; used as Entry-point context and low-depth origin for the forward movement.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural evening ambient light appropriate to the bus interior, rendered in muted navy, aged beige, and subdued neutral tones with restrained brightness and fine grain.\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): Bus 2 remains parked with new seat covers, a bare aisle floor without a rubber mat, and only an empty pedestal beside the driver's seat where a farebox would be. 정임: She wears the same navy conductor's uniform and billed cap, gripping a handhold while bending forward to scan the floor.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
  "roll_prompts": {
   "A": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThe camera is positioned low at the extreme rear of the bus, looking straight forward down the central receding line of the bare, unmatted aisle in a wide shot. Successive rows of passenger seats with new covers run along both the left and right sides of the screen, creating staggered side framing that recedes into the distance. In the distant front-left of the frame at seat 2A, Jeong-im is visible in her navy conductor's uniform and billed cap. She grips a handhold tightly and leans her upper body forward, scanning the exposed floor in the aisle. Far ahead on the right side of the screen sits the driver's station and an empty square farebox pedestal. The inner edges of the rear doorway subtly border the extreme left and right foreground corners, framing the entry point, while no mirrors or reflective surfaces are visible from this deep vantage point.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 고무 매트 없이 맨바닥이 드러난 통로와 새 좌석 커버가 씌워진 시외버스 2호차 안, 손잡이를 단단히 쥐고 상체를 숙여 바닥을 응시하는 정임.\n\nLOCATION (lock): Inside the passenger aisle of the second intercity bus, between the newly covered seats on the bare, unmatted floor. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural evening ambient light appropriate to the bus interior, rendered in muted navy, aged beige, and subdued neutral tones with restrained brightness and fine grain.\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): Bus 2 remains parked with new seat covers, a bare aisle floor without a rubber mat, and only an empty pedestal beside the driver's seat where a farebox would be. 정임: She wears the same navy conductor's uniform and billed cap, gripping a handhold while bending forward to scan the floor.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
   "B": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThe camera is positioned low at the extreme rear of the bus, looking straight forward down the central receding line of the bare, unmatted aisle in a wide shot. Successive rows of passenger seats with new covers run along both the left and right sides of the screen, creating staggered side framing that recedes into the distance. In the distant front-left of the frame at seat 2A, Jeong-im is visible in her navy conductor's uniform and billed cap. She grips a handhold tightly and leans her upper body forward, scanning the exposed floor in the aisle. Far ahead on the right side of the screen sits the driver's station and an empty square farebox pedestal. The inner edges of the rear doorway subtly border the extreme left and right foreground corners, framing the entry point, while no mirrors or reflective surfaces are visible from this deep vantage point.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 고무 매트 없이 맨바닥이 드러난 통로와 새 좌석 커버가 씌워진 시외버스 2호차 안, 손잡이를 단단히 쥐고 상체를 숙여 바닥을 응시하는 정임.\n\nLOCATION (lock): Inside the passenger aisle of the second intercity bus, between the newly covered seats on the bare, unmatted floor. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural evening ambient light appropriate to the bus interior, rendered in muted navy, aged beige, and subdued neutral tones with restrained brightness and fine grain.\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): Bus 2 remains parked with new seat covers, a bare aisle floor without a rubber mat, and only an empty pedestal beside the driver's seat where a farebox would be. 정임: She wears the same navy conductor's uniform and billed cap, gripping a handhold while bending forward to scan the floor.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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\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": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S2sh1_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_d0f7b4bcb5eda6d3.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    }
   ],
   "B": [
    {
     "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S2sh1_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_d0f7b4bcb5eda6d3.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "A",
    "direction": "정임의 시선은 아래쪽 바닥을 향하고 있음.",
    "built_space": "버스 내부로 카메라는 문 뒤에 위치하며 좌석들은 앞을 향해 있음. 하지만 카메라가 위치한 뒷문 쪽 근경(화면 우측)에 버스 전방에 있어야 할 운전대와 요금함이 배치되어 버스의 앞뒤 구조가 물리적으로 심각하게 모순됨.",
    "entities": "정임은 지시된 남색 유니폼과 모자를 착용함. 좌석마다 하얀색 새 커버가 씌워져 있음.",
    "hard_violations": [
     "[gemini-pro] 물리적으로 불가능한 공간 구조 (버스의 뒷문 프레이밍 영역에 앞쪽 구조물인 운전대와 요금함이 겹쳐서 배치됨)",
     "[gpt] 운전대와 전방 출입구 설비가 전경에 있는 전방 시점으로 구성되어, 뒷문 바로 안쪽에서 전방을 보는 것으로 고정된 카메라 위치를 위반했다.",
     "[gpt] 요금함이 없어야 하고 빈 받침대만 있어야 하는데, 전면 패널과 상부 투입부를 갖춘 대형 요금함 형태의 물체를 추가했다."
    ],
    "physics": "정임은 양손으로 손잡이를 쥐고 있으나 하체가 좌석 위나 공중에 엉거주춤하게 떠 있는 것처럼 묘사되어 발이 물리적으로 바닥을 제대로 지지하지 못하는 상태임."
   },
   {
    "label": "B",
    "direction": "정임의 시선은 통로 바닥을 명확히 향하고 있음.",
    "built_space": "버스 내부로 프레임 양옆에 뒷문 입구가 걸려 있어 제시된 뒷문 카메라 앵글을 정확히 구현함. 중앙의 맨바닥 통로와 새 커버가 씌워진 좌석들의 배열이 깊이감을 주며, 전방에 빈 요금함 받침대가 배치됨 (단, 운전석이 다이어그램 기준 우측이 아닌 좌측에 구현됨).",
    "entities": "정임은 남색 유니폼과 모자를 착용하고 있으며 외모가 레퍼런스와 일치함. 통로에는 고무 매트가 없는 맨바닥이 드러나 있고, 모든 좌석에 새 하얀 커버가 씌워져 있음.",
    "hard_violations": [],
    "physics": "정임은 두 발을 바닥에 안정적으로 딛고 섰으며, 오른손으로 수직 기둥 손잡이를 단단히 쥐고 왼손은 무릎을 짚어 상체를 숙인 자세를 사실적이고 자연스럽게 지탱함."
   }
  ],
  "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": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 9,
        "verdict_ko": "뒷문 입구에서 바라본 완벽한 프레이밍, 맨바닥 통로, 정임의 정확한 복장과 바닥을 응시하는 자세 및 빈 요금함 받침대 등 프롬프트의 핵심 요소를 훌륭하게 구현함 (단, 운전석이 다이어그램과 반대로 좌측에 배치됨)."
       },
       {
        "label": "A",
        "score": 2,
        "verdict_ko": "카메라가 뒷문 쪽에 위치함에도 버스 앞쪽에 있어야 할 운전대와 요금함을 근경에 배치하는 치명적인 구조적 오류(Hard Violation)를 범했으며, 정임의 자세 또한 지지 기반이 불명확함."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임의 시선은 아래쪽 바닥을 향하고 있음.",
        "built_space": "버스 내부로 카메라는 문 뒤에 위치하며 좌석들은 앞을 향해 있음. 하지만 카메라가 위치한 뒷문 쪽 근경(화면 우측)에 버스 전방에 있어야 할 운전대와 요금함이 배치되어 버스의 앞뒤 구조가 물리적으로 심각하게 모순됨.",
        "entities": "정임은 지시된 남색 유니폼과 모자를 착용함. 좌석마다 하얀색 새 커버가 씌워져 있음.",
        "hard_violations": [
         "물리적으로 불가능한 공간 구조 (버스의 뒷문 프레이밍 영역에 앞쪽 구조물인 운전대와 요금함이 겹쳐서 배치됨)"
        ],
        "physics": "정임은 양손으로 손잡이를 쥐고 있으나 하체가 좌석 위나 공중에 엉거주춤하게 떠 있는 것처럼 묘사되어 발이 물리적으로 바닥을 제대로 지지하지 못하는 상태임."
       },
       {
        "label": "B",
        "direction": "정임의 시선은 통로 바닥을 명확히 향하고 있음.",
        "built_space": "버스 내부로 프레임 양옆에 뒷문 입구가 걸려 있어 제시된 뒷문 카메라 앵글을 정확히 구현함. 중앙의 맨바닥 통로와 새 커버가 씌워진 좌석들의 배열이 깊이감을 주며, 전방에 빈 요금함 받침대가 배치됨 (단, 운전석이 다이어그램 기준 우측이 아닌 좌측에 구현됨).",
        "entities": "정임은 남색 유니폼과 모자를 착용하고 있으며 외모가 레퍼런스와 일치함. 통로에는 고무 매트가 없는 맨바닥이 드러나 있고, 모든 좌석에 새 하얀 커버가 씌워져 있음.",
        "hard_violations": [],
        "physics": "정임은 두 발을 바닥에 안정적으로 딛고 섰으며, 오른손으로 수직 기둥 손잡이를 단단히 쥐고 왼손은 무릎을 짚어 상체를 숙인 자세를 사실적이고 자연스럽게 지탱함."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 9,
        "verdict_ko": "뒷문 입구에서 바라본 완벽한 프레이밍, 맨바닥 통로, 정임의 정확한 복장과 바닥을 응시하는 자세 및 빈 요금함 받침대 등 프롬프트의 핵심 요소를 훌륭하게 구현함 (단, 운전석이 다이어그램과 반대로 좌측에 배치됨)."
       },
       {
        "label": "A",
        "score": 2,
        "verdict_ko": "카메라가 뒷문 쪽에 위치함에도 버스 앞쪽에 있어야 할 운전대와 요금함을 근경에 배치하는 치명적인 구조적 오류(Hard Violation)를 범했으며, 정임의 자세 또한 지지 기반이 불명확함."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "정임의 시선은 아래쪽 바닥을 향하고 있음.",
        "built_space": "버스 내부로 카메라는 문 뒤에 위치하며 좌석들은 앞을 향해 있음. 하지만 카메라가 위치한 뒷문 쪽 근경(화면 우측)에 버스 전방에 있어야 할 운전대와 요금함이 배치되어 버스의 앞뒤 구조가 물리적으로 심각하게 모순됨.",
        "entities": "정임은 지시된 남색 유니폼과 모자를 착용함. 좌석마다 하얀색 새 커버가 씌워져 있음.",
        "hard_violations": [
         "물리적으로 불가능한 공간 구조 (버스의 뒷문 프레이밍 영역에 앞쪽 구조물인 운전대와 요금함이 겹쳐서 배치됨)"
        ],
        "physics": "정임은 양손으로 손잡이를 쥐고 있으나 하체가 좌석 위나 공중에 엉거주춤하게 떠 있는 것처럼 묘사되어 발이 물리적으로 바닥을 제대로 지지하지 못하는 상태임."
       },
       {
        "label": "B",
        "direction": "정임의 시선은 통로 바닥을 명확히 향하고 있음.",
        "built_space": "버스 내부로 프레임 양옆에 뒷문 입구가 걸려 있어 제시된 뒷문 카메라 앵글을 정확히 구현함. 중앙의 맨바닥 통로와 새 커버가 씌워진 좌석들의 배열이 깊이감을 주며, 전방에 빈 요금함 받침대가 배치됨 (단, 운전석이 다이어그램 기준 우측이 아닌 좌측에 구현됨).",
        "entities": "정임은 남색 유니폼과 모자를 착용하고 있으며 외모가 레퍼런스와 일치함. 통로에는 고무 매트가 없는 맨바닥이 드러나 있고, 모든 좌석에 새 하얀 커버가 씌워져 있음.",
        "hard_violations": [],
        "physics": "정임은 두 발을 바닥에 안정적으로 딛고 섰으며, 오른손으로 수직 기둥 손잡이를 단단히 쥐고 왼손은 무릎을 짚어 상체를 숙인 자세를 사실적이고 자연스럽게 지탱함."
       }
      ]
     },
     "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": "낮은 후방 시점의 와이드 숏에서 정임이 손잡이를 쥐고 몸을 숙여 맨바닥 통로를 정확히 응시하며, 새 좌석 커버와 저녁 버스 내부도 충실히 구현했다."
       },
       {
        "label": "B",
        "score": 2,
        "verdict_ko": "정임의 복장과 동작은 대체로 맞지만 운전대와 요금함이 전경에 놓인 전방 출입구 시점이라 지정된 후방 카메라 위치를 거스르고, 빈 받침대 대신 요금함을 보여 준다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임의 얼굴과 눈은 카메라가 아니라 자신의 발 앞에서 전방으로 이어지는 노출된 통로 바닥을 향한다. 왼손은 수직 손잡이를 정확히 감싸 쥐며, 무기나 별도의 지시 물체 및 이동 중인 신체는 없다.",
        "built_space": "카메라는 뒷문 바로 안쪽의 낮은 위치에서 버스 전방을 바라보고, 후방 문의 안쪽 가장자리가 좌우 프레임 가장자리에 걸린다. 중앙에는 매트가 없는 평평한 바닥 통로가 길게 후퇴하며, 좌우에 약 7열씩의 좌석 등받이와 새 흰색 커버, 천장 수하물 선반, 수직 손잡이들이 보인다. 정임은 좌우 좌석 사이 통로의 전방부에 서 있고 운전석 주변의 받침 구조도 먼 전경에 작게 놓여 있어 평면도와 카메라 위치가 성립한다.",
        "entities": "등장인물은 정임 한 명뿐이다. 한국인으로 보이는 20대 여성이고, 단정한 중간 길이 검은 머리와 남색 챙모자, 남색 승무원 제복, 갈색 작업화를 착용해 인물 참조와 일치한다. 버스는 1980년대 한국 시외버스형 좌석, 금속 선반과 창호를 갖추며 좌석마다 새 흰색 커버가 씌워져 있다. 읽을 수 있는 문구나 추가 인물은 없다.",
        "hard_violations": [],
        "physics": "정임의 두 발이 통로 바닥에 완전히 닿아 체중을 받치고, 왼손은 천장과 바닥에 고정된 수직 봉을 단단히 잡는다. 오른손은 굽힌 무릎에 얹혀 숙인 상체를 보조하므로 바닥을 살피는 자세가 물리적으로 자연스럽고 떠 있는 신체나 물체가 없다."
       },
       {
        "label": "B",
        "direction": "정임은 고개를 숙여 자신의 오른쪽 아래, 앞쪽 통로 바닥 부근을 바라본다. 양손은 서로 다른 높이에서 수직 손잡이를 감싸고 있어 손잡이를 잡는 방향은 맞지만, 시선은 긴 중앙 검색 구역보다 가까운 입구 바닥 쪽에 치우쳐 있다.",
        "built_space": "프레임 오른쪽 전경에 운전대와 대형 요금함 형태의 금속 상자가 있고, 정임은 그 맞은편 전방 출입구 계단 옆에 있다. 따라서 카메라는 버스 뒤쪽이 아니라 운전석과 전방 출입구 쪽에서 후방 좌석을 바라본다. 좌우에는 여러 열의 새 흰색 좌석 커버와 천장 선반이 있으나 중앙 통로 바닥은 강한 평행 홈이 있는 덮개처럼 보이며, 전방에는 단차도 있다. 지정된 빈 받침대 대신 완성된 수납·투입 구조의 요금함이 보인다.",
        "entities": "정임 한 명만 등장하며 한국인으로 보이는 20대 여성, 중간 길이 검은 머리, 남색 챙모자와 남색 제복, 갈색 신발은 참조와 대체로 맞는다. 시대형 버스 좌석과 흰색 새 커버는 보이지만, 프롬프트가 비어 있어야 한다고 한 요금함 위치에 실제 요금함 같은 상자가 추가되어 있다. 읽을 수 있는 글자는 없다.",
        "hard_violations": [
         "운전대와 전방 출입구 설비가 전경에 있는 전방 시점으로 구성되어, 뒷문 바로 안쪽에서 전방을 보는 것으로 고정된 카메라 위치를 위반했다.",
         "요금함이 없어야 하고 빈 받침대만 있어야 하는데, 전면 패널과 상부 투입부를 갖춘 대형 요금함 형태의 물체를 추가했다."
        ],
        "physics": "정임의 신발은 입구 계단과 바닥에 닿아 있고 양손은 고정된 금속 봉을 실제로 쥐고 있어 몸의 지지가 확인된다. 웅크린 자세는 다소 비좁지만 손과 발로 지탱 가능하며, 공중에 뜬 신체나 지지되지 않은 물체는 없다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 9,
        "verdict_ko": "낮은 후방 시점의 와이드 숏에서 정임이 손잡이를 쥐고 몸을 숙여 맨바닥 통로를 정확히 응시하며, 새 좌석 커버와 저녁 버스 내부도 충실히 구현했다."
       },
       {
        "label": "A",
        "score": 2,
        "verdict_ko": "정임의 복장과 동작은 대체로 맞지만 운전대와 요금함이 전경에 놓인 전방 출입구 시점이라 지정된 후방 카메라 위치를 거스르고, 빈 받침대 대신 요금함을 보여 준다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "B",
        "direction": "정임의 얼굴과 눈은 카메라가 아니라 자신의 발 앞에서 전방으로 이어지는 노출된 통로 바닥을 향한다. 왼손은 수직 손잡이를 정확히 감싸 쥐며, 무기나 별도의 지시 물체 및 이동 중인 신체는 없다.",
        "built_space": "카메라는 뒷문 바로 안쪽의 낮은 위치에서 버스 전방을 바라보고, 후방 문의 안쪽 가장자리가 좌우 프레임 가장자리에 걸린다. 중앙에는 매트가 없는 평평한 바닥 통로가 길게 후퇴하며, 좌우에 약 7열씩의 좌석 등받이와 새 흰색 커버, 천장 수하물 선반, 수직 손잡이들이 보인다. 정임은 좌우 좌석 사이 통로의 전방부에 서 있고 운전석 주변의 받침 구조도 먼 전경에 작게 놓여 있어 평면도와 카메라 위치가 성립한다.",
        "entities": "등장인물은 정임 한 명뿐이다. 한국인으로 보이는 20대 여성이고, 단정한 중간 길이 검은 머리와 남색 챙모자, 남색 승무원 제복, 갈색 작업화를 착용해 인물 참조와 일치한다. 버스는 1980년대 한국 시외버스형 좌석, 금속 선반과 창호를 갖추며 좌석마다 새 흰색 커버가 씌워져 있다. 읽을 수 있는 문구나 추가 인물은 없다.",
        "hard_violations": [],
        "physics": "정임의 두 발이 통로 바닥에 완전히 닿아 체중을 받치고, 왼손은 천장과 바닥에 고정된 수직 봉을 단단히 잡는다. 오른손은 굽힌 무릎에 얹혀 숙인 상체를 보조하므로 바닥을 살피는 자세가 물리적으로 자연스럽고 떠 있는 신체나 물체가 없다."
       },
       {
        "label": "A",
        "direction": "정임은 고개를 숙여 자신의 오른쪽 아래, 앞쪽 통로 바닥 부근을 바라본다. 양손은 서로 다른 높이에서 수직 손잡이를 감싸고 있어 손잡이를 잡는 방향은 맞지만, 시선은 긴 중앙 검색 구역보다 가까운 입구 바닥 쪽에 치우쳐 있다.",
        "built_space": "프레임 오른쪽 전경에 운전대와 대형 요금함 형태의 금속 상자가 있고, 정임은 그 맞은편 전방 출입구 계단 옆에 있다. 따라서 카메라는 버스 뒤쪽이 아니라 운전석과 전방 출입구 쪽에서 후방 좌석을 바라본다. 좌우에는 여러 열의 새 흰색 좌석 커버와 천장 선반이 있으나 중앙 통로 바닥은 강한 평행 홈이 있는 덮개처럼 보이며, 전방에는 단차도 있다. 지정된 빈 받침대 대신 완성된 수납·투입 구조의 요금함이 보인다.",
        "entities": "정임 한 명만 등장하며 한국인으로 보이는 20대 여성, 중간 길이 검은 머리, 남색 챙모자와 남색 제복, 갈색 신발은 참조와 대체로 맞는다. 시대형 버스 좌석과 흰색 새 커버는 보이지만, 프롬프트가 비어 있어야 한다고 한 요금함 위치에 실제 요금함 같은 상자가 추가되어 있다. 읽을 수 있는 글자는 없다.",
        "hard_violations": [
         "운전대와 전방 출입구 설비가 전경에 있는 전방 시점으로 구성되어, 뒷문 바로 안쪽에서 전방을 보는 것으로 고정된 카메라 위치를 위반했다.",
         "요금함이 없어야 하고 빈 받침대만 있어야 하는데, 전면 패널과 상부 투입부를 갖춘 대형 요금함 형태의 물체를 추가했다."
        ],
        "physics": "정임의 신발은 입구 계단과 바닥에 닿아 있고 양손은 고정된 금속 봉을 실제로 쥐고 있어 몸의 지지가 확인된다. 웅크린 자세는 다소 비좁지만 손과 발로 지탱 가능하며, 공중에 뜬 신체나 지지되지 않은 물체는 없다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": true,
   "slot_winner": {
    "gemini-pro": "B",
    "gpt": "B"
   },
   "route": "cross_slot_agree"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 0.444,
    "B": 2.0
   },
   "adjusted": {
    "A": 0.194,
    "B": 2.0
   },
   "violations": {
    "A": [
     "[gemini-pro] 물리적으로 불가능한 공간 구조 (버스의 뒷문 프레이밍 영역에 앞쪽 구조물인 운전대와 요금함이 겹쳐서 배치됨)",
     "[gpt] 운전대와 전방 출입구 설비가 전경에 있는 전방 시점으로 구성되어, 뒷문 바로 안쪽에서 전방을 보는 것으로 고정된 카메라 위치를 위반했다.",
     "[gpt] 요금함이 없어야 하고 빈 받침대만 있어야 하는데, 전면 패널과 상부 투입부를 갖춘 대형 요금함 형태의 물체를 추가했다."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "B",
    "gpt": "B"
   },
   "agreed": true
  },
  "totals": {
   "B": 2000,
   "A": 194
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 2000,
    "verdict_ko": "뒷문 입구에서 바라본 완벽한 프레이밍, 맨바닥 통로, 정임의 정확한 복장과 바닥을 응시하는 자세 및 빈 요금함 받침대 등 프롬프트의 핵심 요소를 훌륭하게 구현함 (단, 운전석이 다이어그램과 반대로 좌측에 배치됨)."
   },
   {
    "label": "A",
    "score": 194,
    "verdict_ko": "카메라가 뒷문 쪽에 위치함에도 버스 앞쪽에 있어야 할 운전대와 요금함을 근경에 배치하는 치명적인 구조적 오류(Hard Violation)를 범했으며, 정임의 자세 또한 지지 기반이 불명확함.  ★위반: [gemini-pro] 물리적으로 불가능한 공간 구조 (버스의 뒷문 프레이밍 영역에 앞쪽 구조물인 운전대와 요금함이 겹쳐서 배치됨) / [gpt] 운전대와 전방 출입구 설비가 전경에 있는 전방 시점으로 구성되어, 뒷문 바로 안쪽에서 전방을 보는 것으로 고정된 카메라 위치를 위반했다. / [gpt] 요금함이 없어야 하고 빈 받침대만 있어야 하는데, 전면 패널과 상부 투입부를 갖춘 대형 요금함 형태의 물체를 추가했다."
   }
  ],
  "refs": [
   {
    "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S2sh1_confinedfp.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_d0f7b4bcb5eda6d3.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "정임",
    "path": "<bytes:812895>",
    "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
    "role": "character_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a92a91-5c40-73ab-bcfc-536dd1b49b3d",
  "confined_fp": {
   "base_key": "confinedfp::c27b91b795db",
   "apt_reason": "샷이 버스 내부(통로 및 좌석)를 배경으로 하고 있으며, 차량 내부의 좌석 배치와 통로, 손잡이 등의 공간 구조가 정확하게 표현되어야 하므로 평면도 레이아웃 보조가 유용합니다.",
   "fixed": true,
   "mismatches": [
    "The diagram places Jeong-im explicitly inside seat 2A, whereas the text describes her gripping a handhold and bending forward to scan the floor, which suggests she is standing or leaning deeply into the central aisle rather than sitting."
   ],
   "era_research": {
    "subject": "1980년대 한국 시외버스 내부",
    "queries": [
     [
      "1980년대 한국 시외버스 내부 실내 사진",
      "80년대 한국 고속버스 내부 직행버스 실내"
     ],
     [
      "1980년대 고속버스 안내양 내부 한국",
      "1980년대 시외버스 차내 한국 사진"
     ],
     [
      "\"고속버스 내부\" \"1988\" 한국",
      "\"시외버스 내부\" \"1980년대\"",
      "\"고속버스 실내\" \"1980년대\" 한국",
      "\"직행버스 내부\" 옛날 사진"
     ]
    ],
    "picked_url": "https://www.kwnews.co.kr/photos/2024/06/12/2024061211260506422_x.jpg",
    "sha256": "ee4a8afeeb2143bab2e2a46aaef4fa51c49964533662a5164851cd3723e17ae0",
    "file": "eraref_d0f7b4bcb5eda6d3.png"
   }
  },
  "ref_mode": "confined_fp: 도면+장면설명+엔티티",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S2sh1::cine": {
  "applied": true,
  "attempted_at": "2026-08-29T09:47:45.373564+00:00",
  "fingerprint": "045164d0dd48fdacf3d27a4fe9321bfd1a67eb6a6bce4f80574546e743cd103e",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S2sh1_sel.png",
  "source_sha256": "a3927656a4afcdec113272a01f24d0e0cce6d0654644665eabb58b4ed6c5731c",
  "file": "S2sh1_cine.png",
  "staged_sha256": "389893b5ef5f76c6b9f51773d63acc105489b907f8661440d4f5ab7bf20d9504",
  "latency_ms": 85105,
  "request_id": "01a04ceb-0dc5-7950-b8a4-b565f8a56151",
  "status_url": "https://queue.fal.run/reve/2.1/requests/01a04ceb-0dc5-7950-b8a4-b565f8a56151/status",
  "response_url": "https://queue.fal.run/reve/2.1/requests/01a04ceb-0dc5-7950-b8a4-b565f8a56151",
  "submitted_at": "2026-08-29T09:47:48.141302+00:00"
 },
 "S2sh4::signage": {
  "fp": "dd92f7bcd3052701",
  "inscriptions": [],
  "cues": [],
  "dropped": []
 },
 "S2sh4": {
  "input_fingerprint": "f56a0304a6784def",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 차창 밖에서 유리창 너머로 정임을 향해 한 손을 높이 들어 올린 자세로 멈춰 선 기사 최씨.\n\nLOCATION (lock): Outside the second intercity bus in the garage, directly alongside one of its passenger windows and looking in through the glass. 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- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 정임 in the lower-left of the frame, foreground, looks toward 기사 최씨 beyond the side window; 기사 최씨 in the middle-right of the frame, midground, looks toward 정임 inside the bus; side window plane in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옆 유리창 (기사 최씨와 정임 사이에 놓여 있음) — The camera looks diagonally through the window surface toward 기사 최씨 outside, with 정임 remaining on the interior side; used as Transparent separator and compositional plane between the characters; 시외버스 2호차 좌석 (새 좌석 커버가 씌워져 있음) — A nearby seat back presents its oblique interior side at the left edge; used as Interior scale reference that confirms the camera remains inside the bus.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural evening ambient light distinguishes the interior and exterior with restrained brightness, subdued neutrals, moderate contrast, and subtle analog grain.\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\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 the new seat covers, uncovered aisle floor, bus windows, handrails, and the empty fare-box pedestal. Exclude the rubber aisle mat, chained fare box, and worn seat covers of the other bus, and do not retain a duplicate figure performing the earlier floor-search action.\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): Bus 2 still has new seat covers, no aisle mat, and an empty farebox pedestal beside the driver's seat. Its window glass separates the bus interior from the garage outside. 기사 최씨: He wears his work jumper and work gloves, standing outside the bus window with one hand raised in a signaling gesture.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 기사 최씨 (한국인 외모, 50대 남성, 나이대에 맞는 얼굴, 짧고 단정한 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 차창 밖에서 유리창 너머로 정임을 향해 한 손을 높이 들어 올린 자세로 멈춰 선 기사 최씨.\n\nLOCATION (lock): Outside the second intercity bus in the garage, directly alongside one of its passenger windows and looking in through the glass. 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- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 정임 in the lower-left of the frame, foreground, looks toward 기사 최씨 beyond the side window; 기사 최씨 in the middle-right of the frame, midground, looks toward 정임 inside the bus; side window plane in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옆 유리창 (기사 최씨와 정임 사이에 놓여 있음) — The camera looks diagonally through the window surface toward 기사 최씨 outside, with 정임 remaining on the interior side; used as Transparent separator and compositional plane between the characters; 시외버스 2호차 좌석 (새 좌석 커버가 씌워져 있음) — A nearby seat back presents its oblique interior side at the left edge; used as Interior scale reference that confirms the camera remains inside the bus.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural evening ambient light distinguishes the interior and exterior with restrained brightness, subdued neutrals, moderate contrast, and subtle analog grain.\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\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 the new seat covers, uncovered aisle floor, bus windows, handrails, and the empty fare-box pedestal. Exclude the rubber aisle mat, chained fare box, and worn seat covers of the other bus, and do not retain a duplicate figure performing the earlier floor-search action.\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): Bus 2 still has new seat covers, no aisle mat, and an empty farebox pedestal beside the driver's seat. Its window glass separates the bus interior from the garage outside. 기사 최씨: He wears his work jumper and work gloves, standing outside the bus window with one hand raised in a signaling gesture.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 기사 최씨 (한국인 외모, 50대 남성, 나이대에 맞는 얼굴, 짧고 단정한 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): evening.\n\nSHOT TEXT (authoritative, Korean): 차창 밖에서 유리창 너머로 정임을 향해 한 손을 높이 들어 올린 자세로 멈춰 선 기사 최씨.\n\nLOCATION (lock): Outside the second intercity bus in the garage, directly alongside one of its passenger windows and looking in through the glass. 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- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 정임 in the lower-left of the frame, foreground, looks toward 기사 최씨 beyond the side window; 기사 최씨 in the middle-right of the frame, midground, looks toward 정임 inside the bus; side window plane in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: 옆 유리창 (기사 최씨와 정임 사이에 놓여 있음) — The camera looks diagonally through the window surface toward 기사 최씨 outside, with 정임 remaining on the interior side; used as Transparent separator and compositional plane between the characters; 시외버스 2호차 좌석 (새 좌석 커버가 씌워져 있음) — A nearby seat back presents its oblique interior side at the left edge; used as Interior scale reference that confirms the camera remains inside the bus.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Natural evening ambient light distinguishes the interior and exterior with restrained brightness, subdued neutrals, moderate contrast, and subtle analog grain.\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\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 the new seat covers, uncovered aisle floor, bus windows, handrails, and the empty fare-box pedestal. Exclude the rubber aisle mat, chained fare box, and worn seat covers of the other bus, and do not retain a duplicate figure performing the earlier floor-search action.\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): Bus 2 still has new seat covers, no aisle mat, and an empty farebox pedestal beside the driver's seat. Its window glass separates the bus interior from the garage outside. 기사 최씨: He wears his work jumper and work gloves, standing outside the bus window with one hand raised in a signaling gesture.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 기사 최씨 (한국인 외모, 50대 남성, 나이대에 맞는 얼굴, 짧고 단정한 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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\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": "좌측 하단의 정임은 창문 너머 최씨를 향해 시선을 두고, 우측 중단의 최씨는 버스 안의 정임을 응시합니다. 최씨의 오른손은 정임을 향해 유리창에 높이 들려 있습니다.",
    "built_space": "버스 내부에서 창밖을 대각선으로 바라보는 뷰를 정확히 구현했습니다. 좌측에 하얀 커버가 씌워진 버스 좌석들이 공간감을 형성하며, 유리창에는 정임의 얼굴이 광학적으로 알맞게 반사되어 있습니다. 창밖 배경은 어두운 차고지 내부로 묘사되었습니다.",
    "entities": "기사 최씨의 얼굴은 캐릭터 레퍼런스와 일치하나, 지시된 작업용 점퍼가 아닌 이전 샷의 인물이 입었던 파란색 커버올을 입고 있습니다. 정임 역시 이전 샷의 인물과 동일한 파란색 작업복과 모자를 쓰고 있어, 이전 샷의 의상을 복사하지 말라는 지시를 위반했습니다.",
    "hard_violations": [],
    "physics": "최씨는 두 발로 서서 오른팔을 들어 유리에 손을 대고 있으며, 정임은 좌석 옆에 안정적으로 서(앉아) 있습니다. 중력에 어긋나거나 지지되지 않고 떠 있는 물체나 신체 부위는 없습니다."
   },
   {
    "label": "B",
    "direction": "정임은 우측 창밖의 최씨를 향해 시선을 두고, 최씨는 유리창 너머로 정임을 마주 봅니다. 최씨의 오른손은 유리에 닿아 있습니다.",
    "built_space": "버스 내부에서 창밖을 보는 구도이며 좌측에 하얀 좌석 커버가 있습니다. 그러나 창밖 배경이 실내 차고지가 아닌 다른 버스들이 주차된 야외 공간으로 묘사되어 장소 잠정(Location Lock)을 위반했습니다.",
    "entities": "기사 최씨의 얼굴은 레퍼런스와 일치하지만 올리브색 스웨터를 입고 있어 의상 지시와 다릅니다. 정임은 베이지색 니트를 착용하고 있습니다.",
    "hard_violations": [
     "[gemini-pro] 우측 하단 창틀에 최씨의 팔과 전혀 연결되지 않은, 잘린 형태의 흰색 장갑 낀 손이 놓여 있음 (물리적으로 불가능한 신체 구조 및 여분의 신체 부위)"
    ],
    "physics": "최씨의 오른팔은 유리를 짚고 몸을 지탱하고 있으나, 우측 하단 창틀에 놓인 흰색 장갑은 어느 팔과도 연결되지 않은 채 허공/창틀에 떠 있어 물리적으로 불가능한 상태입니다."
   }
  ],
  "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": 6,
        "verdict_ko": "지시된 카메라 구도, 샷의 앵글, 차고지 내부라는 장소 설정을 물리적 오류 없이 정확히 구현했으나, 이전 샷 인물의 의상(파란색 작업복과 모자)을 두 사람에게 그대로 복사하여 부정 지시사항을 크게 위반한 점이 감점 요인입니다."
       },
       {
        "label": "B",
        "score": 2,
        "verdict_ko": "지시된 구도는 따랐으나, 우측 하단 창틀에 팔과 연결되지 않은 정체불명의 장갑 낀 손이 등장하는 치명적인 물리적 오류(Hard Violation)가 발생했으며, 배경이 실내 차고지가 아닌 야외로 묘사되었습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "좌측 하단의 정임은 창문 너머 최씨를 향해 시선을 두고, 우측 중단의 최씨는 버스 안의 정임을 응시합니다. 최씨의 오른손은 정임을 향해 유리창에 높이 들려 있습니다.",
        "built_space": "버스 내부에서 창밖을 대각선으로 바라보는 뷰를 정확히 구현했습니다. 좌측에 하얀 커버가 씌워진 버스 좌석들이 공간감을 형성하며, 유리창에는 정임의 얼굴이 광학적으로 알맞게 반사되어 있습니다. 창밖 배경은 어두운 차고지 내부로 묘사되었습니다.",
        "entities": "기사 최씨의 얼굴은 캐릭터 레퍼런스와 일치하나, 지시된 작업용 점퍼가 아닌 이전 샷의 인물이 입었던 파란색 커버올을 입고 있습니다. 정임 역시 이전 샷의 인물과 동일한 파란색 작업복과 모자를 쓰고 있어, 이전 샷의 의상을 복사하지 말라는 지시를 위반했습니다.",
        "hard_violations": [],
        "physics": "최씨는 두 발로 서서 오른팔을 들어 유리에 손을 대고 있으며, 정임은 좌석 옆에 안정적으로 서(앉아) 있습니다. 중력에 어긋나거나 지지되지 않고 떠 있는 물체나 신체 부위는 없습니다."
       },
       {
        "label": "B",
        "direction": "정임은 우측 창밖의 최씨를 향해 시선을 두고, 최씨는 유리창 너머로 정임을 마주 봅니다. 최씨의 오른손은 유리에 닿아 있습니다.",
        "built_space": "버스 내부에서 창밖을 보는 구도이며 좌측에 하얀 좌석 커버가 있습니다. 그러나 창밖 배경이 실내 차고지가 아닌 다른 버스들이 주차된 야외 공간으로 묘사되어 장소 잠정(Location Lock)을 위반했습니다.",
        "entities": "기사 최씨의 얼굴은 레퍼런스와 일치하지만 올리브색 스웨터를 입고 있어 의상 지시와 다릅니다. 정임은 베이지색 니트를 착용하고 있습니다.",
        "hard_violations": [
         "우측 하단 창틀에 최씨의 팔과 전혀 연결되지 않은, 잘린 형태의 흰색 장갑 낀 손이 놓여 있음 (물리적으로 불가능한 신체 구조 및 여분의 신체 부위)"
        ],
        "physics": "최씨의 오른팔은 유리를 짚고 몸을 지탱하고 있으나, 우측 하단 창틀에 놓인 흰색 장갑은 어느 팔과도 연결되지 않은 채 허공/창틀에 떠 있어 물리적으로 불가능한 상태입니다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 6,
        "verdict_ko": "지시된 카메라 구도, 샷의 앵글, 차고지 내부라는 장소 설정을 물리적 오류 없이 정확히 구현했으나, 이전 샷 인물의 의상(파란색 작업복과 모자)을 두 사람에게 그대로 복사하여 부정 지시사항을 크게 위반한 점이 감점 요인입니다."
       },
       {
        "label": "B",
        "score": 2,
        "verdict_ko": "지시된 구도는 따랐으나, 우측 하단 창틀에 팔과 연결되지 않은 정체불명의 장갑 낀 손이 등장하는 치명적인 물리적 오류(Hard Violation)가 발생했으며, 배경이 실내 차고지가 아닌 야외로 묘사되었습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "좌측 하단의 정임은 창문 너머 최씨를 향해 시선을 두고, 우측 중단의 최씨는 버스 안의 정임을 응시합니다. 최씨의 오른손은 정임을 향해 유리창에 높이 들려 있습니다.",
        "built_space": "버스 내부에서 창밖을 대각선으로 바라보는 뷰를 정확히 구현했습니다. 좌측에 하얀 커버가 씌워진 버스 좌석들이 공간감을 형성하며, 유리창에는 정임의 얼굴이 광학적으로 알맞게 반사되어 있습니다. 창밖 배경은 어두운 차고지 내부로 묘사되었습니다.",
        "entities": "기사 최씨의 얼굴은 캐릭터 레퍼런스와 일치하나, 지시된 작업용 점퍼가 아닌 이전 샷의 인물이 입었던 파란색 커버올을 입고 있습니다. 정임 역시 이전 샷의 인물과 동일한 파란색 작업복과 모자를 쓰고 있어, 이전 샷의 의상을 복사하지 말라는 지시를 위반했습니다.",
        "hard_violations": [],
        "physics": "최씨는 두 발로 서서 오른팔을 들어 유리에 손을 대고 있으며, 정임은 좌석 옆에 안정적으로 서(앉아) 있습니다. 중력에 어긋나거나 지지되지 않고 떠 있는 물체나 신체 부위는 없습니다."
       },
       {
        "label": "B",
        "direction": "정임은 우측 창밖의 최씨를 향해 시선을 두고, 최씨는 유리창 너머로 정임을 마주 봅니다. 최씨의 오른손은 유리에 닿아 있습니다.",
        "built_space": "버스 내부에서 창밖을 보는 구도이며 좌측에 하얀 좌석 커버가 있습니다. 그러나 창밖 배경이 실내 차고지가 아닌 다른 버스들이 주차된 야외 공간으로 묘사되어 장소 잠정(Location Lock)을 위반했습니다.",
        "entities": "기사 최씨의 얼굴은 레퍼런스와 일치하지만 올리브색 스웨터를 입고 있어 의상 지시와 다릅니다. 정임은 베이지색 니트를 착용하고 있습니다.",
        "hard_violations": [
         "우측 하단 창틀에 최씨의 팔과 전혀 연결되지 않은, 잘린 형태의 흰색 장갑 낀 손이 놓여 있음 (물리적으로 불가능한 신체 구조 및 여분의 신체 부위)"
        ],
        "physics": "최씨의 오른팔은 유리를 짚고 몸을 지탱하고 있으나, 우측 하단 창틀에 놓인 흰색 장갑은 어느 팔과도 연결되지 않은 채 허공/창틀에 떠 있어 물리적으로 불가능한 상태입니다."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 7,
        "verdict_ko": "정임의 좌하단 배치, 창을 사이에 둔 상호 시선, 최씨의 높은 한 손 신호와 차고 배경을 가장 충실히 구현했지만, 최씨의 회색 스웨터는 참조의 노란 작업 점퍼와 다르다."
       },
       {
        "label": "B",
        "score": 6,
        "verdict_ko": "창을 사이에 둔 시선과 신호 동작은 정확하나, 정임에게 이전 장면 인물의 모자와 청색 작업복을 옮겨 입힌 점과 최씨의 참조 의상 불일치가 결정적인 감점이다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "좌하단의 정임은 고개와 시선을 오른쪽 창밖의 최씨에게 향하고, 중우측의 최씨도 창 안 정임의 얼굴을 바라본다. 높이 든 장갑 낀 손바닥은 정임을 향한 정지 신호로 읽히며 두 사람의 시선 목표가 서로 맞는다.",
        "built_space": "카메라는 버스 내부에 있고, 중앙의 큰 측면 유리창 한 면이 두 사람을 분리한다. 왼쪽 가장자리에는 새 흰색 커버가 씌워진 좌석 등받이 약 3개가 사선으로 보이며 정임은 그중 창가 좌석에 앉아 있다. 창 아래에는 고정 손잡이 1개가 있고, 창밖에는 차고 바닥과 주차된 버스들이 보여 장소가 명확하다. 유리에 약한 내부 반사가 있으나 카메라 위치상 가능한 반사다.",
        "entities": "정임은 좌하단 전경의 한국인 여성으로 구체적인 인간 형태로 표현되며, 최씨는 중우측 창밖의 50대 한국인 남성으로 짧고 단정한 검은 머리와 참조 인물에 가까운 얼굴을 지녔다. 최씨는 작업 장갑을 양손에 착용했지만 참조의 노란 작업 점퍼 대신 회색 니트형 상의를 입었다. 좌석에는 새 흰색 커버가 있고, 창밖은 버스 차고로 보인다.",
        "hard_violations": [],
        "physics": "정임은 엉덩이와 등이 실제 좌석에 지지된 자연스러운 착석 자세다. 최씨는 창밖 바닥에 서 있으며 발은 프레임 밖이지만 몸통의 수직 하중과 자세가 정상적인 기립으로 이어진다. 든 팔은 어깨에 연결되고 장갑 낀 손과 팔의 관절도 신호 동작을 물리적으로 지탱한다."
       },
       {
        "label": "B",
        "direction": "좌측 전경의 정임은 오른쪽 창밖 최씨를 바라보고, 중우측의 최씨도 정임의 얼굴 쪽으로 시선을 둔다. 최씨의 장갑 낀 손바닥은 정임 쪽으로 높이 들려 있어 목표와 신호 방향은 정확하다.",
        "built_space": "카메라는 버스 내부에 있으며 중앙 측면 유리창이 두 사람 사이의 투명한 분리면을 이룬다. 왼쪽에는 새 흰색 커버 좌석 등받이 약 2개가 사선으로 보이고 정임은 창가 좌석에 앉아 있다. 창 아래 고정 손잡이 1개와 창틀·잠금장치가 보인다. 외부는 차고 건물과 콘크리트 바닥으로 읽히지만 A보다 다른 버스와 차고의 고정 특징이 적다. 유리 속 좌석과 정임의 반사는 카메라 각도상 가능한 범위다.",
        "entities": "최씨는 참조와 가까운 50대 한국인 남성의 얼굴과 머리를 지녔고 작업 장갑을 착용했으나, 노란 작업 점퍼가 아니라 청색 일체형 작업복을 입었다. 정임은 한국인 여성으로 보이지만 이전 장면 인물과 매우 유사한 남색 모자와 청색 작업복을 착용해 이전 인물의 의복을 이 장면으로 가져오지 말라는 지시와 충돌한다. 좌석의 새 흰색 커버와 측면 창은 맞는다.",
        "hard_violations": [],
        "physics": "정임은 좌석에 엉덩이를 두고 등받이 앞에 안정적으로 앉아 있다. 최씨는 창밖 바닥에 선 수직 자세이며 하체 일부가 프레임에 포함되어 기립 지지가 자연스럽다. 높이 든 팔은 어깨와 팔꿈치가 정상적으로 연결되어 있고, 다른 손도 몸 옆에 자연스럽게 내려와 있어 떠 있거나 지지 없는 물체는 없다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 7,
        "verdict_ko": "정임의 좌하단 배치, 창을 사이에 둔 상호 시선, 최씨의 높은 한 손 신호와 차고 배경을 가장 충실히 구현했지만, 최씨의 회색 스웨터는 참조의 노란 작업 점퍼와 다르다."
       },
       {
        "label": "A",
        "score": 6,
        "verdict_ko": "창을 사이에 둔 시선과 신호 동작은 정확하나, 정임에게 이전 장면 인물의 모자와 청색 작업복을 옮겨 입힌 점과 최씨의 참조 의상 불일치가 결정적인 감점이다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "B",
        "direction": "좌하단의 정임은 고개와 시선을 오른쪽 창밖의 최씨에게 향하고, 중우측의 최씨도 창 안 정임의 얼굴을 바라본다. 높이 든 장갑 낀 손바닥은 정임을 향한 정지 신호로 읽히며 두 사람의 시선 목표가 서로 맞는다.",
        "built_space": "카메라는 버스 내부에 있고, 중앙의 큰 측면 유리창 한 면이 두 사람을 분리한다. 왼쪽 가장자리에는 새 흰색 커버가 씌워진 좌석 등받이 약 3개가 사선으로 보이며 정임은 그중 창가 좌석에 앉아 있다. 창 아래에는 고정 손잡이 1개가 있고, 창밖에는 차고 바닥과 주차된 버스들이 보여 장소가 명확하다. 유리에 약한 내부 반사가 있으나 카메라 위치상 가능한 반사다.",
        "entities": "정임은 좌하단 전경의 한국인 여성으로 구체적인 인간 형태로 표현되며, 최씨는 중우측 창밖의 50대 한국인 남성으로 짧고 단정한 검은 머리와 참조 인물에 가까운 얼굴을 지녔다. 최씨는 작업 장갑을 양손에 착용했지만 참조의 노란 작업 점퍼 대신 회색 니트형 상의를 입었다. 좌석에는 새 흰색 커버가 있고, 창밖은 버스 차고로 보인다.",
        "hard_violations": [],
        "physics": "정임은 엉덩이와 등이 실제 좌석에 지지된 자연스러운 착석 자세다. 최씨는 창밖 바닥에 서 있으며 발은 프레임 밖이지만 몸통의 수직 하중과 자세가 정상적인 기립으로 이어진다. 든 팔은 어깨에 연결되고 장갑 낀 손과 팔의 관절도 신호 동작을 물리적으로 지탱한다."
       },
       {
        "label": "A",
        "direction": "좌측 전경의 정임은 오른쪽 창밖 최씨를 바라보고, 중우측의 최씨도 정임의 얼굴 쪽으로 시선을 둔다. 최씨의 장갑 낀 손바닥은 정임 쪽으로 높이 들려 있어 목표와 신호 방향은 정확하다.",
        "built_space": "카메라는 버스 내부에 있으며 중앙 측면 유리창이 두 사람 사이의 투명한 분리면을 이룬다. 왼쪽에는 새 흰색 커버 좌석 등받이 약 2개가 사선으로 보이고 정임은 창가 좌석에 앉아 있다. 창 아래 고정 손잡이 1개와 창틀·잠금장치가 보인다. 외부는 차고 건물과 콘크리트 바닥으로 읽히지만 A보다 다른 버스와 차고의 고정 특징이 적다. 유리 속 좌석과 정임의 반사는 카메라 각도상 가능한 범위다.",
        "entities": "최씨는 참조와 가까운 50대 한국인 남성의 얼굴과 머리를 지녔고 작업 장갑을 착용했으나, 노란 작업 점퍼가 아니라 청색 일체형 작업복을 입었다. 정임은 한국인 여성으로 보이지만 이전 장면 인물과 매우 유사한 남색 모자와 청색 작업복을 착용해 이전 인물의 의복을 이 장면으로 가져오지 말라는 지시와 충돌한다. 좌석의 새 흰색 커버와 측면 창은 맞는다.",
        "hard_violations": [],
        "physics": "정임은 좌석에 엉덩이를 두고 등받이 앞에 안정적으로 앉아 있다. 최씨는 창밖 바닥에 선 수직 자세이며 하체 일부가 프레임에 포함되어 기립 지지가 자연스럽다. 높이 든 팔은 어깨와 팔꿈치가 정상적으로 연결되어 있고, 다른 손도 몸 옆에 자연스럽게 내려와 있어 떠 있거나 지지 없는 물체는 없다."
       }
      ]
     },
     "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": 1.857,
    "B": 1.333
   },
   "adjusted": {
    "A": 1.857,
    "B": 1.083
   },
   "violations": {
    "B": [
     "[gemini-pro] 우측 하단 창틀에 최씨의 팔과 전혀 연결되지 않은, 잘린 형태의 흰색 장갑 낀 손이 놓여 있음 (물리적으로 불가능한 신체 구조 및 여분의 신체 부위)"
    ]
   },
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "B"
   },
   "agreed": false
  },
  "totals": {
   "A": 1857,
   "B": 1083
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1857,
    "verdict_ko": "지시된 카메라 구도, 샷의 앵글, 차고지 내부라는 장소 설정을 물리적 오류 없이 정확히 구현했으나, 이전 샷 인물의 의상(파란색 작업복과 모자)을 두 사람에게 그대로 복사하여 부정 지시사항을 크게 위반한 점이 감점 요인입니다."
   },
   {
    "label": "B",
    "score": 1083,
    "verdict_ko": "지시된 구도는 따랐으나, 우측 하단 창틀에 팔과 연결되지 않은 정체불명의 장갑 낀 손이 등장하는 치명적인 물리적 오류(Hard Violation)가 발생했으며, 배경이 실내 차고지가 아닌 야외로 묘사되었습니다.  ★위반: [gemini-pro] 우측 하단 창틀에 최씨의 팔과 전혀 연결되지 않은, 잘린 형태의 흰색 장갑 낀 손이 놓여 있음 (물리적으로 불가능한 신체 구조 및 여분의 신체 부위)"
   }
  ],
  "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/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S2sh1_sel.png",
    "asset_id": "ffc1788e-4931-49c9-ab4d-90669d754d9c",
    "role": "prev_still"
   },
   {
    "label": "CHARACTER REFERENCE — 기사 최씨: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:952050>",
    "asset_id": "e0d2ec85-1698-4efc-ab24-a73e30964534",
    "role": "character_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a92ab1-6893-71fc-a7e8-48c0ca9591d2",
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S2sh1"
  }
 },
 "S2sh4::cine": {
  "applied": true,
  "attempted_at": "2026-08-29T09:51:11.244359+00:00",
  "fingerprint": "0eb5a2de312c8967352a0ae98e7010388f489b6179c766e6ed6b405d1a28bd7e",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S2sh4_sel.png",
  "source_sha256": "f65e5055b1f0c03776563bf61a82b89497afa16992166f76191209ff4632327c",
  "file": "S2sh4_cine.png",
  "staged_sha256": "abb55d5d8c51bfa1f5cf21cc933eb8e2a3d0bce64b0cb4575b6672c4bd8d3a85",
  "latency_ms": 84496,
  "request_id": "01a04cee-325c-7af0-8a6f-f2f4458ee1a5",
  "status_url": "https://queue.fal.run/reve/2.1/requests/01a04cee-325c-7af0-8a6f-f2f4458ee1a5/status",
  "response_url": "https://queue.fal.run/reve/2.1/requests/01a04cee-325c-7af0-8a6f-f2f4458ee1a5",
  "submitted_at": "2026-08-29T09:51:14.077434+00:00"
 },
 "S3sh4::confined_fp_apt": {
  "applies": false,
  "reason_ko": "이 샷은 회수권 뭉치를 들고 있는 손을 보여주는 타이트한 클로즈업이므로, 버스 내부의 좌석 배치나 통제 구역의 위치 관계를 파악하기 위한 평면도 레이아웃 보조가 필요하지 않습니다.",
  "input_fingerprint": "d533bee0eef7c9c8"
 },
 "S3sh4::signage": {
  "fp": "e22fce1c4ddb36be",
  "inscriptions": [],
  "cues": [
   {
    "text_native": "",
    "source": "scene_text_implied",
    "source_quote": "회수권 뭉치"
   }
  ],
  "dropped": []
 },
 "S3sh4::bgfirst_bg": {
  "input_fingerprint": "99ceceaa3c0fe59a",
  "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): 정임의 펼쳐진 한쪽 손바닥 위에 놓인, 표지가 낡게 접힌 회수권 뭉치를 다른 손 검지로 짚고 있는 클로즈업.\n\nLOCATION (lock): Inside the first intercity bus beside the seat directly behind the driver, near the front aisle and chained fare box. The windows beyond are dark with night.\n\nTIME OF DAY (lock): night.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 회수권 뭉치 (표지가 접혀 있고 고무줄로 묶인 채 펼친 손바닥 위에 놓여 있음) — The folded cover and layered ticket edges face upward toward the camera while 정임's index finger marks the count; used as Central insert subject, framed together with both hands and sleeve edges for physical scale; 좌석 아래쪽 (정임이 회수권 뭉치를 찾아낸 좌석 주변임) — Only an oblique portion remains visible at the upper frame edge; used as Minimal environmental context preventing the hand detail from becoming abstract; 고무 매트가 깔린 통로 (1호차 통로에 그대로 깔려 있음); used as Peripheral ground reference beneath the hands.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. 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: Low-key ambient exposure appropriate to the nighttime bus interior uses muted navy, aged beige, subdued paper tones, moderate contrast, and tactile film grain.\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.\n\nPERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
  "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): 정임의 펼쳐진 한쪽 손바닥 위에 놓인, 표지가 낡게 접힌 회수권 뭉치를 다른 손 검지로 짚고 있는 클로즈업.\n\nLOCATION (lock): Inside the first intercity bus beside the seat directly behind the driver, near the front aisle and chained fare box. The windows beyond are dark with night.\n\nTIME OF DAY (lock): night.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 회수권 뭉치 (표지가 접혀 있고 고무줄로 묶인 채 펼친 손바닥 위에 놓여 있음) — The folded cover and layered ticket edges face upward toward the camera while 정임's index finger marks the count; used as Central insert subject, framed together with both hands and sleeve edges for physical scale; 좌석 아래쪽 (정임이 회수권 뭉치를 찾아낸 좌석 주변임) — Only an oblique portion remains visible at the upper frame edge; used as Minimal environmental context preventing the hand detail from becoming abstract; 고무 매트가 깔린 통로 (1호차 통로에 그대로 깔려 있음); used as Peripheral ground reference beneath the hands.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. 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: Low-key ambient exposure appropriate to the nighttime bus interior uses muted navy, aged beige, subdued paper tones, moderate contrast, and tactile film grain.\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.\n\nPERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
  "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S3sh4__bgfirst_bg.png",
  "asset_id": "a05c8da3-015c-475a-9dfd-fad1096c1e26",
  "input_asset_ids": [
   "0a75eeb5-c326-4b30-a5a7-be04460eba29",
   "c7cff2d3-ee5c-43de-85e5-5b8dad4fe3f2"
  ],
  "era_research": {
   "subject": "1980년대 한국 시외버스 내부 및 요금통",
   "queries": [
    [
     "1980년대 한국 시외버스 내부 좌석 운전석",
     "옛날 한국 버스 요금통 1980년대 시내버스 내부"
    ]
   ],
   "picked_url": "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
   "sha256": "587de5353dace6b5a009871a0c0885a7670336392d63862d50b3f47e7d8fe625",
   "file": "eraref_8f484ee77a47c9d6.png"
  }
 },
 "S3sh4": {
  "input_fingerprint": "2c30236ab310bb12",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 정임의 펼쳐진 한쪽 손바닥 위에 놓인, 표지가 낡게 접힌 회수권 뭉치를 다른 손 검지로 짚고 있는 클로즈업.\n\nLOCATION (lock): Inside the first intercity bus beside the seat directly behind the driver, near the front aisle and chained fare box. The windows beyond are dark with night. 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- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 회수권 뭉치 (표지가 접혀 있고 고무줄로 묶인 채 펼친 손바닥 위에 놓여 있음) — The folded cover and layered ticket edges face upward toward the camera while 정임's index finger marks the count; used as Central insert subject, framed together with both hands and sleeve edges for physical scale; 좌석 아래쪽 (정임이 회수권 뭉치를 찾아낸 좌석 주변임) — Only an oblique portion remains visible at the upper frame edge; used as Minimal environmental context preventing the hand detail from becoming abstract; 고무 매트가 깔린 통로 (1호차 통로에 그대로 깔려 있음); used as Peripheral ground reference beneath the hands.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. 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: Low-key ambient exposure appropriate to the nighttime bus interior uses muted navy, aged beige, subdued paper tones, moderate contrast, and tactile film grain.\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): Bus 1 retains its worn vinyl seat covers, rubber aisle mat, and chained farebox in the same position, while the windows are dark outside. The rubber-banded ticket bundle has been removed from beneath the seat, and its worn folded cover is visible. 정임: She wears her navy conductor's uniform and billed cap, standing near the farebox with the ticket bundle supported on one open palm and counting it with the opposite index finger.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
  "roll_prompts": {
   "A": "Stage the shot inside a tight, built interior. The FIRST attached image (SHOT BACKGROUND) is the finished empty interior of this shot, and in a space this cramped its geometry is the truth of the shot — keep it EXACTLY: its camera, perspective and every fixed fitting it has stay untouched, in the same place, at the same angle, in the same number.\n\nBefore you place anyone, count what the background shows: every fixed fitting this space has, how many of each, which way each faces, and what any reflective surface among them could return from this camera. Those counts and placements are what you must still be able to make after the people are in. Adding a second one of anything, moving a fitting or growing a new surface is a failure even when the person looks right.\n\nThe SECOND attached image (LAYOUT SKETCH) tells you where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Its background lines are not decoration — they are the same structure seen in line form, so use them to place each person correctly with respect to it: where the body rests and on what, which side of a fitting the hands are on, what the body passes in front of and what it passes behind. Where sketch and background disagree about the structure itself, the background wins.\n\nThe CHARACTER REFERENCE photographs show the real people.\n\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. A person may cover part of the structure — that is expected, and covering is not redrawing. What the body hides stays hidden; what remains visible stays exactly as the background had it. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 정임의 펼쳐진 한쪽 손바닥 위에 놓인, 표지가 낡게 접힌 회수권 뭉치를 다른 손 검지로 짚고 있는 클로즈업.\n\nLOCATION (lock): Inside the first intercity bus beside the seat directly behind the driver, near the front aisle and chained fare box. The windows beyond are dark with night. 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- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 회수권 뭉치 (표지가 접혀 있고 고무줄로 묶인 채 펼친 손바닥 위에 놓여 있음) — The folded cover and layered ticket edges face upward toward the camera while 정임's index finger marks the count; used as Central insert subject, framed together with both hands and sleeve edges for physical scale; 좌석 아래쪽 (정임이 회수권 뭉치를 찾아낸 좌석 주변임) — Only an oblique portion remains visible at the upper frame edge; used as Minimal environmental context preventing the hand detail from becoming abstract; 고무 매트가 깔린 통로 (1호차 통로에 그대로 깔려 있음); used as Peripheral ground reference beneath the hands.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. 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: Low-key ambient exposure appropriate to the nighttime bus interior uses muted navy, aged beige, subdued paper tones, moderate contrast, and tactile film grain.\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): Bus 1 retains its worn vinyl seat covers, rubber aisle mat, and chained farebox in the same position, while the windows are dark outside. The rubber-banded ticket bundle has been removed from beneath the seat, and its worn folded cover is visible. 정임: She wears her navy conductor's uniform and billed cap, standing near the farebox with the ticket bundle supported on one open palm and counting it with the opposite index finger.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 정임의 펼쳐진 한쪽 손바닥 위에 놓인, 표지가 낡게 접힌 회수권 뭉치를 다른 손 검지로 짚고 있는 클로즈업.\n\nLOCATION (lock): Inside the first intercity bus beside the seat directly behind the driver, near the front aisle and chained fare box. The windows beyond are dark with night. 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- FRAMING SCALE: insert close-up on a detail\n- KEY BACKGROUND ELEMENTS: 회수권 뭉치 (표지가 접혀 있고 고무줄로 묶인 채 펼친 손바닥 위에 놓여 있음) — The folded cover and layered ticket edges face upward toward the camera while 정임's index finger marks the count; used as Central insert subject, framed together with both hands and sleeve edges for physical scale; 좌석 아래쪽 (정임이 회수권 뭉치를 찾아낸 좌석 주변임) — Only an oblique portion remains visible at the upper frame edge; used as Minimal environmental context preventing the hand detail from becoming abstract; 고무 매트가 깔린 통로 (1호차 통로에 그대로 깔려 있음); used as Peripheral ground reference beneath the hands.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. 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: Low-key ambient exposure appropriate to the nighttime bus interior uses muted navy, aged beige, subdued paper tones, moderate contrast, and tactile film grain.\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): Bus 1 retains its worn vinyl seat covers, rubber aisle mat, and chained farebox in the same position, while the windows are dark outside. The rubber-banded ticket bundle has been removed from beneath the seat, and its worn folded cover is visible. 정임: She wears her navy conductor's uniform and billed cap, standing near the farebox with the ticket bundle supported on one open palm and counting it with the opposite index finger.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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\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/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S3sh4__bgfirst_bg.png",
     "asset_id": "a05c8da3-015c-475a-9dfd-fad1096c1e26",
     "role": "bgfirst_bg"
    },
    {
     "label": "LAYOUT SKETCH (people placement AND the structure they sit inside — the sketched fixed fittings and openings are the same ones the background photograph shows, drawn as lines; read them to place each body correctly against that structure, and where the two disagree about the structure itself the background photograph wins)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/conti/conti_S3sh4.png",
     "asset_id": "0a75eeb5-c326-4b30-a5a7-be04460eba29",
     "role": "conti_light"
    },
    {
     "label": "CHARACTER REFERENCE — 정임: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "PROP REFERENCE — 고무줄로 묶인 낡은 종이 회수권 뭉치: the exact object appearing in this shot; match its look, material and wear exactly.",
     "path": "<bytes:1135067>",
     "asset_id": "6e82ce93-8876-47c1-ac1d-48cf6c9a06d7",
     "role": "prop_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/da049582-2c6d-492c-979d-f468d61bab6e/episodes/97375a4b-b465-4fd6-9a2d-800e0eb01685/images/background_chain/L01B02.png",
     "asset_id": "c7cff2d3-ee5c-43de-85e5-5b8dad4fe3f2",
     "role": "location_plate"
    },
    {
     "label": "CHARACTER REFERENCE — 정임: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "PROP REFERENCE — 고무줄로 묶인 낡은 종이 회수권 뭉치: the exact object appearing in this shot; match its look, material and wear exactly.",
     "path": "<bytes:1135067>",
     "asset_id": "6e82ce93-8876-47c1-ac1d-48cf6c9a06d7",
     "role": "prop_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "judge_flip_skipped": "judge_fn_owns_order",
  "readings": [
   {
    "label": "A",
    "direction": "오른손 검지가 왼손 바닥 위에 놓인 회수권 뭉치의 모서리를 가리키고 있음.",
    "built_space": "버스 내부 통로에 고무 매트가 깔려 있으며, 왼쪽에는 자주색 좌석의 하단부가, 오른쪽에는 체인이 달린 금속 요금함이 레퍼런스와 동일한 형태와 위치로 존재함.",
    "entities": "네이비색 유니폼을 입은 인물(정임)의 두 손이 프레임에 잡히며, 고무줄로 묶이고 낡은 종이 질감을 띤 회수권 뭉치가 레퍼런스와 정확히 일치함.",
    "hard_violations": [],
    "physics": "회수권 뭉치는 위를 향해 펼쳐진 왼손 바닥에 의해 안정적으로 지지되고 있으며, 포즈와 중력의 작용이 자연스러움."
   },
   {
    "label": "B",
    "direction": "왼손 검지가 오른손 바닥 위에 놓인 회수권 뭉치의 윗면을 짚고 있음.",
    "built_space": "바닥에 고무 매트가 보이나 좌석 하단부의 재질과 색상이 낡은 회색조로 렌더링되어 레퍼런스와 어긋나며, 지정된 요금함이 프레임 내에 전혀 보이지 않음.",
    "entities": "네이비색 유니폼을 입은 양손, 고무줄로 묶인 회수권 뭉치가 레퍼런스와 일치하게 묘사됨.",
    "hard_violations": [],
    "physics": "오른손 바닥이 회수권 뭉치의 무게를 받치고 있으며 물리적으로 어긋남 없이 지지됨."
   }
  ],
  "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": "프롬프트가 요구한 클로즈업 앵글과 손의 자세를 훌륭하게 구현했으며, 레퍼런스에 있는 자주색 좌석, 고무 매트, 체인이 달린 요금함을 지정된 위치와 스케일에 맞게 완벽히 재현했습니다."
       },
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "회수권 뭉치와 손의 묘사는 양호하나, 필수 배경 요소인 체인이 달린 요금함이 누락되었고 좌석의 색상과 재질이 레퍼런스(자주색)와 일치하지 않습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "오른손 검지가 왼손 바닥 위에 놓인 회수권 뭉치의 모서리를 가리키고 있음.",
        "built_space": "버스 내부 통로에 고무 매트가 깔려 있으며, 왼쪽에는 자주색 좌석의 하단부가, 오른쪽에는 체인이 달린 금속 요금함이 레퍼런스와 동일한 형태와 위치로 존재함.",
        "entities": "네이비색 유니폼을 입은 인물(정임)의 두 손이 프레임에 잡히며, 고무줄로 묶이고 낡은 종이 질감을 띤 회수권 뭉치가 레퍼런스와 정확히 일치함.",
        "hard_violations": [],
        "physics": "회수권 뭉치는 위를 향해 펼쳐진 왼손 바닥에 의해 안정적으로 지지되고 있으며, 포즈와 중력의 작용이 자연스러움."
       },
       {
        "label": "B",
        "direction": "왼손 검지가 오른손 바닥 위에 놓인 회수권 뭉치의 윗면을 짚고 있음.",
        "built_space": "바닥에 고무 매트가 보이나 좌석 하단부의 재질과 색상이 낡은 회색조로 렌더링되어 레퍼런스와 어긋나며, 지정된 요금함이 프레임 내에 전혀 보이지 않음.",
        "entities": "네이비색 유니폼을 입은 양손, 고무줄로 묶인 회수권 뭉치가 레퍼런스와 일치하게 묘사됨.",
        "hard_violations": [],
        "physics": "오른손 바닥이 회수권 뭉치의 무게를 받치고 있으며 물리적으로 어긋남 없이 지지됨."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 9,
        "verdict_ko": "프롬프트가 요구한 클로즈업 앵글과 손의 자세를 훌륭하게 구현했으며, 레퍼런스에 있는 자주색 좌석, 고무 매트, 체인이 달린 요금함을 지정된 위치와 스케일에 맞게 완벽히 재현했습니다."
       },
       {
        "label": "B",
        "score": 5,
        "verdict_ko": "회수권 뭉치와 손의 묘사는 양호하나, 필수 배경 요소인 체인이 달린 요금함이 누락되었고 좌석의 색상과 재질이 레퍼런스(자주색)와 일치하지 않습니다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "A",
        "direction": "오른손 검지가 왼손 바닥 위에 놓인 회수권 뭉치의 모서리를 가리키고 있음.",
        "built_space": "버스 내부 통로에 고무 매트가 깔려 있으며, 왼쪽에는 자주색 좌석의 하단부가, 오른쪽에는 체인이 달린 금속 요금함이 레퍼런스와 동일한 형태와 위치로 존재함.",
        "entities": "네이비색 유니폼을 입은 인물(정임)의 두 손이 프레임에 잡히며, 고무줄로 묶이고 낡은 종이 질감을 띤 회수권 뭉치가 레퍼런스와 정확히 일치함.",
        "hard_violations": [],
        "physics": "회수권 뭉치는 위를 향해 펼쳐진 왼손 바닥에 의해 안정적으로 지지되고 있으며, 포즈와 중력의 작용이 자연스러움."
       },
       {
        "label": "B",
        "direction": "왼손 검지가 오른손 바닥 위에 놓인 회수권 뭉치의 윗면을 짚고 있음.",
        "built_space": "바닥에 고무 매트가 보이나 좌석 하단부의 재질과 색상이 낡은 회색조로 렌더링되어 레퍼런스와 어긋나며, 지정된 요금함이 프레임 내에 전혀 보이지 않음.",
        "entities": "네이비색 유니폼을 입은 양손, 고무줄로 묶인 회수권 뭉치가 레퍼런스와 일치하게 묘사됨.",
        "hard_violations": [],
        "physics": "오른손 바닥이 회수권 뭉치의 무게를 받치고 있으며 물리적으로 어긋남 없이 지지됨."
       }
      ]
     },
     "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": "펼친 손바닥 위의 낡고 접힌 고무줄 회수권 뭉치를 반대편 검지로 짚는 삽입 클로즈업과 남색 제복 소매, 최소한의 좌석·고무 통로 맥락을 가장 정확히 구현했으나 표면 숫자가 일부 식별된다."
       },
       {
        "label": "B",
        "score": 6,
        "verdict_ko": "손바닥과 검지의 기본 동작은 맞지만 검지가 표지 아래쪽을 짚어 계수 동작이 덜 명확하고, 회수권 형태와 흰 셔츠가 드러난 재킷형 소매가 참고물과 다르며 요금함이 지나치게 크게 포함되어 삽입 프레이밍이 약해졌다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "시선이나 무기는 보이지 않는다. 위쪽 손의 검지가 손바닥 위 회수권 뭉치의 상단 인쇄·층 부분을 정확히 향해 닿아 있어 표를 세는 동작으로 읽힌다.",
        "built_space": "화면 위쪽에 좌석 하부와 금속 프레임 일부가 하나 보이고, 손 아래와 왼쪽 대부분에는 홈이 난 고무 통로 매트가 보인다. 요금함과 창문은 타이트한 삽입 화면 밖이며, 좌석 뒤편 전방 통로에서 찍은 배치로 무리 없이 성립한다. 반사 문제는 없다.",
        "entities": "인물은 얼굴 없이 정임의 두 손·손목·팔 일부만 보인다. 손은 젊은 여성으로 자연스럽게 읽히고 남색 작업복형 제복 소매도 참고 인물과 부합한다. 중앙 물체는 낡은 종이층, 접힌 표지와 갈색 고무줄을 갖춘 회수권 뭉치로 참고 소품과 매우 가깝다. 다만 표면의 일부 숫자는 식별 가능해 ‘읽을 수 있는 글자 없음’ 조건에는 어긋난다.",
        "hard_violations": [],
        "physics": "회수권 뭉치는 아래쪽의 펼친 손바닥 전체가 받치고 있으며 고무줄이 종이층을 고정한다. 반대편 검지는 실제로 표면에 접촉하고 두 손 모두 소매와 팔에 자연스럽게 이어진다. 떠 있거나 지지되지 않은 물체는 없다."
       },
       {
        "label": "B",
        "direction": "시선이나 무기는 보이지 않는다. 아래쪽 손의 검지가 손바닥 위 회수권의 아래쪽 접힌 표지 모서리를 향해 닿는다. 대상 자체는 맞지만 표의 층이나 계수 칸보다 빈 표지 부분을 짚어 ‘수를 세는’ 방향성은 A보다 약하다.",
        "built_space": "왼쪽 위에 좌석 하나의 하부와 금속 지지대가 보이고, 중앙에는 고무 통로 매트, 오른쪽에는 체인이 달린 요금함 하나와 출입문 가장자리가 보인다. 전방 통로의 구조 관계는 가능하고 고정물 중복이나 불가능한 반사는 없다. 다만 요금함이 화면 오른쪽을 크게 차지해 좌석 가장자리만 최소 맥락으로 남기라는 삽입 구성보다 환경을 과도하게 보여준다.",
        "entities": "정임으로 지정된 인물은 두 손과 팔 일부만 보이며 별도 인물은 없다. 손은 젊은 성인 여성으로 읽힐 수 있으나 남색 소매가 흰 셔츠 소맷단을 갖춘 재킷처럼 보여 참고한 작업복형 차장 제복과 차이가 크다. 회수권은 낡은 종이와 고무줄을 갖췄지만 참고 소품보다 길고 평평하며, 감싸듯 접힌 두꺼운 표지 형태가 덜 일치한다. 요금함과 좌석, 통로 매트는 버스 전방 장소를 나타낸다.",
        "hard_violations": [],
        "physics": "회수권은 펼친 손바닥과 손가락 위에 안정적으로 놓였고 고무줄로 묶여 있다. 반대 검지가 표면에 접촉하며 손과 팔의 연결도 자연스럽다. 모든 물체에 보이는 지지가 있고 부유하거나 불가능한 자세는 없다."
       }
      ],
      "all_candidates_fail": false
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 9,
        "verdict_ko": "펼친 손바닥 위의 낡고 접힌 고무줄 회수권 뭉치를 반대편 검지로 짚는 삽입 클로즈업과 남색 제복 소매, 최소한의 좌석·고무 통로 맥락을 가장 정확히 구현했으나 표면 숫자가 일부 식별된다."
       },
       {
        "label": "A",
        "score": 6,
        "verdict_ko": "손바닥과 검지의 기본 동작은 맞지만 검지가 표지 아래쪽을 짚어 계수 동작이 덜 명확하고, 회수권 형태와 흰 셔츠가 드러난 재킷형 소매가 참고물과 다르며 요금함이 지나치게 크게 포함되어 삽입 프레이밍이 약해졌다."
       }
      ],
      "all_candidates_fail": false,
      "readings": [
       {
        "label": "B",
        "direction": "시선이나 무기는 보이지 않는다. 위쪽 손의 검지가 손바닥 위 회수권 뭉치의 상단 인쇄·층 부분을 정확히 향해 닿아 있어 표를 세는 동작으로 읽힌다.",
        "built_space": "화면 위쪽에 좌석 하부와 금속 프레임 일부가 하나 보이고, 손 아래와 왼쪽 대부분에는 홈이 난 고무 통로 매트가 보인다. 요금함과 창문은 타이트한 삽입 화면 밖이며, 좌석 뒤편 전방 통로에서 찍은 배치로 무리 없이 성립한다. 반사 문제는 없다.",
        "entities": "인물은 얼굴 없이 정임의 두 손·손목·팔 일부만 보인다. 손은 젊은 여성으로 자연스럽게 읽히고 남색 작업복형 제복 소매도 참고 인물과 부합한다. 중앙 물체는 낡은 종이층, 접힌 표지와 갈색 고무줄을 갖춘 회수권 뭉치로 참고 소품과 매우 가깝다. 다만 표면의 일부 숫자는 식별 가능해 ‘읽을 수 있는 글자 없음’ 조건에는 어긋난다.",
        "hard_violations": [],
        "physics": "회수권 뭉치는 아래쪽의 펼친 손바닥 전체가 받치고 있으며 고무줄이 종이층을 고정한다. 반대편 검지는 실제로 표면에 접촉하고 두 손 모두 소매와 팔에 자연스럽게 이어진다. 떠 있거나 지지되지 않은 물체는 없다."
       },
       {
        "label": "A",
        "direction": "시선이나 무기는 보이지 않는다. 아래쪽 손의 검지가 손바닥 위 회수권의 아래쪽 접힌 표지 모서리를 향해 닿는다. 대상 자체는 맞지만 표의 층이나 계수 칸보다 빈 표지 부분을 짚어 ‘수를 세는’ 방향성은 A보다 약하다.",
        "built_space": "왼쪽 위에 좌석 하나의 하부와 금속 지지대가 보이고, 중앙에는 고무 통로 매트, 오른쪽에는 체인이 달린 요금함 하나와 출입문 가장자리가 보인다. 전방 통로의 구조 관계는 가능하고 고정물 중복이나 불가능한 반사는 없다. 다만 요금함이 화면 오른쪽을 크게 차지해 좌석 가장자리만 최소 맥락으로 남기라는 삽입 구성보다 환경을 과도하게 보여준다.",
        "entities": "정임으로 지정된 인물은 두 손과 팔 일부만 보이며 별도 인물은 없다. 손은 젊은 성인 여성으로 읽힐 수 있으나 남색 소매가 흰 셔츠 소맷단을 갖춘 재킷처럼 보여 참고한 작업복형 차장 제복과 차이가 크다. 회수권은 낡은 종이와 고무줄을 갖췄지만 참고 소품보다 길고 평평하며, 감싸듯 접힌 두꺼운 표지 형태가 덜 일치한다. 요금함과 좌석, 통로 매트는 버스 전방 장소를 나타낸다.",
        "hard_violations": [],
        "physics": "회수권은 펼친 손바닥과 손가락 위에 안정적으로 놓였고 고무줄로 묶여 있다. 반대 검지가 표면에 접촉하며 손과 팔의 연결도 자연스럽다. 모든 물체에 보이는 지지가 있고 부유하거나 불가능한 자세는 없다."
       }
      ]
     },
     "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": 1.667,
    "B": 1.556
   },
   "adjusted": {
    "A": 1.667,
    "B": 1.556
   },
   "violations": {},
   "per_model_winner": {
    "gemini-pro": "A",
    "gpt": "B"
   },
   "agreed": false
  },
  "totals": {
   "A": 1667,
   "B": 1556
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1667,
    "verdict_ko": "프롬프트가 요구한 클로즈업 앵글과 손의 자세를 훌륭하게 구현했으며, 레퍼런스에 있는 자주색 좌석, 고무 매트, 체인이 달린 요금함을 지정된 위치와 스케일에 맞게 완벽히 재현했습니다."
   },
   {
    "label": "B",
    "score": 1556,
    "verdict_ko": "회수권 뭉치와 손의 묘사는 양호하나, 필수 배경 요소인 체인이 달린 요금함이 누락되었고 좌석의 색상과 재질이 레퍼런스(자주색)와 일치하지 않습니다."
   }
  ],
  "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/da049582-2c6d-492c-979d-f468d61bab6e/episodes/97375a4b-b465-4fd6-9a2d-800e0eb01685/images/background_chain/L01B02.png",
    "asset_id": "c7cff2d3-ee5c-43de-85e5-5b8dad4fe3f2",
    "role": "location_plate"
   },
   {
    "label": "CHARACTER REFERENCE — 정임: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:812895>",
    "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
    "role": "character_ref"
   },
   {
    "label": "PROP REFERENCE — 고무줄로 묶인 낡은 종이 회수권 뭉치: the exact object appearing in this shot; match its look, material and wear exactly.",
    "path": "<bytes:1135067>",
    "asset_id": "6e82ce93-8876-47c1-ac1d-48cf6c9a06d7",
    "role": "prop_ref"
   }
  ],
  "critique_skipped": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a92abe-3ced-7384-81d9-64665f00c624",
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S3sh4__bgfirst_bg.png",
   "bg_asset_id": "a05c8da3-015c-475a-9dfd-fad1096c1e26",
   "bg_record_key": "S3sh4::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S3sh4::cine": {
  "applied": true,
  "attempted_at": "2026-08-29T09:55:10.659602+00:00",
  "fingerprint": "328138ebdf092fe813529f550c2e09433ff4849a5d035147f96e7753b36c6f72",
  "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": "d4dadbb410ebe7470e15f484280e51ea52c8268efd7d0828cefe47e00f69ffb0",
  "file": "S3sh4_cine.png",
  "staged_sha256": "476b547ac75b0cb3de40e8deadf71c09838fb9581480ab229e4c7a743541b480",
  "latency_ms": 68164,
  "request_id": "01a04cf1-dcc7-72d0-90dd-5f4cb44d6bcb",
  "status_url": "https://queue.fal.run/reve/2.1/requests/01a04cf1-dcc7-72d0-90dd-5f4cb44d6bcb/status",
  "response_url": "https://queue.fal.run/reve/2.1/requests/01a04cf1-dcc7-72d0-90dd-5f4cb44d6bcb",
  "submitted_at": "2026-08-29T09:55:14.316507+00:00"
 },
 "S3sh6::confined_fp_apt": {
  "applies": true,
  "reason_ko": "버스 내부의 운전석, 앞문, 승객석 등 구체적인 공간 배치가 명확해야 하며, 두 인물의 위치와 시선 방향이 중요하게 작용하는 장면이므로 평면도 보조가 필요합니다.",
  "input_fingerprint": "a341cd47521a460d"
 },
 "S3sh6::signage": {
  "fp": "d894b0fce9db5347",
  "inscriptions": [],
  "cues": [
   {
    "text_native": "",
    "source": "scene_text_implied",
    "source_quote": "회수권 뭉치"
   }
  ],
  "dropped": []
 },
 "confinedfp::0586889cbde5": {
  "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/confinedfp_base_0586889cbde5.png",
  "place_text": "Inside the front passenger area of the first intercity bus, beside the seat behind the driver and close to the front entry. The bus windows are dark at night.",
  "input_fingerprint": "8e1035dd9d3669e6"
 },
 "S3sh6::confined_fp": {
  "reads": {
   "controls": "A bell control is attached to the Front Passenger Seat.",
   "mirrors": "No mirrors are present in the diagram.",
   "camera": "The camera stands in the central aisle behind row 3, pointing straight forward toward the front of the bus.",
   "occupants": "Seat 1B is occupied by 정임. The Front Entry Area is occupied by 기사 최씨."
  },
  "mismatches": [],
  "scene_description_en": "The camera is positioned in the central aisle, facing directly forward to capture a medium shot of the dimly lit bus interior. On the near left side of the frame, 정임 sits in the aisle-adjacent Seat 1B, turning her head rightward across the aisle. In the lower-center of the view, she tightly grips a folded, rubber-banded bundle of bus tickets. On the near right side of the frame, 기사 최씨 stands in the Front Entry Area, facing left toward 정임 while resting one hand on a worn seatback. The rubber-matted aisle extends straight up the center of the screen from near to far, dividing the two characters. In the far right background, the Front Passenger Seat is visible bearing a bell control. The side windows on the far left and right reveal a pitch-black exterior, isolating the subjects within the aged interior.",
  "fixed": true,
  "input_fingerprint": "c69418fbec4ffbdb"
 },
 "S3sh6": {
  "input_fingerprint": "d0f66e0a5a522aa4",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 손바닥 위의 회수권 뭉치를 꼭 쥔 채 곁에 선 기사 최씨 쪽으로 고개를 돌린 자세로 시선을 맞춘 정임.\n\nLOCATION (lock): Inside the front passenger area of the first intercity bus, beside the seat behind the driver and close to the front entry. The bus windows are dark at night. 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- FRAMING SCALE: medium shot\n- KEY BACKGROUND ELEMENTS: 회수권 뭉치 (정임의 한 손에 꼭 쥐어져 있음) — The folded cover and compressed ticket edges angle toward the camera beneath 정임's face; used as Lower-center evidence retained within the reaction framing; 낡은 비닐 좌석 커버 (1호차 좌석 등받이에 그대로 씌워져 있음) — The seat beside 기사 최씨 presents its aisle-side plane, supporting his resting hand; used as Shared physical anchor between 기사 최씨 and 정임; 차창 (창밖이 이미 깜깜함) — The window presents a dark exterior beyond the bus interior; used as Background field that isolates the two-character exchange; 고무 매트가 깔린 통로 (1호차 통로에 그대로 깔려 있음); used as Lower background continuity beneath the two characters.\nCompose the frame exactly as specified above — subject scale and screen placement. The viewing angle is deliberately left open here; choose the angle that serves the shot. 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: Low-key ambient light appropriate to the nighttime interior is set against the explicitly dark windows, with muted navy, aged beige, restrained highlights, and subtle film grain.\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 정임 — 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\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): Bus 1 still has its worn vinyl seat covers, rubber aisle mat, and farebox chained in its original position beneath the gauges; the exterior beyond the windows remains dark. The rubber-banded ticket bundle remains out from under the seat with its worn folded cover intact. 정임: She remains in her navy conductor's uniform and billed cap, gripping the ticket bundle in one hand and turning her head sideways with her gaze raised. 기사 최씨: He wears his work jumper and work gloves, standing beside a seat near the front door with one hand resting on the seatback.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리); 기사 최씨 (한국인 외모, 50대 남성, 나이대에 맞는 얼굴, 짧고 단정한 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
  "roll_prompts": {
   "A": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThe camera is positioned in the central aisle, facing directly forward to capture a medium shot of the dimly lit bus interior. On the near left side of the frame, 정임 sits in the aisle-adjacent Seat 1B, turning her head rightward across the aisle. In the lower-center of the view, she tightly grips a folded, rubber-banded bundle of bus tickets. On the near right side of the frame, 기사 최씨 stands in the Front Entry Area, facing left toward 정임 while resting one hand on a worn seatback. The rubber-matted aisle extends straight up the center of the screen from near to far, dividing the two characters. In the far right background, the Front Passenger Seat is visible bearing a bell control. The side windows on the far left and right reveal a pitch-black exterior, isolating the subjects within the aged interior.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 손바닥 위의 회수권 뭉치를 꼭 쥔 채 곁에 선 기사 최씨 쪽으로 고개를 돌린 자세로 시선을 맞춘 정임.\n\nLOCATION (lock): Inside the front passenger area of the first intercity bus, beside the seat behind the driver and close to the front entry. The bus windows are dark at night. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Low-key ambient light appropriate to the nighttime interior is set against the explicitly dark windows, with muted navy, aged beige, restrained highlights, and subtle film grain.\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): Bus 1 still has its worn vinyl seat covers, rubber aisle mat, and farebox chained in its original position beneath the gauges; the exterior beyond the windows remains dark. The rubber-banded ticket bundle remains out from under the seat with its worn folded cover intact. 정임: She remains in her navy conductor's uniform and billed cap, gripping the ticket bundle in one hand and turning her head sideways with her gaze raised. 기사 최씨: He wears his work jumper and work gloves, standing beside a seat near the front door with one hand resting on the seatback.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리); 기사 최씨 (한국인 외모, 50대 남성, 나이대에 맞는 얼굴, 짧고 단정한 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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.",
   "B": "Create ONE 16:9 photorealistic film still for the brief below.\n\nThe FIRST attached image is a top-down FLOOR PLAN of this interior and\nthe SCENE LAYOUT text below is what a careful reader saw in it.\nTogether they are the ONLY authority for physical arrangement: which\nseat/station each person occupies, which station every primary control\nbelongs to, where any mirror/reflective surface sits and what it can\nphysically reflect, where the camera stands and what appears on which\nside of the screen. If any other sentence seems to contradict them, the\nfloor plan wins. The floor plan is a diagram, not scenery — none of its\nlines, arrows or labels may appear in the photograph. WHO the people\nare and what they do comes from the SHOT TEXT and the attached\nCHARACTER/PROP references — never add a person the SHOT TEXT does not\nplace here. No text, no watermarks.\n\nSCENE LAYOUT (what a careful reader saw in the attached floor plan):\nThe camera is positioned in the central aisle, facing directly forward to capture a medium shot of the dimly lit bus interior. On the near left side of the frame, 정임 sits in the aisle-adjacent Seat 1B, turning her head rightward across the aisle. In the lower-center of the view, she tightly grips a folded, rubber-banded bundle of bus tickets. On the near right side of the frame, 기사 최씨 stands in the Front Entry Area, facing left toward 정임 while resting one hand on a worn seatback. The rubber-matted aisle extends straight up the center of the screen from near to far, dividing the two characters. In the far right background, the Front Passenger Seat is visible bearing a bell control. The side windows on the far left and right reveal a pitch-black exterior, isolating the subjects within the aged interior.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — late-20th-century South Korea; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 손바닥 위의 회수권 뭉치를 꼭 쥔 채 곁에 선 기사 최씨 쪽으로 고개를 돌린 자세로 시선을 맞춘 정임.\n\nLOCATION (lock): Inside the front passenger area of the first intercity bus, beside the seat behind the driver and close to the front entry. The bus windows are dark at night. The shot takes place here.\n\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Low-key ambient light appropriate to the nighttime interior is set against the explicitly dark windows, with muted navy, aged beige, restrained highlights, and subtle film grain.\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): Bus 1 still has its worn vinyl seat covers, rubber aisle mat, and farebox chained in its original position beneath the gauges; the exterior beyond the windows remains dark. The rubber-banded ticket bundle remains out from under the seat with its worn folded cover intact. 정임: She remains in her navy conductor's uniform and billed cap, gripping the ticket bundle in one hand and turning her head sideways with her gaze raised. 기사 최씨: He wears his work jumper and work gloves, standing beside a seat near the front door with one hand resting on the seatback.\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 정임 right now, so 정임'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 정임: 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: 정임 (한국인 외모, 20대 여성, 자연스러운 얼굴형, 단정한 중간 길이의 검은 머리); 기사 최씨 (한국인 외모, 50대 남성, 나이대에 맞는 얼굴, 짧고 단정한 검은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\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\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": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S3sh6_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "기사 최씨",
     "path": "<bytes:952050>",
     "asset_id": "e0d2ec85-1698-4efc-ab24-a73e30964534",
     "role": "character_ref"
    },
    {
     "label": "고무줄로 묶인 낡은 종이 회수권 뭉치",
     "path": "<bytes:1135067>",
     "asset_id": "6e82ce93-8876-47c1-ac1d-48cf6c9a06d7",
     "role": "prop_ref"
    }
   ],
   "B": [
    {
     "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S3sh6_confinedfp.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
     "asset_id": null,
     "role": null
    },
    {
     "label": "정임",
     "path": "<bytes:812895>",
     "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
     "role": "character_ref"
    },
    {
     "label": "기사 최씨",
     "path": "<bytes:952050>",
     "asset_id": "e0d2ec85-1698-4efc-ab24-a73e30964534",
     "role": "character_ref"
    },
    {
     "label": "고무줄로 묶인 낡은 종이 회수권 뭉치",
     "path": "<bytes:1135067>",
     "asset_id": "6e82ce93-8876-47c1-ac1d-48cf6c9a06d7",
     "role": "prop_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "initial_roll_all_fail": true,
  "readings": [
   {
    "label": "A",
    "direction": "정임의 시선은 최씨를 향하고 있으나, 최씨는 정임을 보지 않고 허공을 응시하고 있어 '시선을 맞춘' 상태를 구현하지 못했습니다.",
    "built_space": "카메라가 버스 전면을 향하고 있으나, 엔진룸 왼쪽 공간에 있어야 할 운전석과 운전대가 완전히 누락된 채 빈 벽으로 처리되어 버스 내부 구조를 왜곡했습니다.",
    "entities": "정임은 유니폼을 입었으나 들고 있는 물건이 낡은 종이 회수권 뭉치가 아닌 단단한 직육면체 블록처럼 보입니다. 최씨는 레퍼런스의 노란색 점퍼가 아닌 회갈색 재킷을 입었습니다.",
    "hard_violations": [
     "[gemini-pro] 해부학적 오류: 최씨의 왼팔 끝에 오른손이 달려 있음.",
     "[gemini-pro] 구조 누락: 버스 전면을 비추고 있음에도 좌측의 운전석과 운전대가 완전히 누락됨."
    ],
    "physics": "최씨가 서서 두 손을 각각 다른 의자 등받이에 얹고 있으나, 왼팔 끝에 달린 손의 엄지손가락 방향이 잘못되어 물리적으로 불가능한 해부학적 구조(두 개의 오른손)를 띠고 있습니다."
   },
   {
    "label": "B",
    "direction": "정임과 최씨의 시선이 서로를 향해 정확히 교차하며 프롬프트가 요구한 시선 맞춤이 제대로 연출되었습니다.",
    "built_space": "카메라가 버스 뒤쪽을 향한 구도이지만, 도면 상 왼쪽에 있어야 할 정임이 우측에 있고, 우측에 있어야 할 출입구와 요금통이 좌측에 있는 등 버스의 좌우 구조가 완전히 반전되었습니다.",
    "entities": "정임이 쥐고 있는 회수권 뭉치는 종이의 질감, 접힌 모양, 고무줄까지 레퍼런스를 완벽히 재현했습니다. 다만 최씨가 노란색 점퍼 대신 짙은 갈색 스웨터를 입고 있습니다.",
    "hard_violations": [
     "[gemini-pro] 잘못된 위치 배치: 도면과 달리 정임이 버스의 우측에, 요금통이 좌측에 배치되어 버스 내부 구조가 좌우 반전됨.",
     "[gemini-pro] 금지된 텍스트 노출: 정임의 가슴 명찰에 읽을 수 있는(좌우 반전된) 뚜렷한 글자가 나타남."
    ],
    "physics": "정임은 무릎 위에 손을 모아 회수권을 자연스럽게 쥐고 있으며, 최씨는 한 손으로 기둥을 잡고 다른 손은 의자 등받이에 자연스럽게 기댄 채 지탱하고 있어 물리적 오류가 없습니다."
   }
  ],
  "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": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 4,
        "verdict_ko": "도면과 반대로 버스의 좌우 배치가 뒤집혀 있고 명찰에 글자가 노출되는 치명적 오류가 있으나, 종이 회수권 뭉치의 디테일과 두 인물의 시선 교환, 조명 분위기는 프롬프트를 잘 따랐습니다."
       },
       {
        "label": "A",
        "score": 2,
        "verdict_ko": "최씨의 왼팔에 오른손이 달려있는 해부학적 오류가 있고, 버스 앞부분에 필수적인 운전석이 누락되었으며, 회수권이 딱딱한 블록처럼 묘사되어 프롬프트 반영에 크게 실패했습니다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임의 시선은 최씨를 향하고 있으나, 최씨는 정임을 보지 않고 허공을 응시하고 있어 '시선을 맞춘' 상태를 구현하지 못했습니다.",
        "built_space": "카메라가 버스 전면을 향하고 있으나, 엔진룸 왼쪽 공간에 있어야 할 운전석과 운전대가 완전히 누락된 채 빈 벽으로 처리되어 버스 내부 구조를 왜곡했습니다.",
        "entities": "정임은 유니폼을 입었으나 들고 있는 물건이 낡은 종이 회수권 뭉치가 아닌 단단한 직육면체 블록처럼 보입니다. 최씨는 레퍼런스의 노란색 점퍼가 아닌 회갈색 재킷을 입었습니다.",
        "hard_violations": [
         "해부학적 오류: 최씨의 왼팔 끝에 오른손이 달려 있음.",
         "구조 누락: 버스 전면을 비추고 있음에도 좌측의 운전석과 운전대가 완전히 누락됨."
        ],
        "physics": "최씨가 서서 두 손을 각각 다른 의자 등받이에 얹고 있으나, 왼팔 끝에 달린 손의 엄지손가락 방향이 잘못되어 물리적으로 불가능한 해부학적 구조(두 개의 오른손)를 띠고 있습니다."
       },
       {
        "label": "B",
        "direction": "정임과 최씨의 시선이 서로를 향해 정확히 교차하며 프롬프트가 요구한 시선 맞춤이 제대로 연출되었습니다.",
        "built_space": "카메라가 버스 뒤쪽을 향한 구도이지만, 도면 상 왼쪽에 있어야 할 정임이 우측에 있고, 우측에 있어야 할 출입구와 요금통이 좌측에 있는 등 버스의 좌우 구조가 완전히 반전되었습니다.",
        "entities": "정임이 쥐고 있는 회수권 뭉치는 종이의 질감, 접힌 모양, 고무줄까지 레퍼런스를 완벽히 재현했습니다. 다만 최씨가 노란색 점퍼 대신 짙은 갈색 스웨터를 입고 있습니다.",
        "hard_violations": [
         "잘못된 위치 배치: 도면과 달리 정임이 버스의 우측에, 요금통이 좌측에 배치되어 버스 내부 구조가 좌우 반전됨.",
         "금지된 텍스트 노출: 정임의 가슴 명찰에 읽을 수 있는(좌우 반전된) 뚜렷한 글자가 나타남."
        ],
        "physics": "정임은 무릎 위에 손을 모아 회수권을 자연스럽게 쥐고 있으며, 최씨는 한 손으로 기둥을 잡고 다른 손은 의자 등받이에 자연스럽게 기댄 채 지탱하고 있어 물리적 오류가 없습니다."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 4,
        "verdict_ko": "도면과 반대로 버스의 좌우 배치가 뒤집혀 있고 명찰에 글자가 노출되는 치명적 오류가 있으나, 종이 회수권 뭉치의 디테일과 두 인물의 시선 교환, 조명 분위기는 프롬프트를 잘 따랐습니다."
       },
       {
        "label": "A",
        "score": 2,
        "verdict_ko": "최씨의 왼팔에 오른손이 달려있는 해부학적 오류가 있고, 버스 앞부분에 필수적인 운전석이 누락되었으며, 회수권이 딱딱한 블록처럼 묘사되어 프롬프트 반영에 크게 실패했습니다."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "A",
        "direction": "정임의 시선은 최씨를 향하고 있으나, 최씨는 정임을 보지 않고 허공을 응시하고 있어 '시선을 맞춘' 상태를 구현하지 못했습니다.",
        "built_space": "카메라가 버스 전면을 향하고 있으나, 엔진룸 왼쪽 공간에 있어야 할 운전석과 운전대가 완전히 누락된 채 빈 벽으로 처리되어 버스 내부 구조를 왜곡했습니다.",
        "entities": "정임은 유니폼을 입었으나 들고 있는 물건이 낡은 종이 회수권 뭉치가 아닌 단단한 직육면체 블록처럼 보입니다. 최씨는 레퍼런스의 노란색 점퍼가 아닌 회갈색 재킷을 입었습니다.",
        "hard_violations": [
         "해부학적 오류: 최씨의 왼팔 끝에 오른손이 달려 있음.",
         "구조 누락: 버스 전면을 비추고 있음에도 좌측의 운전석과 운전대가 완전히 누락됨."
        ],
        "physics": "최씨가 서서 두 손을 각각 다른 의자 등받이에 얹고 있으나, 왼팔 끝에 달린 손의 엄지손가락 방향이 잘못되어 물리적으로 불가능한 해부학적 구조(두 개의 오른손)를 띠고 있습니다."
       },
       {
        "label": "B",
        "direction": "정임과 최씨의 시선이 서로를 향해 정확히 교차하며 프롬프트가 요구한 시선 맞춤이 제대로 연출되었습니다.",
        "built_space": "카메라가 버스 뒤쪽을 향한 구도이지만, 도면 상 왼쪽에 있어야 할 정임이 우측에 있고, 우측에 있어야 할 출입구와 요금통이 좌측에 있는 등 버스의 좌우 구조가 완전히 반전되었습니다.",
        "entities": "정임이 쥐고 있는 회수권 뭉치는 종이의 질감, 접힌 모양, 고무줄까지 레퍼런스를 완벽히 재현했습니다. 다만 최씨가 노란색 점퍼 대신 짙은 갈색 스웨터를 입고 있습니다.",
        "hard_violations": [
         "잘못된 위치 배치: 도면과 달리 정임이 버스의 우측에, 요금통이 좌측에 배치되어 버스 내부 구조가 좌우 반전됨.",
         "금지된 텍스트 노출: 정임의 가슴 명찰에 읽을 수 있는(좌우 반전된) 뚜렷한 글자가 나타남."
        ],
        "physics": "정임은 무릎 위에 손을 모아 회수권을 자연스럽게 쥐고 있으며, 최씨는 한 손으로 기둥을 잡고 다른 손은 의자 등받이에 자연스럽게 기댄 채 지탱하고 있어 물리적 오류가 없습니다."
       }
      ]
     },
     "ok": true
    },
    {
     "model": "gpt",
     "order": "reverse",
     "display_to_canonical": {
      "A": "B",
      "B": "A"
     },
     "raw": {
      "winner": "B",
      "ranking": [
       "B",
       "A"
      ],
      "verdicts": [
       {
        "label": "B",
        "score": 7,
        "verdict_ko": "정임과 최씨의 시선 교환, 전방 승객 구역의 배치, 고무줄로 묶인 낡은 회수권은 가장 충실하지만 정임이 한 손이 아니라 두 손으로 뭉치를 잡고 있으며 잠긴 작업복 형태도 다르다."
       },
       {
        "label": "A",
        "score": 6,
        "verdict_ko": "두 인물의 시선과 최씨의 좌석 등받이 접촉은 맞지만, 전방 구조가 덜 명확하고 회수권이 고무줄과 접힌 표지 없이 풀린 종이 묶음처럼 보이며 역시 두 손으로 들고 있다."
       }
      ],
      "readings": [
       {
        "label": "A",
        "direction": "정임은 화면 오른쪽의 기사 최씨 얼굴을 올려다보고, 최씨는 화면 왼쪽 아래의 정임 얼굴을 내려다봐 서로 시선이 정확히 만난다. 무기나 별도의 조준 물체는 없다.",
        "built_space": "카메라는 버스 전방에서 뒤쪽을 바라보는 듯하며, 양쪽의 어두운 차창, 중앙의 고무 통로 매트, 여러 줄의 낡은 비닐 좌석, 중앙 지지봉 1개와 오른쪽의 사슬 달린 요금통 1개가 보인다. 정임은 운전석 뒤쪽에 해당할 수 있는 왼쪽 좌석에 앉아 있고 최씨는 오른쪽 전방 좌석 옆에 서서 한 손을 그 등받이에 얹고 다른 손은 지지봉을 잡는다. 창에는 불가능한 인물 반사가 없다.",
        "entities": "정임은 한국인 외모의 20대 여성으로 보이며 중간 길이 검은 머리, 남색 모자와 참고 이미지에 가까운 남색 작업복을 착용한다. 최씨는 한국인 외모의 50대 남성, 짧은 검은 머리와 회색 작업 점퍼 및 작업 장갑으로 보인다. 정임 손의 물체는 낡은 종이표 여러 장이지만 참고물의 뚜렷한 고무줄, 압축된 표 가장자리와 접힌 표지가 충분히 재현되지 않았다. 두 사람 외 인물은 없다.",
        "hard_violations": [],
        "physics": "정임의 엉덩이는 좌석에 놓이고 몸은 자연스럽게 앉아 지지된다. 최씨는 하체가 프레임 아래로 잘렸지만 서 있는 자세이며, 오른손은 좌석 등받이, 왼손은 봉에 실제로 접촉한다. 회수권은 정임의 두 손에 받쳐져 떠 있지 않지만, 지시된 한 손의 강한 움켜쥠은 아니다."
       },
       {
        "label": "B",
        "direction": "정임은 고개를 화면 오른쪽으로 돌려 기사 최씨의 얼굴을 올려다보고, 최씨도 정임 얼굴을 내려다봐 시선이 정확히 서로에게 닿는다. 다른 조준 또는 이동 물체는 없다.",
        "built_space": "카메라는 통로에서 버스 전방을 바라본다. 전면의 어두운 차창 1면, 오른쪽 전면 출입문, 중앙 전방 덮개와 계기부, 좌우의 낡은 좌석들, 고무 통로 매트 1개, 세로봉 여러 개와 손잡이가 보인다. 정임은 운전석 뒤편에 해당하는 왼쪽 첫 좌석에 앉고, 최씨는 오른쪽 전면 출입구 가까이 서서 앞쪽 좌석 등받이에 장갑 낀 손을 얹는다. 요금통의 사슬은 보이지 않지만 전방 구조와 인물 배치는 평면도 및 시대 참고와 잘 맞으며, 차창의 어두운 반사도 광학적으로 모순되지 않는다.",
        "entities": "정임은 한국인 외모의 20대 여성, 단정한 검은 머리와 남색 모자를 갖췄으나 참고 이미지의 남색 작업용 커버올 대신 정장형 승무원 제복을 입었다. 최씨는 한국인 외모의 50대 남성, 짧은 검은 머리, 갈색 작업 점퍼와 장갑을 착용한다. 회수권은 낡은 접힌 종이표가 고무줄로 압축된 뭉치여서 참고 소품과 명확히 일치한다. 두 사람 외 인물은 없다.",
        "hard_violations": [],
        "physics": "정임은 좌석에 정상적으로 앉아 좌판과 등받이의 지지를 받는다. 최씨의 발은 프레임 밖이지만 몸은 수직으로 서 있고 양손이 각각 좌석 등받이에 닿아 있어 부유하지 않는다. 회수권은 정임의 두 손이 확실히 지지하지만, 프롬프트의 한 손으로 꼭 쥔 상태와는 다르다."
       }
      ],
      "all_candidates_fail": true
     },
     "normalized": {
      "winner": "A",
      "ranking": [
       "A",
       "B"
      ],
      "verdicts": [
       {
        "label": "A",
        "score": 7,
        "verdict_ko": "정임과 최씨의 시선 교환, 전방 승객 구역의 배치, 고무줄로 묶인 낡은 회수권은 가장 충실하지만 정임이 한 손이 아니라 두 손으로 뭉치를 잡고 있으며 잠긴 작업복 형태도 다르다."
       },
       {
        "label": "B",
        "score": 6,
        "verdict_ko": "두 인물의 시선과 최씨의 좌석 등받이 접촉은 맞지만, 전방 구조가 덜 명확하고 회수권이 고무줄과 접힌 표지 없이 풀린 종이 묶음처럼 보이며 역시 두 손으로 들고 있다."
       }
      ],
      "all_candidates_fail": true,
      "readings": [
       {
        "label": "B",
        "direction": "정임은 화면 오른쪽의 기사 최씨 얼굴을 올려다보고, 최씨는 화면 왼쪽 아래의 정임 얼굴을 내려다봐 서로 시선이 정확히 만난다. 무기나 별도의 조준 물체는 없다.",
        "built_space": "카메라는 버스 전방에서 뒤쪽을 바라보는 듯하며, 양쪽의 어두운 차창, 중앙의 고무 통로 매트, 여러 줄의 낡은 비닐 좌석, 중앙 지지봉 1개와 오른쪽의 사슬 달린 요금통 1개가 보인다. 정임은 운전석 뒤쪽에 해당할 수 있는 왼쪽 좌석에 앉아 있고 최씨는 오른쪽 전방 좌석 옆에 서서 한 손을 그 등받이에 얹고 다른 손은 지지봉을 잡는다. 창에는 불가능한 인물 반사가 없다.",
        "entities": "정임은 한국인 외모의 20대 여성으로 보이며 중간 길이 검은 머리, 남색 모자와 참고 이미지에 가까운 남색 작업복을 착용한다. 최씨는 한국인 외모의 50대 남성, 짧은 검은 머리와 회색 작업 점퍼 및 작업 장갑으로 보인다. 정임 손의 물체는 낡은 종이표 여러 장이지만 참고물의 뚜렷한 고무줄, 압축된 표 가장자리와 접힌 표지가 충분히 재현되지 않았다. 두 사람 외 인물은 없다.",
        "hard_violations": [],
        "physics": "정임의 엉덩이는 좌석에 놓이고 몸은 자연스럽게 앉아 지지된다. 최씨는 하체가 프레임 아래로 잘렸지만 서 있는 자세이며, 오른손은 좌석 등받이, 왼손은 봉에 실제로 접촉한다. 회수권은 정임의 두 손에 받쳐져 떠 있지 않지만, 지시된 한 손의 강한 움켜쥠은 아니다."
       },
       {
        "label": "A",
        "direction": "정임은 고개를 화면 오른쪽으로 돌려 기사 최씨의 얼굴을 올려다보고, 최씨도 정임 얼굴을 내려다봐 시선이 정확히 서로에게 닿는다. 다른 조준 또는 이동 물체는 없다.",
        "built_space": "카메라는 통로에서 버스 전방을 바라본다. 전면의 어두운 차창 1면, 오른쪽 전면 출입문, 중앙 전방 덮개와 계기부, 좌우의 낡은 좌석들, 고무 통로 매트 1개, 세로봉 여러 개와 손잡이가 보인다. 정임은 운전석 뒤편에 해당하는 왼쪽 첫 좌석에 앉고, 최씨는 오른쪽 전면 출입구 가까이 서서 앞쪽 좌석 등받이에 장갑 낀 손을 얹는다. 요금통의 사슬은 보이지 않지만 전방 구조와 인물 배치는 평면도 및 시대 참고와 잘 맞으며, 차창의 어두운 반사도 광학적으로 모순되지 않는다.",
        "entities": "정임은 한국인 외모의 20대 여성, 단정한 검은 머리와 남색 모자를 갖췄으나 참고 이미지의 남색 작업용 커버올 대신 정장형 승무원 제복을 입었다. 최씨는 한국인 외모의 50대 남성, 짧은 검은 머리, 갈색 작업 점퍼와 장갑을 착용한다. 회수권은 낡은 접힌 종이표가 고무줄로 압축된 뭉치여서 참고 소품과 명확히 일치한다. 두 사람 외 인물은 없다.",
        "hard_violations": [],
        "physics": "정임은 좌석에 정상적으로 앉아 좌판과 등받이의 지지를 받는다. 최씨의 발은 프레임 밖이지만 몸은 수직으로 서 있고 양손이 각각 좌석 등받이에 닿아 있어 부유하지 않는다. 회수권은 정임의 두 손이 확실히 지지하지만, 프롬프트의 한 손으로 꼭 쥔 상태와는 다르다."
       }
      ]
     },
     "ok": true
    }
   ],
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "slot_winner_match": false,
   "slot_winner": {
    "gemini-pro": "B",
    "gpt": "A"
   },
   "route": "cross_slot_combined"
  },
  "dual": {
   "models": [
    "gemini-pro",
    "gpt"
   ],
   "normalized": {
    "A": 1.5,
    "B": 1.857
   },
   "adjusted": {
    "A": 1.25,
    "B": 1.607
   },
   "violations": {
    "A": [
     "[gemini-pro] 해부학적 오류: 최씨의 왼팔 끝에 오른손이 달려 있음.",
     "[gemini-pro] 구조 누락: 버스 전면을 비추고 있음에도 좌측의 운전석과 운전대가 완전히 누락됨."
    ],
    "B": [
     "[gemini-pro] 잘못된 위치 배치: 도면과 달리 정임이 버스의 우측에, 요금통이 좌측에 배치되어 버스 내부 구조가 좌우 반전됨.",
     "[gemini-pro] 금지된 텍스트 노출: 정임의 가슴 명찰에 읽을 수 있는(좌우 반전된) 뚜렷한 글자가 나타남."
    ]
   },
   "per_model_winner": {
    "gemini-pro": "B",
    "gpt": "A"
   },
   "agreed": false
  },
  "totals": {
   "B": 1607,
   "A": 1250
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 1607,
    "verdict_ko": "도면과 반대로 버스의 좌우 배치가 뒤집혀 있고 명찰에 글자가 노출되는 치명적 오류가 있으나, 종이 회수권 뭉치의 디테일과 두 인물의 시선 교환, 조명 분위기는 프롬프트를 잘 따랐습니다.  ★위반: [gemini-pro] 잘못된 위치 배치: 도면과 달리 정임이 버스의 우측에, 요금통이 좌측에 배치되어 버스 내부 구조가 좌우 반전됨. / [gemini-pro] 금지된 텍스트 노출: 정임의 가슴 명찰에 읽을 수 있는(좌우 반전된) 뚜렷한 글자가 나타남."
   },
   {
    "label": "A",
    "score": 1250,
    "verdict_ko": "최씨의 왼팔에 오른손이 달려있는 해부학적 오류가 있고, 버스 앞부분에 필수적인 운전석이 누락되었으며, 회수권이 딱딱한 블록처럼 묘사되어 프롬프트 반영에 크게 실패했습니다.  ★위반: [gemini-pro] 해부학적 오류: 최씨의 왼팔 끝에 오른손이 달려 있음. / [gemini-pro] 구조 누락: 버스 전면을 비추고 있음에도 좌측의 운전석과 운전대가 완전히 누락됨."
   }
  ],
  "refs": [
   {
    "label": "FLOOR PLAN — layout authority, a diagram, never scenery",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/S3sh6_confinedfp.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "PERIOD REFERENCE — 1980년대 한국 시외버스 내부 및 요금통: a researched photograph of the real\nthing. Copy its era-accurate form, proportions, materials, fittings\nand styling exactly. It decides ONLY what this thing looks like —\nnever the composition, framing, lighting, people or their placement.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/da049582-2c6d-492c-979d-f468d61bab6e/images/97375a4b-b465-4fd6-9a2d-800e0eb01685/scene/recipe/eraref_8f484ee77a47c9d6.png",
    "asset_id": null,
    "role": null
   },
   {
    "label": "정임",
    "path": "<bytes:812895>",
    "asset_id": "08a49c6d-61a7-4e2f-a8ba-1559d0a1a0b8",
    "role": "character_ref"
   },
   {
    "label": "기사 최씨",
    "path": "<bytes:952050>",
    "asset_id": "e0d2ec85-1698-4efc-ab24-a73e30964534",
    "role": "character_ref"
   },
   {
    "label": "고무줄로 묶인 낡은 종이 회수권 뭉치",
    "path": "<bytes:1135067>",
    "asset_id": "6e82ce93-8876-47c1-ac1d-48cf6c9a06d7",
    "role": "prop_ref"
   }
  ],
  "critique_skipped": true,
  "needs_reshoot": true,
  "shot_run_produced": true,
  "shot_run_uid": "06a92acc-2e31-7e4b-88c4-ba759742a708",
  "confined_fp": {
   "base_key": "confinedfp::0586889cbde5",
   "apt_reason": "버스 내부의 운전석, 앞문, 승객석 등 구체적인 공간 배치가 명확해야 하며, 두 인물의 위치와 시선 방향이 중요하게 작용하는 장면이므로 평면도 보조가 필요합니다.",
   "fixed": true,
   "mismatches": [],
   "era_research": {
    "subject": "1980년대 한국 시외버스 내부 및 요금통",
    "queries": [
     [
      "1980년대 한국 시외버스 내부 좌석 운전석",
      "옛날 한국 버스 요금통 1980년대 시내버스 내부"
     ]
    ],
    "picked_url": "https://file1.bobaedream.co.kr/multi_image/truck/2020/01/15/19/CQA5e1eee56bb76d.jpg",
    "sha256": "587de5353dace6b5a009871a0c0885a7670336392d63862d50b3f47e7d8fe625",
    "file": "eraref_8f484ee77a47c9d6.png"
   }
  },
  "ref_mode": "confined_fp: 도면+장면설명+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S3sh4"
  }
 },
 "S3sh6::cine": {
  "applied": true,
  "attempted_at": "2026-08-29T10:03:28.174839+00:00",
  "fingerprint": "b54442875f3c35221fda821bc8f2d2c23d090968917d1d5f6c287abf8fc4de01",
  "fingerprint_version": 2,
  "provider": "reve",
  "endpoint": "reve/2.1/edit",
  "model": "reve/2.1/edit",
  "pack": "24.202608252115",
  "source_file": "S3sh6_sel.png",
  "source_sha256": "718e66921d86c465dee9ea104d8497dd999a3cf9b9db9e17a727a1a07ce52660",
  "file": "S3sh6_cine.png",
  "staged_sha256": "d5f127f88cd1616867463c304148564a02eb09a833d2c3891252db4d61a7cc66",
  "latency_ms": 92703,
  "request_id": "01a04cf9-6ed5-7e41-a0e3-3b1a0c445a66",
  "status_url": "https://queue.fal.run/reve/2.1/requests/01a04cf9-6ed5-7e41-a0e3-3b1a0c445a66/status",
  "response_url": "https://queue.fal.run/reve/2.1/requests/01a04cf9-6ed5-7e41-a0e3-3b1a0c445a66",
  "submitted_at": "2026-08-29T10:03:30.467670+00:00"
 }
}