{
 "S1sh4::signage": {
  "fp": "fb8ebc518eeffb84",
  "inscriptions": []
 },
 "era_assess::2bc3cb6892c8590e": {
  "subjects": [
   {
    "subject_native": "한국 고등학교 교실 내부 (대한민국)",
    "search_terms_native": [
     "고등학교 교실 내부",
     "한국 학교 교실",
     "고등학교 교실 창문"
    ],
    "language_lock_native": "모든 검색어는 반드시 한국어로만 작성되어야 하며, 다른 언어로 번역하거나 추가해서는 안 됩니다.",
    "reason_ko": "한국의 고등학교 교실은 슬라이딩 도어, 독특한 창문 배치, 책걸상 구조 등 서구권 교실과 구별되는 명확한 지역적 레이아웃과 특징을 가지고 있습니다."
   }
  ]
 },
 "era_ref::628d21fb9055b07e": {
  "subject": "한국 고등학교 교실 내부 (대한민국)",
  "terms": [
   "고등학교 교실 내부",
   "한국 학교 교실",
   "고등학교 교실 창문"
  ],
  "queries": [
   [
    "대한민국 고등학교 교실 내부 창문",
    "한국 학교 고등학교 교실 창가"
   ]
  ],
  "candidates": 4,
  "picked_index": 3,
  "picked_url": "https://news.nateimg.co.kr/orgImg/ns/2025/11/16/NISI20250731_0001907820_web.jpg",
  "picked_reason_ko": "3번은 관광·전시용 꾸밈 없이 실제로 사용하는 한국 고등학교 교실의 현대적이고 일상적인 구성과 재료를 가장 완전하고 선명하게 보여 준다.",
  "sha256": "c73d2d504515288074fba5529a68f252a63eed4785955501f41fed8fc2048464",
  "file": "eraref_628d21fb9055b07e.png"
 },
 "S1sh4::bgfirst_bg": {
  "input_fingerprint": "2c42e80c102a1119",
  "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): 창밖을 향해 고개를 돌린 현석의 눈동자에 희미한 생기가 도는 측면 얼굴.\n\nLOCATION (lock): Inside a high-school classroom, beside the window overlooking the sky. White LED classroom lighting fills the room.\n\nTIME OF DAY (lock): morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Hyunseok's seated eye height, the tracking camera briefly tightens from just behind his shoulder into a clean profile close-up, with his face held in the left half and generous look space opening toward the classroom window on the right. His head is caught midway through the turn, and shallow focus holds on the faint vitality returning to his eyes as he watches the birds outside.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: classroom window (visible beside Hyunseok; opening state unspecified) — The camera looks toward the window area with the flying flock visible outside beyond it; used as Provides right-side look space and the destination of Hyunseok's turning gaze; rows of seated students (wearing transparent AR glasses and watching the same government promotional video) — Only soft, partial figures remain behind Hyunseok along the classroom rows; used as Recedes out of focus to contrast collective uniformity with Hyunseok's private response.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: White LED classroom light is kept subdued and cool, with controlled contrast across Hyunseok's profile.\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. Signage and other writing that permanently belongs to this place may appear, in the place's native language and period style; no captions, watermarks or overlay text.\n\nPERIOD REFERENCE — 한국 고등학교 교실 내부 (대한민국): 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): 창밖을 향해 고개를 돌린 현석의 눈동자에 희미한 생기가 도는 측면 얼굴.\n\nLOCATION (lock): Inside a high-school classroom, beside the window overlooking the sky. White LED classroom lighting fills the room.\n\nTIME OF DAY (lock): morning.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Hyunseok's seated eye height, the tracking camera briefly tightens from just behind his shoulder into a clean profile close-up, with his face held in the left half and generous look space opening toward the classroom window on the right. His head is caught midway through the turn, and shallow focus holds on the faint vitality returning to his eyes as he watches the birds outside.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: classroom window (visible beside Hyunseok; opening state unspecified) — The camera looks toward the window area with the flying flock visible outside beyond it; used as Provides right-side look space and the destination of Hyunseok's turning gaze; rows of seated students (wearing transparent AR glasses and watching the same government promotional video) — Only soft, partial figures remain behind Hyunseok along the classroom rows; used as Recedes out of focus to contrast collective uniformity with Hyunseok's private response.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: White LED classroom light is kept subdued and cool, with controlled contrast across Hyunseok's profile.\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. Signage and other writing that permanently belongs to this place may appear, in the place's native language and period style; no captions, watermarks or overlay text.\n\nPERIOD REFERENCE — 한국 고등학교 교실 내부 (대한민국): 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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S1sh4__bgfirst_bg.png",
  "asset_id": "cfac60af-05ee-434e-af2b-89b318579af4",
  "input_asset_ids": [
   "9f6d8dfe-d37d-443b-bb45-56e0d1940a11",
   "04872d30-3ac8-451e-9c07-fa7856fd97f9"
  ],
  "era_research": {
   "subject": "한국 고등학교 교실 내부 (대한민국)",
   "queries": [
    [
     "대한민국 고등학교 교실 내부 창문",
     "한국 학교 고등학교 교실 창가"
    ]
   ],
   "picked_url": "https://news.nateimg.co.kr/orgImg/ns/2025/11/16/NISI20250731_0001907820_web.jpg",
   "sha256": "c73d2d504515288074fba5529a68f252a63eed4785955501f41fed8fc2048464",
   "file": "eraref_628d21fb9055b07e.png"
  }
 },
 "S1sh4": {
  "input_fingerprint": "b99ecafaf2c9ee89",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): morning.\n\nSHOT TEXT (authoritative, Korean): 창밖을 향해 고개를 돌린 현석의 눈동자에 희미한 생기가 도는 측면 얼굴.\n\nLOCATION (lock): Inside a high-school classroom, beside the window overlooking the sky. White LED classroom lighting fills the room. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Hyunseok's seated eye height, the tracking camera briefly tightens from just behind his shoulder into a clean profile close-up, with his face held in the left half and generous look space opening toward the classroom window on the right. His head is caught midway through the turn, and shallow focus holds on the faint vitality returning to his eyes as he watches the birds outside.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: classroom window (visible beside Hyunseok; opening state unspecified) — The camera looks toward the window area with the flying flock visible outside beyond it; used as Provides right-side look space and the destination of Hyunseok's turning gaze; rows of seated students (wearing transparent AR glasses and watching the same government promotional video) — Only soft, partial figures remain behind Hyunseok along the classroom rows; used as Recedes out of focus to contrast collective uniformity with Hyunseok's private response.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: White LED classroom light is kept subdued and cool, with controlled contrast across Hyunseok's profile.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok is no longer wearing his transparent AR glasses; they remain discarded on his desk. Jeho remains seated in his own AR glasses with a writing instrument in his tense hand.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 현석 (한국인 남성, 10대 후반의 젊은 얼굴, 헝클어진 짧은 머리, 자연스러운 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\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): morning.\n\nSHOT TEXT (authoritative, Korean): 창밖을 향해 고개를 돌린 현석의 눈동자에 희미한 생기가 도는 측면 얼굴.\n\nLOCATION (lock): Inside a high-school classroom, beside the window overlooking the sky. White LED classroom lighting fills the room. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Hyunseok's seated eye height, the tracking camera briefly tightens from just behind his shoulder into a clean profile close-up, with his face held in the left half and generous look space opening toward the classroom window on the right. His head is caught midway through the turn, and shallow focus holds on the faint vitality returning to his eyes as he watches the birds outside.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: classroom window (visible beside Hyunseok; opening state unspecified) — The camera looks toward the window area with the flying flock visible outside beyond it; used as Provides right-side look space and the destination of Hyunseok's turning gaze; rows of seated students (wearing transparent AR glasses and watching the same government promotional video) — Only soft, partial figures remain behind Hyunseok along the classroom rows; used as Recedes out of focus to contrast collective uniformity with Hyunseok's private response.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: White LED classroom light is kept subdued and cool, with controlled contrast across Hyunseok's profile.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok is no longer wearing his transparent AR glasses; they remain discarded on his desk. Jeho remains seated in his own AR glasses with a writing instrument in his tense hand.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 현석 (한국인 남성, 10대 후반의 젊은 얼굴, 헝클어진 짧은 머리, 자연스러운 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\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): morning.\n\nSHOT TEXT (authoritative, Korean): 창밖을 향해 고개를 돌린 현석의 눈동자에 희미한 생기가 도는 측면 얼굴.\n\nLOCATION (lock): Inside a high-school classroom, beside the window overlooking the sky. White LED classroom lighting fills the room. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Hyunseok's seated eye height, the tracking camera briefly tightens from just behind his shoulder into a clean profile close-up, with his face held in the left half and generous look space opening toward the classroom window on the right. His head is caught midway through the turn, and shallow focus holds on the faint vitality returning to his eyes as he watches the birds outside.\n- FRAMING SCALE: close-up\n- KEY BACKGROUND ELEMENTS: classroom window (visible beside Hyunseok; opening state unspecified) — The camera looks toward the window area with the flying flock visible outside beyond it; used as Provides right-side look space and the destination of Hyunseok's turning gaze; rows of seated students (wearing transparent AR glasses and watching the same government promotional video) — Only soft, partial figures remain behind Hyunseok along the classroom rows; used as Recedes out of focus to contrast collective uniformity with Hyunseok's private response.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: White LED classroom light is kept subdued and cool, with controlled contrast across Hyunseok's profile.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok is no longer wearing his transparent AR glasses; they remain discarded on his desk. Jeho remains seated in his own AR glasses with a writing instrument in his tense hand.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 현석 (한국인 남성, 10대 후반의 젊은 얼굴, 헝클어진 짧은 머리, 자연스러운 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\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, but reframe like a film director choosing a different setup — a different angle, height, distance or foreground layer from the obvious one. For this candidate only, treat the CAMERA & FRAME contract above as a baseline to depart from, not a lock; the references and the shot text fix WHAT is seen, not WHERE the camera stands. 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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S1sh4__bgfirst_bg.png",
     "asset_id": "cfac60af-05ee-434e-af2b-89b318579af4",
     "role": "bgfirst_bg"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/conti/conti_S1sh4.png",
     "asset_id": "9f6d8dfe-d37d-443b-bb45-56e0d1940a11",
     "role": "conti_light"
    },
    {
     "label": "CHARACTER REFERENCE — 현석: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:973165>",
     "asset_id": "37299bdd-ae7a-4ecb-a36a-05c753ff9d51",
     "role": "character_ref"
    }
   ],
   "B": [
    {
     "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/46575226-4c2f-4334-b16a-429b11371e24/episodes/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/images/background_chain/L01B01.png",
     "asset_id": "04872d30-3ac8-451e-9c07-fa7856fd97f9",
     "role": "location_plate"
    },
    {
     "label": "CHARACTER REFERENCE — 현석: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:973165>",
     "asset_id": "37299bdd-ae7a-4ecb-a36a-05c753ff9d51",
     "role": "character_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "창가 바로 옆 좌석 배치, 클로즈업 구도, 책상 위 AR 안경, 뒤에서 필기하는 제호 등 프롬프트의 까다로운 세부 사항을 훌륭하게 구현했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "인물이 창가 옆이 아닌 교실 중앙에 배치되어 위치 제약을 어겼으며, 뒤에서 펜을 쥐고 있는 인물(제호)의 지정된 행동 묘사도 누락되었습니다."
     }
    ],
    "readings": [
     {
      "label": "A",
      "direction": "현석의 시선은 오른쪽 창밖 하늘에 있는 새 떼를 정확히 향함. 뒤쪽 남학생은 책상 위를 내려다보며 펜으로 글을 씀.",
      "built_space": "지정된 교실의 창가 바로 옆 줄에 책상이 위치하며, 카메라 시점과 창문의 공간적 배열이 프롬프트 및 기준 사진과 일치함.",
      "entities": "현석의 얼굴, 헤어스타일, 복장(조끼, 타이)이 기준과 일치함. 책상 위 벗어둔 안경, 뒤에서 AR 안경을 쓰고 필기하는 제호, 창밖 새 떼가 모두 묘사됨.",
      "hard_violations": [],
      "physics": "현석과 배경 인물들 모두 의자에 안정적으로 앉아 몸을 지지하고 있으며, 펜을 쥔 손과 책상의 접촉이 자연스러움."
     },
     {
      "label": "B",
      "direction": "현석은 오른쪽을 응시하나 창문이 멀어 시선의 목적지가 다소 불분명함. 배경의 학생들은 대부분 정면이나 모호한 곳을 응시함.",
      "built_space": "현석이 창가 자리가 아닌 교실 중앙 열에 앉아 있어, '창가 옆(beside the window)'이라는 위치 설정을 위반함.",
      "entities": "현석의 인물 특징과 책상 위 AR 안경은 존재하나, 프롬프트가 요구한 '펜을 쥐고 긴장한 채 글을 쓰는 제호'가 뒤쪽에 묘사되지 않음.",
      "hard_violations": [],
      "physics": "학생들이 의자에 앉아 책상에 팔을 기대고 있는 등 물리적 지지 상태는 정상적임."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "agree",
     "gap": 0.0,
     "per_model_winner": {
      "gemini-pro": "A",
      "openrouter:x-ai/grok-4.6": "A"
     },
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "창가 바로 옆 좌석 배치, 클로즈업 구도, 책상 위 AR 안경, 뒤에서 필기하는 제호 등 프롬프트의 까다로운 세부 사항을 훌륭하게 구현했습니다."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "인물이 창가 옆이 아닌 교실 중앙에 배치되어 위치 제약을 어겼으며, 뒤에서 펜을 쥐고 있는 인물(제호)의 지정된 행동 묘사도 누락되었습니다."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "A",
      "direction": "현석의 시선은 오른쪽 창밖 하늘에 있는 새 떼를 정확히 향함. 뒤쪽 남학생은 책상 위를 내려다보며 펜으로 글을 씀.",
      "built_space": "지정된 교실의 창가 바로 옆 줄에 책상이 위치하며, 카메라 시점과 창문의 공간적 배열이 프롬프트 및 기준 사진과 일치함.",
      "entities": "현석의 얼굴, 헤어스타일, 복장(조끼, 타이)이 기준과 일치함. 책상 위 벗어둔 안경, 뒤에서 AR 안경을 쓰고 필기하는 제호, 창밖 새 떼가 모두 묘사됨.",
      "hard_violations": [],
      "physics": "현석과 배경 인물들 모두 의자에 안정적으로 앉아 몸을 지지하고 있으며, 펜을 쥔 손과 책상의 접촉이 자연스러움."
     },
     {
      "label": "B",
      "direction": "현석은 오른쪽을 응시하나 창문이 멀어 시선의 목적지가 다소 불분명함. 배경의 학생들은 대부분 정면이나 모호한 곳을 응시함.",
      "built_space": "현석이 창가 자리가 아닌 교실 중앙 열에 앉아 있어, '창가 옆(beside the window)'이라는 위치 설정을 위반함.",
      "entities": "현석의 인물 특징과 책상 위 AR 안경은 존재하나, 프롬프트가 요구한 '펜을 쥐고 긴장한 채 글을 쓰는 제호'가 뒤쪽에 묘사되지 않음.",
      "hard_violations": [],
      "physics": "학생들이 의자에 앉아 책상에 팔을 기대고 있는 등 물리적 지지 상태는 정상적임."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 8,
      "verdict_ko": "요구된 완벽한 측면(clean profile) 구도와 화면 좌우 배분을 정확히 구현했으며, 배경 학생이 AR 안경을 쓰고 펜을 쥐고 있는 세부 지시까지 충실히 반영했습니다."
     },
     {
      "label": "A",
      "score": 5,
      "verdict_ko": "지시된 완벽한 측면이 아닌 반측면 구도이며, 배경 학생의 펜과 투명 AR 안경 등 세부 묘사가 프롬프트와 다소 차이가 있습니다."
     }
    ],
    "readings": [
     {
      "label": "B",
      "direction": "현석의 시선은 화면 우측 창밖 하늘의 새 떼를 정확히 향하고 있습니다.",
      "built_space": "교실 내부, 창문 벽을 정면으로 바라보는 측면 구도이며 인물들이 자기 자리에 바르게 착석해 있습니다.",
      "entities": "현석의 외모와 의상은 레퍼런스와 일치하며, 배경의 학생은 투명 AR 안경을 착용하고 펜을 쥔 채 앉아 있습니다. 현석의 책상 위에는 안경이 놓여 있습니다.",
      "hard_violations": [],
      "physics": "안정적인 착석 자세와 손의 위치, 책상 위 사물들의 지지가 물리적으로 자연스럽습니다."
     },
     {
      "label": "A",
      "direction": "현석의 시선은 우측 창밖의 새 떼를 향하고 있습니다.",
      "built_space": "교실 내부, 뒤에서 앞쪽 측면을 바라보는 구도로 창문과 칠판의 위치가 적절하게 배치되어 있습니다.",
      "entities": "현석의 얼굴과 의상은 일치하나, 배경 학생들의 안경이 불투명한 고글 형태에 가깝고 지시된 필기구를 쥔 손이 뚜렷하게 보이지 않습니다.",
      "hard_violations": [],
      "physics": "착석 자세와 사물들의 물리적 지지는 모두 정상적입니다."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "agree",
     "gap": 0.0,
     "per_model_winner": {
      "gemini-pro": "B",
      "openrouter:x-ai/grok-4.6": "B"
     },
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 8,
      "verdict_ko": "요구된 완벽한 측면(clean profile) 구도와 화면 좌우 배분을 정확히 구현했으며, 배경 학생이 AR 안경을 쓰고 펜을 쥐고 있는 세부 지시까지 충실히 반영했습니다."
     },
     {
      "label": "B",
      "score": 5,
      "verdict_ko": "지시된 완벽한 측면이 아닌 반측면 구도이며, 배경 학생의 펜과 투명 AR 안경 등 세부 묘사가 프롬프트와 다소 차이가 있습니다."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "A",
      "direction": "현석의 시선은 화면 우측 창밖 하늘의 새 떼를 정확히 향하고 있습니다.",
      "built_space": "교실 내부, 창문 벽을 정면으로 바라보는 측면 구도이며 인물들이 자기 자리에 바르게 착석해 있습니다.",
      "entities": "현석의 외모와 의상은 레퍼런스와 일치하며, 배경의 학생은 투명 AR 안경을 착용하고 펜을 쥔 채 앉아 있습니다. 현석의 책상 위에는 안경이 놓여 있습니다.",
      "hard_violations": [],
      "physics": "안정적인 착석 자세와 손의 위치, 책상 위 사물들의 지지가 물리적으로 자연스럽습니다."
     },
     {
      "label": "B",
      "direction": "현석의 시선은 우측 창밖의 새 떼를 향하고 있습니다.",
      "built_space": "교실 내부, 뒤에서 앞쪽 측면을 바라보는 구도로 창문과 칠판의 위치가 적절하게 배치되어 있습니다.",
      "entities": "현석의 얼굴과 의상은 일치하나, 배경 학생들의 안경이 불투명한 고글 형태에 가깝고 지시된 필기구를 쥔 손이 뚜렷하게 보이지 않습니다.",
      "hard_violations": [],
      "physics": "착석 자세와 사물들의 물리적 지지는 모두 정상적입니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 15,
     "B": 9
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   }
  },
  "readings": [
   {
    "label": "A",
    "direction": "현석의 시선은 오른쪽 창밖 하늘에 있는 새 떼를 정확히 향함. 뒤쪽 남학생은 책상 위를 내려다보며 펜으로 글을 씀.",
    "built_space": "지정된 교실의 창가 바로 옆 줄에 책상이 위치하며, 카메라 시점과 창문의 공간적 배열이 프롬프트 및 기준 사진과 일치함.",
    "entities": "현석의 얼굴, 헤어스타일, 복장(조끼, 타이)이 기준과 일치함. 책상 위 벗어둔 안경, 뒤에서 AR 안경을 쓰고 필기하는 제호, 창밖 새 떼가 모두 묘사됨.",
    "hard_violations": [],
    "physics": "현석과 배경 인물들 모두 의자에 안정적으로 앉아 몸을 지지하고 있으며, 펜을 쥔 손과 책상의 접촉이 자연스러움."
   },
   {
    "label": "B",
    "direction": "현석은 오른쪽을 응시하나 창문이 멀어 시선의 목적지가 다소 불분명함. 배경의 학생들은 대부분 정면이나 모호한 곳을 응시함.",
    "built_space": "현석이 창가 자리가 아닌 교실 중앙 열에 앉아 있어, '창가 옆(beside the window)'이라는 위치 설정을 위반함.",
    "entities": "현석의 인물 특징과 책상 위 AR 안경은 존재하나, 프롬프트가 요구한 '펜을 쥐고 긴장한 채 글을 쓰는 제호'가 뒤쪽에 묘사되지 않음.",
    "hard_violations": [],
    "physics": "학생들이 의자에 앉아 책상에 팔을 기대고 있는 등 물리적 지지 상태는 정상적임."
   }
  ],
  "totals": {
   "A": 15,
   "B": 9
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 7,
    "verdict_ko": "창가 바로 옆 좌석 배치, 클로즈업 구도, 책상 위 AR 안경, 뒤에서 필기하는 제호 등 프롬프트의 까다로운 세부 사항을 훌륭하게 구현했습니다."
   },
   {
    "label": "B",
    "score": 4,
    "verdict_ko": "인물이 창가 옆이 아닌 교실 중앙에 배치되어 위치 제약을 어겼으며, 뒤에서 펜을 쥐고 있는 인물(제호)의 지정된 행동 묘사도 누락되었습니다."
   }
  ],
  "refs": [
   {
    "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.",
    "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/46575226-4c2f-4334-b16a-429b11371e24/episodes/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/images/background_chain/L01B01.png",
    "asset_id": "04872d30-3ac8-451e-9c07-fa7856fd97f9",
    "role": "location_plate"
   },
   {
    "label": "CHARACTER REFERENCE — 현석: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:973165>",
    "asset_id": "37299bdd-ae7a-4ecb-a36a-05c753ff9d51",
    "role": "character_ref"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "원본 배경 이미지를 변경 없이 정확히 유지하라는 지시와 달리, 좌측 벽면의 화이트보드, 수납장 구조 및 천장 조명의 형태와 위치가 다르게 렌더링되었습니다.",
     "fix_en": "Restore the whiteboard, cabinets, and ceiling lights on the left to match the background reference. Preserve Hyunseok, his clothing, the students on the right, the lighting, and the framing.",
     "severity": "major",
     "observation_index": 0,
     "needs_ref_indices": [
      1
     ],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": true,
     "unfixable": false
    },
    {
     "issue_ko": "레이아웃 스케치에 지정된 위치에만 인물을 배치하라는 지시와 달리, 화면 좌측(창가 반대편 열)에 스케치에 존재하지 않는 추가 학생 인물들이 배치되었습니다.",
     "fix_en": "Remove the extra students from the left side of the classroom. Preserve Hyunseok, his clothing, the students in the right row, the classroom architecture, the lighting, and the framing.",
     "severity": "major",
     "observation_index": 1,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "현석 뒤 창가에서 필기하는 학생 등 뒷줄 인물들이 초점에서 물러난 흐릿한 부분 형상이 아니라 또렷이 보인다.",
     "fix_en": "Blur the students behind Hyunseok to appear as soft, out-of-focus figures. Preserve Hyunseok, his clothing, the classroom architecture, the lighting, and the framing.",
     "severity": "major",
     "observation_index": 2,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "얕은 초점이 현석의 눈에 걸려 있지 않고 창밖 도시와 오른쪽 창의 새 떼까지 선명하다.",
     "fix_en": "Apply shallow focus to blur the window, city, and birds, keeping only Hyunseok's face sharp. Preserve Hyunseok, his clothing, the students behind him, the lighting, and the framing.",
     "severity": "major",
     "observation_index": 3,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "현석의 시선이 새 떼가 있는 창 상단이 아니라 창 중간의 도시 쪽을 향한다.",
     "fix_en": "Adjust Hyunseok's eyes to look upward toward the birds. Preserve his head position, his clothing, the students behind him, the classroom, the lighting, and the framing.",
     "severity": "major",
     "observation_index": 4,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    }
   ],
   "observer_observations": [
    {
     "issue_ko": "원본 배경 이미지를 변경 없이 정확히 유지하라는 지시와 달리, 좌측 벽면의 화이트보드, 수납장 구조 및 천장 조명의 형태와 위치가 다르게 렌더링되었습니다.",
     "severity": "major"
    },
    {
     "issue_ko": "레이아웃 스케치에 지정된 위치에만 인물을 배치하라는 지시와 달리, 화면 좌측(창가 반대편 열)에 스케치에 존재하지 않는 추가 학생 인물들이 배치되었습니다.",
     "severity": "major"
    },
    {
     "issue_ko": "현석 뒤 창가에서 필기하는 학생 등 뒷줄 인물들이 초점에서 물러난 흐릿한 부분 형상이 아니라 또렷이 보인다.",
     "severity": "major"
    },
    {
     "issue_ko": "얕은 초점이 현석의 눈에 걸려 있지 않고 창밖 도시와 오른쪽 창의 새 떼까지 선명하다.",
     "severity": "major"
    },
    {
     "issue_ko": "현석의 시선이 새 떼가 있는 창 상단이 아니라 창 중간의 도시 쪽을 향한다.",
     "severity": "major"
    }
   ],
   "observer_models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ],
   "observer_counts": {
    "gemini-pro": 2,
    "openrouter:x-ai/grok-4.6": 3
   }
  },
  "fix_severity_skipped_count": 5,
  "fix_severity_skipped": [
   {
    "issue_ko": "원본 배경 이미지를 변경 없이 정확히 유지하라는 지시와 달리, 좌측 벽면의 화이트보드, 수납장 구조 및 천장 조명의 형태와 위치가 다르게 렌더링되었습니다.",
    "fix_en": "Restore the whiteboard, cabinets, and ceiling lights on the left to match the background reference. Preserve Hyunseok, his clothing, the students on the right, the lighting, and the framing.",
    "severity": "major",
    "observation_index": 0,
    "needs_ref_indices": [
     1
    ],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": true,
    "unfixable": false
   },
   {
    "issue_ko": "레이아웃 스케치에 지정된 위치에만 인물을 배치하라는 지시와 달리, 화면 좌측(창가 반대편 열)에 스케치에 존재하지 않는 추가 학생 인물들이 배치되었습니다.",
    "fix_en": "Remove the extra students from the left side of the classroom. Preserve Hyunseok, his clothing, the students in the right row, the classroom architecture, the lighting, and the framing.",
    "severity": "major",
    "observation_index": 1,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "현석 뒤 창가에서 필기하는 학생 등 뒷줄 인물들이 초점에서 물러난 흐릿한 부분 형상이 아니라 또렷이 보인다.",
    "fix_en": "Blur the students behind Hyunseok to appear as soft, out-of-focus figures. Preserve Hyunseok, his clothing, the classroom architecture, the lighting, and the framing.",
    "severity": "major",
    "observation_index": 2,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "얕은 초점이 현석의 눈에 걸려 있지 않고 창밖 도시와 오른쪽 창의 새 떼까지 선명하다.",
    "fix_en": "Apply shallow focus to blur the window, city, and birds, keeping only Hyunseok's face sharp. Preserve Hyunseok, his clothing, the students behind him, the lighting, and the framing.",
    "severity": "major",
    "observation_index": 3,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "현석의 시선이 새 떼가 있는 창 상단이 아니라 창 중간의 도시 쪽을 향한다.",
    "fix_en": "Adjust Hyunseok's eyes to look upward toward the birds. Preserve his head position, his clothing, the students behind him, the classroom, the lighting, and the framing.",
    "severity": "major",
    "observation_index": 4,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   }
  ],
  "fix_skipped": true,
  "fix_skip_reason": "no_critical_issue",
  "shot_run_produced": true,
  "shot_run_uid": "06a8c290-7277-738e-89f7-10cf92e01523",
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S1sh4__bgfirst_bg.png",
   "bg_asset_id": "cfac60af-05ee-434e-af2b-89b318579af4",
   "bg_record_key": "S1sh4::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S1sh4::cine": {
  "applied": true,
  "fingerprint": "c8e1824baed764f4c289ef05d7a12dcc9ba6ad0fdeeaf75c8305540e6f0786e0",
  "model": "x-ai/grok-imagine-image-2.0",
  "pack": "20.202608141300",
  "source_file": "S1sh4_sel.png",
  "source_sha256": "265c2a9b0b21236af5fcaf8055366d4b668917bb2758988145c19fa7442fd8f1",
  "file": "S1sh4_cine.png",
  "latency_ms": 16554
 },
 "S1sh8::signage": {
  "fp": "3024e54e5fd3858e",
  "inscriptions": []
 },
 "S1sh8": {
  "input_fingerprint": "4219dbf1a09e6685",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): morning.\n\nSHOT TEXT (authoritative, Korean): 백색 불빛 아래, 제호가 두려움에 찬 굳은 얼굴로 입을 살짝 벌린 채 멈춰 있는 찰나.\n\nLOCATION (lock): Inside the student desk rows of a high-school classroom under white LED lighting. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The dolly-in settles slightly above Jeho's seated eye line and off his front side, angling down into a close three-quarter portrait that keeps his parted mouth and rigid jaw visible. Jeho occupies most of the center-right frame while the classroom falls away behind him, his fixed gaze remaining directed toward the teacher as the announcement lands.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 제호 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: classroom desks and seated students (arranged in rows; students remain largely expressionless) — The rows recede diagonally behind Jeho and are softened by the close focus; used as Supplies restrained institutional context without competing with Jeho's face; transparent AR glasses (worn by Jeho and displaying the government promotional video) — The lenses sit obliquely to camera, leaving Jeho's frightened eyes visible through them; used as Places the imposed media layer directly across Jeho's strained expression.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The white LED illumination stays cool and emotionally flat, exposing Jeho's fear without dramatic shadow.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established about the place — its fixed features, wear and lighting — persists. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok's transparent AR glasses remain on his desk while he stays without them. Jeho remains seated wearing his AR glasses and holding his writing instrument.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): morning.\n\nSHOT TEXT (authoritative, Korean): 백색 불빛 아래, 제호가 두려움에 찬 굳은 얼굴로 입을 살짝 벌린 채 멈춰 있는 찰나.\n\nLOCATION (lock): Inside the student desk rows of a high-school classroom under white LED lighting. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The dolly-in settles slightly above Jeho's seated eye line and off his front side, angling down into a close three-quarter portrait that keeps his parted mouth and rigid jaw visible. Jeho occupies most of the center-right frame while the classroom falls away behind him, his fixed gaze remaining directed toward the teacher as the announcement lands.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 제호 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: classroom desks and seated students (arranged in rows; students remain largely expressionless) — The rows recede diagonally behind Jeho and are softened by the close focus; used as Supplies restrained institutional context without competing with Jeho's face; transparent AR glasses (worn by Jeho and displaying the government promotional video) — The lenses sit obliquely to camera, leaving Jeho's frightened eyes visible through them; used as Places the imposed media layer directly across Jeho's strained expression.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The white LED illumination stays cool and emotionally flat, exposing Jeho's fear without dramatic shadow.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established about the place — its fixed features, wear and lighting — persists. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok's transparent AR glasses remain on his desk while he stays without them. Jeho remains seated wearing his AR glasses and holding his writing instrument.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): morning.\n\nSHOT TEXT (authoritative, Korean): 백색 불빛 아래, 제호가 두려움에 찬 굳은 얼굴로 입을 살짝 벌린 채 멈춰 있는 찰나.\n\nLOCATION (lock): Inside the student desk rows of a high-school classroom under white LED lighting. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The dolly-in settles slightly above Jeho's seated eye line and off his front side, angling down into a close three-quarter portrait that keeps his parted mouth and rigid jaw visible. Jeho occupies most of the center-right frame while the classroom falls away behind him, his fixed gaze remaining directed toward the teacher as the announcement lands.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 제호 in the middle-center of the frame, midground.\n- KEY BACKGROUND ELEMENTS: classroom desks and seated students (arranged in rows; students remain largely expressionless) — The rows recede diagonally behind Jeho and are softened by the close focus; used as Supplies restrained institutional context without competing with Jeho's face; transparent AR glasses (worn by Jeho and displaying the government promotional video) — The lenses sit obliquely to camera, leaving Jeho's frightened eyes visible through them; used as Places the imposed media layer directly across Jeho's strained expression.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: The white LED illumination stays cool and emotionally flat, exposing Jeho's fear without dramatic shadow.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established about the place — its fixed features, wear and lighting — persists. The people visible in that photo are NOT in this shot: never carry their faces, bodies or clothing onto anyone here. Each person in THIS shot is defined solely by the PEOPLE list, their own CHARACTER REFERENCE images and their wardrobe notes. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok's transparent AR glasses remain on his desk while he stays without them. Jeho remains seated wearing his AR glasses and holding his writing instrument.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, but reframe like a film director choosing a different setup — a different angle, height, distance or foreground layer from the obvious one. For this candidate only, treat the CAMERA & FRAME contract above as a baseline to depart from, not a lock; the references and the shot text fix WHAT is seen, not WHERE the camera stands. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "A",
    "direction": "시선이 화면 좌측의 교실 앞쪽(교사 방향)을 주시하고 있음.",
    "built_space": "흰색 LED 조명 아래 교실 책상들이 대각선 뒤쪽으로 부드럽게 멀어지며 배열됨.",
    "entities": "제호가 AR 안경을 쓴 채 입을 살짝 벌리고 굳은 표정으로 프레임의 우측 중앙을 차지하고 있음.",
    "hard_violations": [],
    "physics": "자리에 앉아 안정적으로 자세를 유지하고 있으며, 클로즈업 프레이밍 상 손은 보이지 않음."
   },
   {
    "label": "B",
    "direction": "시선이 교사가 아닌 카메라 정면 렌즈 쪽을 향하고 있음.",
    "built_space": "교실 내부의 책상과 학생들이 카메라를 향해 정면 좌우 대칭으로 배열되어 있음.",
    "entities": "제호가 화면 정중앙에 배치되어 투명 AR 안경을 쓰고 있으나, 입을 굳게 다물고 있음.",
    "hard_violations": [],
    "physics": "자리에 앉아 양손을 책상 위에 올리고 오른손으로 펜을 쥐고 지지함."
   }
  ],
  "gq": {
   "route": "agree",
   "gap": 0.0,
   "per_model_winner": {
    "gemini-pro": "A",
    "openrouter:x-ai/grok-4.6": "A"
   },
   "models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ]
  },
  "totals": {
   "A": 9,
   "B": 3
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 9,
    "verdict_ko": "지정된 3/4 측면 클로즈업 구도, 우측 중앙의 인물 배치 및 대각선으로 멀어지는 교실 배경을 완벽하게 따랐으며, 살짝 벌어진 입과 두려운 표정도 매우 훌륭하게 구현되었습니다."
   },
   {
    "label": "B",
    "score": 3,
    "verdict_ko": "명확히 지시된 3/4 측면 구도와 대각선 배경을 무시하고 정면 대칭 구도를 취했으며, 입을 다물고 있어 표정 지시사항도 충족하지 못했습니다."
   }
  ],
  "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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S1sh4_sel.png",
    "asset_id": "68ae30c4-0399-4150-abdd-94810991e3d5",
    "role": "prev_still"
   },
   {
    "label": "CHARACTER REFERENCE — 제호: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:847626>",
    "asset_id": "7760e702-1e92-4b92-b945-00de7bc3d3c1",
    "role": "character_ref"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "제호가 캐릭터 레퍼런스(파란색 조끼와 넥타이)와 다르게, 프롬프트에서 금지한 이전 샷(Previous Shot Still) 인물의 의상(어두운 조끼와 넥타이)을 그대로 입고 있습니다.",
     "fix_en": "Change Jeho's vest to a solid blue and his tie to blue. Preserve his face, glasses, pose, lighting, and all background students.",
     "severity": "critical",
     "observation_index": 0,
     "needs_ref_indices": [
      2
     ],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "화면 좌측 하단 책상 위에 연필을 쥔 손과 흰 소매가 나타나지만, 그 뒤에 앉은 학생의 내려간 오른쪽 어깨와 해부학적으로 연결되지 않는 분리된 신체 부위입니다.",
     "fix_en": "Erase the disembodied hand and orange pencil in the lower left, replacing them with the blurred desk surface. Preserve Jeho, his clothing, and all other background details.",
     "severity": "critical",
     "observation_index": 1,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "제호가 교사를 보지 않고 렌즈를 거의 정면으로 응시한다.",
     "fix_en": "Shift Jeho's gaze off-camera.",
     "severity": "major",
     "observation_index": 2,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "카메라가 눈높이 위에서 내려다보는 사분면이 아니라 거의 정면 수평 클로즈업이다.",
     "fix_en": "Angle camera slightly downward.",
     "severity": "major",
     "observation_index": 3,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": true,
     "unfixable": false
    },
    {
     "issue_ko": "제호의 AR 안경 렌즈가 비스듬하지 않고 카메라를 정면으로 향한다.",
     "fix_en": "Angle the glasses lenses.",
     "severity": "major",
     "observation_index": 4,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "배경 학생들이 렌즈를 향해 줄 맞춰 앉아 연출 단체 사진처럼 보인다.",
     "fix_en": "Realign background desks to recede diagonally.",
     "severity": "major",
     "observation_index": 5,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": true,
     "unfixable": false
    },
    {
     "issue_ko": "제호의 얼굴이 캐릭터 레퍼런스와 일치하지 않는다.",
     "fix_en": "Match Jeho's facial features to the reference.",
     "severity": "major",
     "observation_index": 6,
     "needs_ref_indices": [
      2
     ],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "왼쪽 뒷문 쪽에 서 있는 인물이 있어 착석한 학생 열과 어긋난다.",
     "fix_en": "Remove the standing person near the back door.",
     "severity": "major",
     "observation_index": 8,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "배경 책상 열이 클로즈 포커스로 흐려지지 않고 학생들이 선명하다.",
     "fix_en": "Apply stronger depth-of-field blur to the background.",
     "severity": "minor",
     "observation_index": 9,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "제호의 앞머리가 단정히 빗긴 상태가 아니라 흐트러져 있다.",
     "fix_en": "Smooth Jeho's bangs.",
     "severity": "minor",
     "observation_index": 10,
     "needs_ref_indices": [
      2
     ],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    }
   ],
   "observer_observations": [
    {
     "issue_ko": "제호가 캐릭터 레퍼런스(파란색 조끼와 넥타이)와 다르게, 프롬프트에서 금지한 이전 샷(Previous Shot Still) 인물의 의상(어두운 조끼와 넥타이)을 그대로 입고 있습니다.",
     "severity": "critical"
    },
    {
     "issue_ko": "화면 좌측 하단 책상 위에 연필을 쥔 손과 흰 소매가 나타나지만, 그 뒤에 앉은 학생의 내려간 오른쪽 어깨와 해부학적으로 연결되지 않는 분리된 신체 부위입니다.",
     "severity": "critical"
    },
    {
     "issue_ko": "제호가 교사를 보지 않고 렌즈를 거의 정면으로 응시한다.",
     "severity": "major"
    },
    {
     "issue_ko": "카메라가 눈높이 위에서 내려다보는 사분면이 아니라 거의 정면 수평 클로즈업이다.",
     "severity": "major"
    },
    {
     "issue_ko": "제호의 AR 안경 렌즈가 비스듬하지 않고 카메라를 정면으로 향한다.",
     "severity": "major"
    },
    {
     "issue_ko": "배경 학생들이 렌즈를 향해 줄 맞춰 앉아 연출 단체 사진처럼 보인다.",
     "severity": "major"
    },
    {
     "issue_ko": "제호의 얼굴이 캐릭터 레퍼런스와 일치하지 않는다.",
     "severity": "major"
    },
    {
     "issue_ko": "제호가 레퍼런스의 파란 조끼·파란 넥타이가 아닌 회색 조끼·줄무늬 넥타이를 입었다.",
     "severity": "major"
    },
    {
     "issue_ko": "왼쪽 뒷문 쪽에 서 있는 인물이 있어 착석한 학생 열과 어긋난다.",
     "severity": "major"
    },
    {
     "issue_ko": "배경 책상 열이 클로즈 포커스로 흐려지지 않고 학생들이 선명하다.",
     "severity": "minor"
    },
    {
     "issue_ko": "제호의 앞머리가 단정히 빗긴 상태가 아니라 흐트러져 있다.",
     "severity": "minor"
    }
   ],
   "observer_models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ],
   "observer_counts": {
    "gemini-pro": 2,
    "openrouter:x-ai/grok-4.6": 9
   }
  },
  "fix_severity_skipped_count": 8,
  "fix_severity_skipped": [
   {
    "issue_ko": "제호가 교사를 보지 않고 렌즈를 거의 정면으로 응시한다.",
    "fix_en": "Shift Jeho's gaze off-camera.",
    "severity": "major",
    "observation_index": 2,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "카메라가 눈높이 위에서 내려다보는 사분면이 아니라 거의 정면 수평 클로즈업이다.",
    "fix_en": "Angle camera slightly downward.",
    "severity": "major",
    "observation_index": 3,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": true,
    "unfixable": false
   },
   {
    "issue_ko": "제호의 AR 안경 렌즈가 비스듬하지 않고 카메라를 정면으로 향한다.",
    "fix_en": "Angle the glasses lenses.",
    "severity": "major",
    "observation_index": 4,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "배경 학생들이 렌즈를 향해 줄 맞춰 앉아 연출 단체 사진처럼 보인다.",
    "fix_en": "Realign background desks to recede diagonally.",
    "severity": "major",
    "observation_index": 5,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": true,
    "unfixable": false
   },
   {
    "issue_ko": "제호의 얼굴이 캐릭터 레퍼런스와 일치하지 않는다.",
    "fix_en": "Match Jeho's facial features to the reference.",
    "severity": "major",
    "observation_index": 6,
    "needs_ref_indices": [
     2
    ],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "왼쪽 뒷문 쪽에 서 있는 인물이 있어 착석한 학생 열과 어긋난다.",
    "fix_en": "Remove the standing person near the back door.",
    "severity": "major",
    "observation_index": 8,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "배경 책상 열이 클로즈 포커스로 흐려지지 않고 학생들이 선명하다.",
    "fix_en": "Apply stronger depth-of-field blur to the background.",
    "severity": "minor",
    "observation_index": 9,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "제호의 앞머리가 단정히 빗긴 상태가 아니라 흐트러져 있다.",
    "fix_en": "Smooth Jeho's bangs.",
    "severity": "minor",
    "observation_index": 10,
    "needs_ref_indices": [
     2
    ],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   }
  ],
  "repair_mode": "edit",
  "fix_ref_gate": {
   "mode": "gated",
   "available": 2,
   "kept": 1,
   "requested_indices": [
    2
   ],
   "kept_labels": [
    "CHARACTER REFERENCE — 제호: the exact person appearing in this shot; match face, hair and build exactly."
   ]
  },
  "fix_ref_count": 2,
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change Jeho's vest to a solid blue and his tie to blue. Preserve his face, glasses, pose, lighting, and all background students.\n- Erase the disembodied hand and orange pencil in the lower left, replacing them with the blurred desk surface. Preserve Jeho, his clothing, and all other background details.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. Real-world writing that belongs\nto the scene — signs, banners, documents — stays exactly as the\noriginal shows it; do not add captions, subtitles, watermarks or\nany overlay text.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "클로즈업·살짝 벌린 입·투명 AR 안경·교탁열 구도가 샷 텍스트에 맞고 제호 얼굴이 레퍼런스에 가깝다. 다만 시선이 교사 쪽이 아닌 카메라 근처를 보고 조끼가 레퍼런스 청색이 아닌 회색이며 배경 여학생을 추가했다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "조끼·넥타이를 레퍼런스 청색으로 맞춘 수리는 있으나 제호 얼굴을 더 다르게 만들고 펜을 든 손을 지워 휴대 상태를 깨뜨렸으며 시선·구도 문제는 그대로라 정체성 손실로 뒤진다."
     }
    ],
    "readings": [
     {
      "label": "A",
      "direction": "제호의 시선은 정면 약간 우측, 카메라에 거의 정면으로 고정되어 있다. 샷이 요구하는 교사 쪽을 향한 고정 시선이 아니라 렌즈 근처를 보고 있다. 입은 살짝 벌어져 있다. 배경 학생들은 대체로 전방을 본다.",
      "built_space": "고등학교 교실 내부. 흰 LED 천장등, 뒤쪽 화이트보드, 왼쪽 목재 문·창, 오른쪽 큰 창. 책상 열이 대각선으로 뒤로 물러난다. 제호는 앞줄에 앉아 프레임 중·우측을 차지한다. 조향장치 등 차량 요소는 없다.",
      "entities": "전경은 한국인 10대 후반 남성으로 짧은 검은 머리, 투명 AR 안경, 회색 조끼·회색 셔츠·줄무늬 넥타이. 얼굴은 캐릭터 레퍼런스에 가깝다. 조끼 색은 레퍼런스 청색이 아니라 이전 스틸의 회색에 가깝다. 오른쪽에 필기구를 든 손이 보인다. 배경에 무표정한 남녀 학생들이 줄지어 앉아 있다. 이전 스틸의 인물은 복제되지 않았다.",
      "hard_violations": [],
      "physics": "제호는 의자에 앉아 있고 엉덩이가 좌석에 있다. 필기구를 쥔 손이 책상 높이에서 지지된다. 공중에 뜬 신체나 물체는 없다."
     },
     {
      "label": "B",
      "direction": "제호의 시선은 A와 같이 거의 정면·카메라 쪽으로, 교사를 향한 고정 시선이 아니다. 입은 살짝 벌어져 있다. 배경 학생들은 전방을 본다.",
      "built_space": "A와 같은 교실 구조: LED등, 화이트보드, 왼쪽 문, 오른쪽 창, 대각선으로 물러나는 책상 열. 제호가 중·우측 클로즈업.",
      "entities": "전경 남성의 얼굴이 A·캐릭터 레퍼런스보다 더 길고 이질적이며 정체성이 어긋난다. 짧은 검은 머리, 투명 AR 안경. 조끼·셔츠·넥타이는 레퍼런스에 가까운 청색으로 바뀌었다. A에 보이던 필기구를 든 손이 사라져 휴대 상태가 깨진다. 배경 학생 조끼도 청색으로 통일됐다.",
      "hard_violations": [],
      "physics": "제호는 앉아 있다. A에서 보이던 필기구 손이 잘려 휴대·지지 관계가 화면에서 확인되지 않는다. 떠 있는 신체는 없다."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "single_openrouter:x-ai/grok-4.6",
     "models": [
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "클로즈업·살짝 벌린 입·투명 AR 안경·교탁열 구도가 샷 텍스트에 맞고 제호 얼굴이 레퍼런스에 가깝다. 다만 시선이 교사 쪽이 아닌 카메라 근처를 보고 조끼가 레퍼런스 청색이 아닌 회색이며 배경 여학생을 추가했다."
     },
     {
      "label": "B",
      "score": 3,
      "verdict_ko": "조끼·넥타이를 레퍼런스 청색으로 맞춘 수리는 있으나 제호 얼굴을 더 다르게 만들고 펜을 든 손을 지워 휴대 상태를 깨뜨렸으며 시선·구도 문제는 그대로라 정체성 손실로 뒤진다."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "A",
      "direction": "제호의 시선은 정면 약간 우측, 카메라에 거의 정면으로 고정되어 있다. 샷이 요구하는 교사 쪽을 향한 고정 시선이 아니라 렌즈 근처를 보고 있다. 입은 살짝 벌어져 있다. 배경 학생들은 대체로 전방을 본다.",
      "built_space": "고등학교 교실 내부. 흰 LED 천장등, 뒤쪽 화이트보드, 왼쪽 목재 문·창, 오른쪽 큰 창. 책상 열이 대각선으로 뒤로 물러난다. 제호는 앞줄에 앉아 프레임 중·우측을 차지한다. 조향장치 등 차량 요소는 없다.",
      "entities": "전경은 한국인 10대 후반 남성으로 짧은 검은 머리, 투명 AR 안경, 회색 조끼·회색 셔츠·줄무늬 넥타이. 얼굴은 캐릭터 레퍼런스에 가깝다. 조끼 색은 레퍼런스 청색이 아니라 이전 스틸의 회색에 가깝다. 오른쪽에 필기구를 든 손이 보인다. 배경에 무표정한 남녀 학생들이 줄지어 앉아 있다. 이전 스틸의 인물은 복제되지 않았다.",
      "hard_violations": [],
      "physics": "제호는 의자에 앉아 있고 엉덩이가 좌석에 있다. 필기구를 쥔 손이 책상 높이에서 지지된다. 공중에 뜬 신체나 물체는 없다."
     },
     {
      "label": "B",
      "direction": "제호의 시선은 A와 같이 거의 정면·카메라 쪽으로, 교사를 향한 고정 시선이 아니다. 입은 살짝 벌어져 있다. 배경 학생들은 전방을 본다.",
      "built_space": "A와 같은 교실 구조: LED등, 화이트보드, 왼쪽 문, 오른쪽 창, 대각선으로 물러나는 책상 열. 제호가 중·우측 클로즈업.",
      "entities": "전경 남성의 얼굴이 A·캐릭터 레퍼런스보다 더 길고 이질적이며 정체성이 어긋난다. 짧은 검은 머리, 투명 AR 안경. 조끼·셔츠·넥타이는 레퍼런스에 가까운 청색으로 바뀌었다. A에 보이던 필기구를 든 손이 사라져 휴대 상태가 깨진다. 배경 학생 조끼도 청색으로 통일됐다.",
      "hard_violations": [],
      "physics": "제호는 앉아 있다. A에서 보이던 필기구 손이 잘려 휴대·지지 관계가 화면에서 확인되지 않는다. 떠 있는 신체는 없다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "클로즈업·벌린 입·AR 안경은 맞지만 제호 교복을 파란 조끼·파란 넥타이로 바꿔 정체성·이전 샷 의상을 깨뜨렸고 필기구도 안 보인다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "제호 얼굴·회색 조끼·줄무늬 넥타이·필기구를 유지한 채 두려운 클로즈업과 AR 안경을 구현했고, A의 의상 개조 손실이 없다."
     }
    ],
    "readings": [
     {
      "label": "A",
      "direction": "제호는 정면 약간 우측, 카메라 너머 교실 앞(교사 방향)을 응시한다. 배경 학생들도 대체로 같은 앞쪽을 본다.",
      "built_space": "고등학교 교실 내부. 전면 화이트보드, 형광등, 좌측 문, 우측 창, 줄지어 앉은 책상. 제호는 앞줄 오른쪽 좌석에 앉은 구도.",
      "entities": "제호로 보이는 한국인 10대 남성, 짧은 검은 머리, 투명 AR 안경, 벌린 입. 그러나 교복이 파란 조끼·파란 줄무늬 넥타이·연회색 셔츠로 바뀌어 캐릭터 레퍼런스·이전 샷의 회색 조끼·회색 넥타이와 불일치. 배경에 다수의 학생(여학생 포함). 제호의 필기구는 보이지 않음.",
      "hard_violations": [],
      "physics": "제호는 좌석에 앉아 있고 상체는 책상 위에 있다. 배경 학생들도 각자의 의자에 앉아 있다. 떠 있는 신체 없음."
     },
     {
      "label": "B",
      "direction": "제호는 정면 약간 우측, 카메라 너머 교실 앞(교사 방향)을 응시한다. 배경 학생들도 대체로 같은 앞쪽을 본다.",
      "built_space": "고등학교 교실 내부. 전면 화이트보드, 형광등, 좌측 문, 우측 창, 줄지어 앉은 책상. 제호는 앞줄 오른쪽 좌석에 앉은 구도. A와 거의 같은 공간.",
      "entities": "제호: 한국인 10대 남성, 짧은 검은 머리, 투명 AR 안경, 벌린 입, 회색 니트 조끼·회색 셔츠·어두운 줄무늬 넥타이로 이전 샷·레퍼런스 계열과 일치. 배경 학생들(여학생 포함). 제호 쪽에서 연필을 쥔 손이 보인다.",
      "hard_violations": [],
      "physics": "제호는 좌석에 앉아 있고 손이 필기구를 쥐고 있다. 배경 학생들도 의자에 앉아 있다. 떠 있는 신체 없음."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "single_openrouter:x-ai/grok-4.6",
     "models": [
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "클로즈업·벌린 입·AR 안경은 맞지만 제호 교복을 파란 조끼·파란 넥타이로 바꿔 정체성·이전 샷 의상을 깨뜨렸고 필기구도 안 보인다."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "제호 얼굴·회색 조끼·줄무늬 넥타이·필기구를 유지한 채 두려운 클로즈업과 AR 안경을 구현했고, A의 의상 개조 손실이 없다."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "B",
      "direction": "제호는 정면 약간 우측, 카메라 너머 교실 앞(교사 방향)을 응시한다. 배경 학생들도 대체로 같은 앞쪽을 본다.",
      "built_space": "고등학교 교실 내부. 전면 화이트보드, 형광등, 좌측 문, 우측 창, 줄지어 앉은 책상. 제호는 앞줄 오른쪽 좌석에 앉은 구도.",
      "entities": "제호로 보이는 한국인 10대 남성, 짧은 검은 머리, 투명 AR 안경, 벌린 입. 그러나 교복이 파란 조끼·파란 줄무늬 넥타이·연회색 셔츠로 바뀌어 캐릭터 레퍼런스·이전 샷의 회색 조끼·회색 넥타이와 불일치. 배경에 다수의 학생(여학생 포함). 제호의 필기구는 보이지 않음.",
      "hard_violations": [],
      "physics": "제호는 좌석에 앉아 있고 상체는 책상 위에 있다. 배경 학생들도 각자의 의자에 앉아 있다. 떠 있는 신체 없음."
     },
     {
      "label": "A",
      "direction": "제호는 정면 약간 우측, 카메라 너머 교실 앞(교사 방향)을 응시한다. 배경 학생들도 대체로 같은 앞쪽을 본다.",
      "built_space": "고등학교 교실 내부. 전면 화이트보드, 형광등, 좌측 문, 우측 창, 줄지어 앉은 책상. 제호는 앞줄 오른쪽 좌석에 앉은 구도. A와 거의 같은 공간.",
      "entities": "제호: 한국인 10대 남성, 짧은 검은 머리, 투명 AR 안경, 벌린 입, 회색 니트 조끼·회색 셔츠·어두운 줄무늬 넥타이로 이전 샷·레퍼런스 계열과 일치. 배경 학생들(여학생 포함). 제호 쪽에서 연필을 쥔 손이 보인다.",
      "hard_violations": [],
      "physics": "제호는 좌석에 앉아 있고 손이 필기구를 쥐고 있다. 배경 학생들도 의자에 앉아 있다. 떠 있는 신체 없음."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 14,
     "B": 7
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "A",
   "fix_won": false,
   "all_candidates_fail": false,
   "policy": "fix_rejudge_v2_orig_on_tie"
  },
  "shot_run_produced": true,
  "shot_run_uid": "06a8c2bb-b4bf-7c54-911c-702d1901ea1a",
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S1sh4"
  }
 },
 "S1sh8::cine": {
  "applied": true,
  "fingerprint": "82930f10640d598161aa2d718980d3ede7c61db144da9c96d319e2f41224ea70",
  "model": "x-ai/grok-imagine-image-2.0",
  "pack": "20.202608141300",
  "source_file": "S1sh8_sel.png",
  "source_sha256": "bd8c57ef03ad9395cf10ebdaa3ff309f612bd3980dd9e6d9463617ba4f7ff6fe",
  "file": "S1sh8_cine.png",
  "latency_ms": 17058
 },
 "S2sh1::signage": {
  "fp": "7e2d107ed47a1c02",
  "inscriptions": [
   {
    "surface_native": "비상구 표지판",
    "text_native": "비상구",
    "reason_ko": "고요한 학교 계단실이라는 공간적 배경의 사실감을 높이기 위해 벽면에 설치된 비상구 안내 표지판이 필요합니다."
   }
  ]
 },
 "era_assess::113cea76f4968e84": {
  "subjects": [
   {
    "subject_native": "2040년대 서울의 학교 계단실",
    "search_terms_native": [
     "한국 학교 계단",
     "학교 계단 테라조",
     "학교 계단실 복도"
    ],
    "language_lock_native": "모든 검색어는 반드시 한국어로만 작성되어야 하며 영어나 다른 언어로 번역해서는 안 됩니다.",
    "reason_ko": "한국 학교 계단실 특유의 테라조(도끼다시) 바닥, 계단 모서리의 황동 논슬립 디테일, 투톤 벽면 페인트 및 소방 안전 표지판의 형태는 서구식 계단실과 명확히 구분됩니다."
   }
  ]
 },
 "era_ref::966de134c15448ad": {
  "subject": "2040년대 서울의 학교 계단실",
  "terms": [
   "한국 학교 계단",
   "학교 계단 테라조",
   "학교 계단실 복도"
  ],
  "queries": [
   [
    "한국 학교 계단실 복도 테라조 계단",
    "서울 학교 실내 계단 테라조 복도"
   ]
  ],
  "candidates": 4,
  "picked_index": 1,
  "picked_url": "https://blog.kakaocdn.net/dna/nz55d/btrYBdM2ckr/AAAAAAAAAAAAAAAAAAAAABXgAhc76WBZMd5ZqOlYT3bxjxWH4ON-rWVL2dsS-J5q/img.png?allow_ip=&allow_referer=&credential=yqXZFxpELC7KVnFOS48ylbz2pIh7yKj8&expires=1774969199&signature=lf4RQbcMXINQt6ZAHE%2Fq70IADfw%3D",
  "picked_reason_ko": "1번이 서울의 일상적인 학교 계단실을 가장 온전하고 중립적으로 보여 주며, 2040년대에도 이어질 법한 구조·마감·안전 설비를 읽기에 가장 좋은 참고 사진이다.",
  "sha256": "9d3ae248b30cf9a12d39ca5680f9e8f0d56fea0826f9ad09ad2a6890c9c498d7",
  "file": "eraref_966de134c15448ad.png"
 },
 "S2sh1::bgfirst_bg": {
  "input_fingerprint": "4b6a4b0039fba187",
  "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): 낡고 건조한 질감의 회색 콘크리트 계단 한가운데에 제호가 홀로 앉아 핸드폰 화면을 뚫어지게 내려다보는 전경.\n\nLOCATION (lock): Inside a quiet school stairwell, midway along the gray concrete steps.\n\nTIME OF DAY (lock): midday.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From high above the lower side of the staircase, a steep downward wide composition holds Jeho alone near the middle of the steps before the crane begins its descent. His seated figure remains small within the gray stair geometry, torso folded toward the phone held between his hands and attention sealed to its screen.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: gray concrete staircase (old and dry in texture; otherwise empty around Jeho) — The flights and step edges run diagonally through the overhead frame around the isolated figure; used as Creates a severe geometric field that emphasizes Jeho's isolation; Jeho's phone (held in both hands with the '조회하기' button on screen) — The screen faces Jeho and is seen only at a steep, distant oblique angle from above; used as Forms the small focal point toward which the later crane descent will tighten.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient light appropriate to the stairwell maintains a restrained cool-gray palette and low-to-moderate contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. Signage and other writing that permanently belongs to this place may appear, in the place's native language and period style; no captions, watermarks or overlay text.\n\nPERIOD REFERENCE — 2040년대 서울의 학교 계단실: 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): 낡고 건조한 질감의 회색 콘크리트 계단 한가운데에 제호가 홀로 앉아 핸드폰 화면을 뚫어지게 내려다보는 전경.\n\nLOCATION (lock): Inside a quiet school stairwell, midway along the gray concrete steps.\n\nTIME OF DAY (lock): midday.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From high above the lower side of the staircase, a steep downward wide composition holds Jeho alone near the middle of the steps before the crane begins its descent. His seated figure remains small within the gray stair geometry, torso folded toward the phone held between his hands and attention sealed to its screen.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: gray concrete staircase (old and dry in texture; otherwise empty around Jeho) — The flights and step edges run diagonally through the overhead frame around the isolated figure; used as Creates a severe geometric field that emphasizes Jeho's isolation; Jeho's phone (held in both hands with the '조회하기' button on screen) — The screen faces Jeho and is seen only at a steep, distant oblique angle from above; used as Forms the small focal point toward which the later crane descent will tighten.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient light appropriate to the stairwell maintains a restrained cool-gray palette and low-to-moderate contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nRender ONE photorealistic empty location photograph, 16:9, that this shot can be staged inside later. Signage and other writing that permanently belongs to this place may appear, in the place's native language and period style; no captions, watermarks or overlay text.\n\nPERIOD REFERENCE — 2040년대 서울의 학교 계단실: 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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S2sh1__bgfirst_bg.png",
  "asset_id": "3ba72b22-4091-4f09-979d-9dd7ed538b13",
  "input_asset_ids": [
   "ec41c5d6-7814-4d0a-a7ec-08eb9db47376",
   "eb569ab2-420f-4d2e-b846-647e6a9d84f5"
  ],
  "era_research": {
   "subject": "2040년대 서울의 학교 계단실",
   "queries": [
    [
     "한국 학교 계단실 복도 테라조 계단",
     "서울 학교 실내 계단 테라조 복도"
    ]
   ],
   "picked_url": "https://blog.kakaocdn.net/dna/nz55d/btrYBdM2ckr/AAAAAAAAAAAAAAAAAAAAABXgAhc76WBZMd5ZqOlYT3bxjxWH4ON-rWVL2dsS-J5q/img.png?allow_ip=&allow_referer=&credential=yqXZFxpELC7KVnFOS48ylbz2pIh7yKj8&expires=1774969199&signature=lf4RQbcMXINQt6ZAHE%2Fq70IADfw%3D",
   "sha256": "9d3ae248b30cf9a12d39ca5680f9e8f0d56fea0826f9ad09ad2a6890c9c498d7",
   "file": "eraref_966de134c15448ad.png"
  }
 },
 "S2sh1": {
  "input_fingerprint": "1c88b9725022240a",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 낡고 건조한 질감의 회색 콘크리트 계단 한가운데에 제호가 홀로 앉아 핸드폰 화면을 뚫어지게 내려다보는 전경.\n\nLOCATION (lock): Inside a quiet school stairwell, midway along the gray concrete steps. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From high above the lower side of the staircase, a steep downward wide composition holds Jeho alone near the middle of the steps before the crane begins its descent. His seated figure remains small within the gray stair geometry, torso folded toward the phone held between his hands and attention sealed to its screen.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: gray concrete staircase (old and dry in texture; otherwise empty around Jeho) — The flights and step edges run diagonally through the overhead frame around the isolated figure; used as Creates a severe geometric field that emphasizes Jeho's isolation; Jeho's phone (held in both hands with the '조회하기' button on screen) — The screen faces Jeho and is seen only at a steep, distant oblique angle from above; used as Forms the small focal point toward which the later crane descent will tighten.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient light appropriate to the stairwell maintains a restrained cool-gray palette and low-to-moderate contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Jeho continues to hold his smartphone, whose screen is still on the central “조회하기” button before he checks the result.\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 제호 right now, so 제호'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 제호: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else. Nothing else of them needs to be in shot; the framing stays on the object.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 비상구 표지판: \"비상구\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
  "roll_prompts": {
   "A": "Stage the shot. The FIRST attached image (SHOT BACKGROUND) is the finished empty background of this shot — keep it EXACTLY: its camera, perspective, architecture, lighting and every fixed feature stay untouched. The SECOND attached image (LAYOUT SKETCH) tells you ONLY where the people go: each sketched person's position, screen size, pose and the gaze/motion arrows. Ignore the sketch's background lines. The CHARACTER REFERENCE photographs show the real people.\nPlace the real people into the background at exactly the sketched positions, sizes and poses, following the arrow directions. No sketch lines or arrows may remain.\n\nCreate ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 낡고 건조한 질감의 회색 콘크리트 계단 한가운데에 제호가 홀로 앉아 핸드폰 화면을 뚫어지게 내려다보는 전경.\n\nLOCATION (lock): Inside a quiet school stairwell, midway along the gray concrete steps. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From high above the lower side of the staircase, a steep downward wide composition holds Jeho alone near the middle of the steps before the crane begins its descent. His seated figure remains small within the gray stair geometry, torso folded toward the phone held between his hands and attention sealed to its screen.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: gray concrete staircase (old and dry in texture; otherwise empty around Jeho) — The flights and step edges run diagonally through the overhead frame around the isolated figure; used as Creates a severe geometric field that emphasizes Jeho's isolation; Jeho's phone (held in both hands with the '조회하기' button on screen) — The screen faces Jeho and is seen only at a steep, distant oblique angle from above; used as Forms the small focal point toward which the later crane descent will tighten.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient light appropriate to the stairwell maintains a restrained cool-gray palette and low-to-moderate contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Jeho continues to hold his smartphone, whose screen is still on the central “조회하기” button before he checks the result.\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 제호 right now, so 제호'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 제호: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else. Nothing else of them needs to be in shot; the framing stays on the object.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 비상구 표지판: \"비상구\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 낡고 건조한 질감의 회색 콘크리트 계단 한가운데에 제호가 홀로 앉아 핸드폰 화면을 뚫어지게 내려다보는 전경.\n\nLOCATION (lock): Inside a quiet school stairwell, midway along the gray concrete steps. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: From high above the lower side of the staircase, a steep downward wide composition holds Jeho alone near the middle of the steps before the crane begins its descent. His seated figure remains small within the gray stair geometry, torso folded toward the phone held between his hands and attention sealed to its screen.\n- FRAMING SCALE: wide shot\n- KEY BACKGROUND ELEMENTS: gray concrete staircase (old and dry in texture; otherwise empty around Jeho) — The flights and step edges run diagonally through the overhead frame around the isolated figure; used as Creates a severe geometric field that emphasizes Jeho's isolation; Jeho's phone (held in both hands with the '조회하기' button on screen) — The screen faces Jeho and is seen only at a steep, distant oblique angle from above; used as Forms the small focal point toward which the later crane descent will tighten.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient light appropriate to the stairwell maintains a restrained cool-gray palette and low-to-moderate contrast.\n- MATERIAL REALISM: every object and surface must read as a real,\n  physical material — correct texture, weight, wear and light response\n  (metal reflects, fabric drapes and creases, liquid is glossy and\n  pools, painted or drawn marks sit ON a surface and follow its\n  curvature and lighting). Nothing may look like a flat sticker, a\n  doodle or a graphic overlay pasted onto the frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Jeho continues to hold his smartphone, whose screen is still on the central “조회하기” button before he checks the result.\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 제호 right now, so 제호'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 제호: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else. Nothing else of them needs to be in shot; the framing stays on the object.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 비상구 표지판: \"비상구\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, but reframe like a film director choosing a different setup — a different angle, height, distance or foreground layer from the obvious one. For this candidate only, treat the CAMERA & FRAME contract above as a baseline to depart from, not a lock; the references and the shot text fix WHAT is seen, not WHERE the camera stands. 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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S2sh1__bgfirst_bg.png",
     "asset_id": "3ba72b22-4091-4f09-979d-9dd7ed538b13",
     "role": "bgfirst_bg"
    },
    {
     "label": "LAYOUT SKETCH (people placement only)",
     "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/conti/conti_S2sh1.png",
     "asset_id": "ec41c5d6-7814-4d0a-a7ec-08eb9db47376",
     "role": "conti_light"
    },
    {
     "label": "CHARACTER REFERENCE — 제호: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:847626>",
     "asset_id": "7760e702-1e92-4b92-b945-00de7bc3d3c1",
     "role": "character_ref"
    },
    {
     "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.",
     "path": "<bytes:621038>",
     "asset_id": "34c4e322-0ded-425f-a7c2-2f126fa257d7",
     "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/46575226-4c2f-4334-b16a-429b11371e24/episodes/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/images/background_chain/L02B01.png",
     "asset_id": "eb569ab2-420f-4d2e-b846-647e6a9d84f5",
     "role": "location_plate"
    },
    {
     "label": "CHARACTER REFERENCE — 제호: the exact person appearing in this shot; match face, hair and build exactly.",
     "path": "<bytes:847626>",
     "asset_id": "7760e702-1e92-4b92-b945-00de7bc3d3c1",
     "role": "character_ref"
    },
    {
     "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.",
     "path": "<bytes:621038>",
     "asset_id": "34c4e322-0ded-425f-a7c2-2f126fa257d7",
     "role": "prop_ref"
    }
   ]
  },
  "shot_run_spend_attempt_count": 1,
  "judge_flip": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 1429,
      "verdict_ko": "지시된 하이앵글 구도(위에서 내려다보는 시점)를 무시하고 레퍼런스의 정면 구도를 그대로 차용하여 프레이밍 조건(우선순위 2)을 위반했으나, 치명적인 구조적 오류가 없어 B보다 낫습니다."
     },
     {
      "label": "B",
      "score": 1250,
      "verdict_ko": "지시된 카메라 각도는 시도했으나, 우측 상단 계단의 불가능한 건축 기하학과 표지판의 불필요한 영문 텍스트 누출 등 하드 위반 요소가 발생해 탈락했습니다.  ★위반: [gemini-pro] 우측 상단 계단의 물리적으로 불가능한 구조 (원근감 붕괴) / [gemini-pro] 표지판에 지시되지 않은 영문 텍스트(EXIT) 노출"
     }
    ],
    "all_candidates_fail": false,
    "dual": {
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ],
     "normalized": {
      "A": 1.429,
      "B": 1.75
     },
     "adjusted": {
      "A": 1.429,
      "B": 1.25
     },
     "violations": {
      "B": [
       "[gemini-pro] 우측 상단 계단의 물리적으로 불가능한 구조 (원근감 붕괴)",
       "[gemini-pro] 표지판에 지시되지 않은 영문 텍스트(EXIT) 노출"
      ]
     },
     "per_model_winner": {
      "gemini-pro": "A",
      "openrouter:x-ai/grok-4.6": "B"
     },
     "agreed": false
    },
    "gq": {
     "route": "combined",
     "gap": 0.571,
     "per_model_winner": {
      "gemini-pro": "A",
      "openrouter:x-ai/grok-4.6": "B"
     },
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 1429,
      "verdict_ko": "지시된 하이앵글 구도(위에서 내려다보는 시점)를 무시하고 레퍼런스의 정면 구도를 그대로 차용하여 프레이밍 조건(우선순위 2)을 위반했으나, 치명적인 구조적 오류가 없어 B보다 낫습니다."
     },
     {
      "label": "B",
      "score": 1250,
      "verdict_ko": "지시된 카메라 각도는 시도했으나, 우측 상단 계단의 불가능한 건축 기하학과 표지판의 불필요한 영문 텍스트 누출 등 하드 위반 요소가 발생해 탈락했습니다.  ★위반: [gemini-pro] 우측 상단 계단의 물리적으로 불가능한 구조 (원근감 붕괴) / [gemini-pro] 표지판에 지시되지 않은 영문 텍스트(EXIT) 노출"
     }
    ],
    "all_candidates_fail": false
   },
   "reverse_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "엄격하게 요구된 카메라의 가파른 하이 앵글 뷰와 하향 프레이밍 지침을 정확히 구현하여 가장 높은 평가를 받음."
     },
     {
      "label": "B",
      "score": 4,
      "verdict_ko": "장소 레퍼런스와 표지판 텍스트는 완벽히 일치하나, 우선순위가 높은 카메라 하이 앵글 지침을 무시하고 정면 눈높이에서 촬영하여 감점됨."
     }
    ],
    "readings": [
     {
      "label": "A",
      "direction": "시선은 양손에 든 스마트폰 화면을 향하고 있음.",
      "built_space": "여러 갈래로 뻗은 회색 콘크리트 계단 구조. 카메라는 프롬프트 지시대로 계단 위쪽에서 아래를 깊게 내려다보는 가파른 각도(하이 앵글)에 위치함.",
      "entities": "제호(짧은 흑발, 회색 셔츠 및 파란색 조끼 착용), 스마트폰(화면 켜짐). 우측 벽에 부착된 초록색 비상구 표지판(텍스트가 '배상구'로 오기됨).",
      "hard_violations": [],
      "physics": "피사체는 계단 중간에 안정적으로 앉아 있으며, 두 발은 아래쪽 계단 디딤판에 놓여 있음. 양손은 스마트폰을 단단히 쥐고 지탱함."
     },
     {
      "label": "B",
      "direction": "시선은 양손에 든 스마트폰 화면을 향하고 있음.",
      "built_space": "레퍼런스 이미지와 완벽히 일치하는 단조로운 구조의 콘크리트 계단실. 카메라는 지시된 하이 앵글 뷰를 따르지 않고 피사체와 비슷한 눈높이(정면)의 계단참에 위치함.",
      "entities": "제호(레퍼런스와 매우 흡사한 외모 및 의상), 스마트폰. 좌측 벽에 부착된 초록색 비상구 표지판('비상구' 텍스트 정확함).",
      "hard_violations": [],
      "physics": "계단에 앉아 체중이 시트에 실려 있으며 두 발은 아래 단에 닿아 있음. 양손이 스마트폰을 안정적으로 받치고 있음."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "agree",
     "gap": 0.0,
     "per_model_winner": {
      "gemini-pro": "A",
      "openrouter:x-ai/grok-4.6": "A"
     },
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "reverse_normalized": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "엄격하게 요구된 카메라의 가파른 하이 앵글 뷰와 하향 프레이밍 지침을 정확히 구현하여 가장 높은 평가를 받음."
     },
     {
      "label": "A",
      "score": 4,
      "verdict_ko": "장소 레퍼런스와 표지판 텍스트는 완벽히 일치하나, 우선순위가 높은 카메라 하이 앵글 지침을 무시하고 정면 눈높이에서 촬영하여 감점됨."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "B",
      "direction": "시선은 양손에 든 스마트폰 화면을 향하고 있음.",
      "built_space": "여러 갈래로 뻗은 회색 콘크리트 계단 구조. 카메라는 프롬프트 지시대로 계단 위쪽에서 아래를 깊게 내려다보는 가파른 각도(하이 앵글)에 위치함.",
      "entities": "제호(짧은 흑발, 회색 셔츠 및 파란색 조끼 착용), 스마트폰(화면 켜짐). 우측 벽에 부착된 초록색 비상구 표지판(텍스트가 '배상구'로 오기됨).",
      "hard_violations": [],
      "physics": "피사체는 계단 중간에 안정적으로 앉아 있으며, 두 발은 아래쪽 계단 디딤판에 놓여 있음. 양손은 스마트폰을 단단히 쥐고 지탱함."
     },
     {
      "label": "A",
      "direction": "시선은 양손에 든 스마트폰 화면을 향하고 있음.",
      "built_space": "레퍼런스 이미지와 완벽히 일치하는 단조로운 구조의 콘크리트 계단실. 카메라는 지시된 하이 앵글 뷰를 따르지 않고 피사체와 비슷한 눈높이(정면)의 계단참에 위치함.",
      "entities": "제호(레퍼런스와 매우 흡사한 외모 및 의상), 스마트폰. 좌측 벽에 부착된 초록색 비상구 표지판('비상구' 텍스트 정확함).",
      "hard_violations": [],
      "physics": "계단에 앉아 체중이 시트에 실려 있으며 두 발은 아래 단에 닿아 있음. 양손이 스마트폰을 안정적으로 받치고 있음."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 1433,
     "B": 1257
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": false,
    "policy": 1
   }
  },
  "totals": {
   "A": 1433,
   "B": 1257
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 1429,
    "verdict_ko": "지시된 하이앵글 구도(위에서 내려다보는 시점)를 무시하고 레퍼런스의 정면 구도를 그대로 차용하여 프레이밍 조건(우선순위 2)을 위반했으나, 치명적인 구조적 오류가 없어 B보다 낫습니다."
   },
   {
    "label": "B",
    "score": 1250,
    "verdict_ko": "지시된 카메라 각도는 시도했으나, 우측 상단 계단의 불가능한 건축 기하학과 표지판의 불필요한 영문 텍스트 누출 등 하드 위반 요소가 발생해 탈락했습니다.  ★위반: [gemini-pro] 우측 상단 계단의 물리적으로 불가능한 구조 (원근감 붕괴) / [gemini-pro] 표지판에 지시되지 않은 영문 텍스트(EXIT) 노출"
   }
  ],
  "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/46575226-4c2f-4334-b16a-429b11371e24/episodes/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/images/background_chain/L02B01.png",
    "asset_id": "eb569ab2-420f-4d2e-b846-647e6a9d84f5",
    "role": "location_plate"
   },
   {
    "label": "CHARACTER REFERENCE — 제호: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:847626>",
    "asset_id": "7760e702-1e92-4b92-b945-00de7bc3d3c1",
    "role": "character_ref"
   },
   {
    "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.",
    "path": "<bytes:621038>",
    "asset_id": "34c4e322-0ded-425f-a7c2-2f126fa257d7",
    "role": "prop_ref"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "스마트폰 화면에 '조회하기' 버튼이 떠 있어야 한다는 지시에도 불구하고, 기기 화면이 켜지지 않은 채 완전히 검은색으로 묘사되어 있습니다.",
     "fix_en": "Illuminate the smartphone screen with a bright, glowing app interface featuring a distinct central button, viewed at a steep oblique angle that obscures exact text. Maintain Jeho, his pose, his hands, his clothing, and the background staircase exactly as they are.",
     "severity": "major",
     "observation_index": 0,
     "needs_ref_indices": [
      4
     ],
     "adds_missing_entity": false,
     "missing_entity_name": ""
    },
    {
     "issue_ko": "왼쪽 벽의 비상구 표지판에 지정된 '비상구' 문구 아래에 형태가 뭉개진 영문(또는 식별 불가능한 문자)이 임의로 추가되어 있습니다.",
     "fix_en": "Erase the garbled text below the word '비상구' on the green exit sign on the left wall, leaving only the clean white background of the sign beneath the main text. Maintain Jeho, his clothing, his pose, and all other background details exactly as they are.",
     "severity": "major",
     "observation_index": 1,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": ""
    }
   ],
   "observer_observations": [
    {
     "issue_ko": "스마트폰 화면에 '조회하기' 버튼이 떠 있어야 한다는 지시에도 불구하고, 기기 화면이 켜지지 않은 채 완전히 검은색으로 묘사되어 있습니다.",
     "severity": "major"
    },
    {
     "issue_ko": "왼쪽 벽의 비상구 표지판에 지정된 '비상구' 문구 아래에 형태가 뭉개진 영문(또는 식별 불가능한 문자)이 임의로 추가되어 있습니다.",
     "severity": "major"
    },
    {
     "issue_ko": "제호가 계단 한가운데가 아니라 아래쪽 단에 가깝게 앉아 있다.",
     "severity": "major"
    }
   ],
   "observer_models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ],
   "observer_counts": {
    "gemini-pro": 2,
    "openrouter:x-ai/grok-4.6": 1
   }
  },
  "fix_severity_skipped_count": 2,
  "fix_severity_skipped": [
   {
    "issue_ko": "스마트폰 화면에 '조회하기' 버튼이 떠 있어야 한다는 지시에도 불구하고, 기기 화면이 켜지지 않은 채 완전히 검은색으로 묘사되어 있습니다.",
    "fix_en": "Illuminate the smartphone screen with a bright, glowing app interface featuring a distinct central button, viewed at a steep oblique angle that obscures exact text. Maintain Jeho, his pose, his hands, his clothing, and the background staircase exactly as they are.",
    "severity": "major",
    "observation_index": 0,
    "needs_ref_indices": [
     4
    ],
    "adds_missing_entity": false,
    "missing_entity_name": ""
   },
   {
    "issue_ko": "왼쪽 벽의 비상구 표지판에 지정된 '비상구' 문구 아래에 형태가 뭉개진 영문(또는 식별 불가능한 문자)이 임의로 추가되어 있습니다.",
    "fix_en": "Erase the garbled text below the word '비상구' on the green exit sign on the left wall, leaving only the clean white background of the sign beneath the main text. Maintain Jeho, his clothing, his pose, and all other background details exactly as they are.",
    "severity": "major",
    "observation_index": 1,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": ""
   }
  ],
  "fix_skipped": true,
  "fix_skip_reason": "no_critical_issue",
  "shot_run_produced": true,
  "shot_run_uid": "06a8c2dd-d672-74c8-86af-a04cfe45cd8a",
  "bgfirst": {
   "bg_path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S2sh1__bgfirst_bg.png",
   "bg_asset_id": "3ba72b22-4091-4f09-979d-9dd7ed538b13",
   "bg_record_key": "S2sh1::bgfirst_bg",
   "chain_winner": true,
   "authority": "plate"
  },
  "ref_mode": "재투영 배경+콘티+엔티티 (2택1: 체인 승)",
  "share_plan": {
   "ref_plan": "background"
  }
 },
 "S2sh1::cine": {
  "applied": true,
  "fingerprint": "63246dc5f9dab2993f868d1859689d53720e9a9a6f3c5ddaaadbf4b1078a67f7",
  "model": "x-ai/grok-imagine-image-2.0",
  "pack": "20.202608141300",
  "source_file": "S2sh1_sel.png",
  "source_sha256": "cd33dfad01b19a3d07a0530a37f633affec050258d59f186275b9b3cb2adebb8",
  "file": "S2sh1_cine.png",
  "latency_ms": 11210
 },
 "S2sh6::signage": {
  "fp": "78172949e2ae73d8",
  "inscriptions": [
   {
    "surface_native": "스마트폰 화면",
    "text_native": "[안내] 불합격",
    "reason_ko": "현석이 씁쓸한 미소를 지으며 제호에게 보여주는 핸드폰 화면의 불합격 통지 문구입니다."
   }
  ]
 },
 "S2sh6": {
  "input_fingerprint": "560abd2cbdbd0301",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 현석이 씁쓸한 미소를 지으며 자신의 핸드폰 화면을 제호를 향해 뻗은 채 들고 있는 상태.\n\nLOCATION (lock): Inside the quiet school stairwell, with one student seated higher on the concrete steps and the other below him. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Jeho's seated head height, the static camera looks diagonally upward from behind and beside his foreground shoulder toward Hyunseok on the higher steps. Hyunseok leans into the middle-left frame with a bitter smile and extends his phone into the open space between them; its screen stays oblique to the lens while Jeho's attention shifts from the offered number to Hyunseok's face.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 제호 in the middle-right of the frame, foreground, looks toward 현석; 현석 in the middle-left of the frame, midground, reaches for space in front of 제호.\n- KEY BACKGROUND ELEMENTS: Hyunseok's phone (extended toward Jeho with 'DJH54' displayed) — The lit content face points toward Jeho and remains oblique rather than square to the filming camera, with the number still readable in perspective; used as Occupies the visual gap between Jeho's foreground shoulder and Hyunseok's expression; gray concrete steps (quiet and unoccupied immediately around the two students) — The steps rise behind Hyunseok and establish his higher position relative to Jeho; used as Maintains the diagonal separation and shared spatial axis between them.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral stairwell ambience keeps the gray setting subdued, with restrained contrast across both faces and the offered screen.\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): Take the dry gray concrete texture, quiet stairwell lighting, and the seated student's uniform and placement from the reference. Exclude the solitary composition and downward phone-checking pose; add the second student higher on the stairs extending his own phone.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok holds out his smartphone with the exam number DJH54 displayed, while his camera remains in his possession after taking Jeho's picture. Jeho still holds his own smartphone.\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 현석 right now, so 현석'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 현석: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else. Nothing else of them needs to be in shot; the framing stays on the object.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 현석 (한국인 남성, 10대 후반의 젊은 얼굴, 헝클어진 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 스마트폰 화면: \"[안내] 불합격\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 현석이 씁쓸한 미소를 지으며 자신의 핸드폰 화면을 제호를 향해 뻗은 채 들고 있는 상태.\n\nLOCATION (lock): Inside the quiet school stairwell, with one student seated higher on the concrete steps and the other below him. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Jeho's seated head height, the static camera looks diagonally upward from behind and beside his foreground shoulder toward Hyunseok on the higher steps. Hyunseok leans into the middle-left frame with a bitter smile and extends his phone into the open space between them; its screen stays oblique to the lens while Jeho's attention shifts from the offered number to Hyunseok's face.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 제호 in the middle-right of the frame, foreground, looks toward 현석; 현석 in the middle-left of the frame, midground, reaches for space in front of 제호.\n- KEY BACKGROUND ELEMENTS: Hyunseok's phone (extended toward Jeho with 'DJH54' displayed) — The lit content face points toward Jeho and remains oblique rather than square to the filming camera, with the number still readable in perspective; used as Occupies the visual gap between Jeho's foreground shoulder and Hyunseok's expression; gray concrete steps (quiet and unoccupied immediately around the two students) — The steps rise behind Hyunseok and establish his higher position relative to Jeho; used as Maintains the diagonal separation and shared spatial axis between them.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral stairwell ambience keeps the gray setting subdued, with restrained contrast across both faces and the offered screen.\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): Take the dry gray concrete texture, quiet stairwell lighting, and the seated student's uniform and placement from the reference. Exclude the solitary composition and downward phone-checking pose; add the second student higher on the stairs extending his own phone.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok holds out his smartphone with the exam number DJH54 displayed, while his camera remains in his possession after taking Jeho's picture. Jeho still holds his own smartphone.\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 현석 right now, so 현석'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 현석: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else. Nothing else of them needs to be in shot; the framing stays on the object.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 현석 (한국인 남성, 10대 후반의 젊은 얼굴, 헝클어진 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 스마트폰 화면: \"[안내] 불합격\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 현석이 씁쓸한 미소를 지으며 자신의 핸드폰 화면을 제호를 향해 뻗은 채 들고 있는 상태.\n\nLOCATION (lock): Inside the quiet school stairwell, with one student seated higher on the concrete steps and the other below him. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: At Jeho's seated head height, the static camera looks diagonally upward from behind and beside his foreground shoulder toward Hyunseok on the higher steps. Hyunseok leans into the middle-left frame with a bitter smile and extends his phone into the open space between them; its screen stays oblique to the lens while Jeho's attention shifts from the offered number to Hyunseok's face.\n- FRAMING SCALE: medium shot\n- FRAME LAYOUT: 제호 in the middle-right of the frame, foreground, looks toward 현석; 현석 in the middle-left of the frame, midground, reaches for space in front of 제호.\n- KEY BACKGROUND ELEMENTS: Hyunseok's phone (extended toward Jeho with 'DJH54' displayed) — The lit content face points toward Jeho and remains oblique rather than square to the filming camera, with the number still readable in perspective; used as Occupies the visual gap between Jeho's foreground shoulder and Hyunseok's expression; gray concrete steps (quiet and unoccupied immediately around the two students) — The steps rise behind Hyunseok and establish his higher position relative to Jeho; used as Maintains the diagonal separation and shared spatial axis between them.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral stairwell ambience keeps the gray setting subdued, with restrained contrast across both faces and the offered screen.\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): Take the dry gray concrete texture, quiet stairwell lighting, and the seated student's uniform and placement from the reference. Exclude the solitary composition and downward phone-checking pose; add the second student higher on the stairs extending his own phone.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Hyunseok holds out his smartphone with the exam number DJH54 displayed, while his camera remains in his possession after taking Jeho's picture. Jeho still holds his own smartphone.\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 현석 right now, so 현석'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 현석: its size, build, skin, age, grooming, sleeve and anything worn on it belong to that person and to no one else. Nothing else of them needs to be in shot; the framing stays on the object.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 현석 (한국인 남성, 10대 후반의 젊은 얼굴, 헝클어진 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 스마트폰 화면: \"[안내] 불합격\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, but reframe like a film director choosing a different setup — a different angle, height, distance or foreground layer from the obvious one. For this candidate only, treat the CAMERA & FRAME contract above as a baseline to depart from, not a lock; the references and the shot text fix WHAT is seen, not WHERE the camera stands. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "B",
    "direction": "현석이 제호를 향해 핸드폰을 내밀고 있으며, 제호의 시선은 현석의 얼굴을 향하고 있음.",
    "built_space": "카메라가 제호의 어깨 뒤에서 위쪽 계단의 현석을 올려다보는 정확한 구도의 콘크리트 계단실임.",
    "entities": "두 학생의 외형과 착장이 레퍼런스와 일치하며, 현석의 폰 화면에 '[안내] 불합격'과 'DJH54'가 모두 정확하게 적혀 있음.",
    "hard_violations": [],
    "physics": "현석은 한쪽 발을 위쪽 계단에 딛고 안정적으로 몸을 굽히고 있으며, 제호는 아래 계단에 자연스럽게 앉아 있음."
   },
   {
    "label": "A",
    "direction": "현석이 제호에게 폰을 내밀고 제호가 현석을 바라봄.",
    "built_space": "콘크리트 계단실이나, 프레이밍이 지시된 어깨 뒤 앵글이 아닌 평면적인 투샷에 가까움.",
    "entities": "인물 외형은 레퍼런스와 대체로 일치함. 폰 화면에 '[안내] 불합격'이 있으나 그 아래에 해독 불가한 문자가 추가됨.",
    "hard_violations": [
     "현석이 핸드폰을 쥐고 있는 손의 손가락이 6개임 (화면 위 엄지, 기기 우측면 검지, 기기 좌측면 뒤로 손가락 4개가 보임)."
    ],
    "physics": "두 인물 모두 계단에 몸을 지탱하여 앉아 있음."
   }
  ],
  "gq": {
   "route": "agree",
   "gap": 0.0,
   "per_model_winner": {
    "gemini-pro": "B",
    "openrouter:x-ai/grok-4.6": "B"
   },
   "models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ]
  },
  "totals": {
   "B": 8,
   "A": 3
  },
  "selected": "B",
  "ranking": [
   "B",
   "A"
  ],
  "verdicts": [
   {
    "label": "B",
    "score": 8,
    "verdict_ko": "요구된 어깨 너머 대각선 앵글과 숏 사이즈를 완벽히 구현했으며, 화면 텍스트('[안내] 불합격', 'DJH54')도 매우 정확하게 렌더링한 훌륭한 결과물입니다."
   },
   {
    "label": "A",
    "score": 3,
    "verdict_ko": "제시된 카메라 구도(제호 어깨 뒤 대각선 앵글)를 따르지 않았으며, 현석의 손가락 구조에 심각한 오류가 있습니다."
   }
  ],
  "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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S2sh1_sel.png",
    "asset_id": "049ebf10-3e23-4701-8920-c92f5133646b",
    "role": "prev_still"
   },
   {
    "label": "CHARACTER REFERENCE — 현석: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:973165>",
    "asset_id": "37299bdd-ae7a-4ecb-a36a-05c753ff9d51",
    "role": "character_ref"
   },
   {
    "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.",
    "path": "<bytes:621038>",
    "asset_id": "34c4e322-0ded-425f-a7c2-2f126fa257d7",
    "role": "prop_ref"
   }
  ],
  "critique": {
   "issues": [
    {
     "issue_ko": "현석이 스마트폰을 쥐고 있는 손의 손가락 형태가 기형적이며 화면을 관통하고 있습니다.",
     "fix_en": "Redraw Hyunseok's hand holding the phone to be anatomically correct, with distinct fingers naturally gripping the edges of the device and the thumb resting outside the lit screen area; preserve the characters, their poses, clothing, lighting, background stairs, and framing exactly as they are.",
     "severity": "critical",
     "observation_index": 0,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "현석이 내민 스마트폰 화면이 제호를 향하지 않고 카메라 렌즈를 정면으로 향하고 있습니다.",
     "fix_en": "Rotate the extended phone in Hyunseok's hand so its lit screen is angled obliquely toward the character on the right, foreshortening the display surface in perspective rather than facing the camera squarely; preserve the characters' identities, poses, expressions, clothing, the stairwell setting, lighting, and framing exactly as they are.",
     "severity": "critical",
     "observation_index": 1,
     "needs_ref_indices": [
      3
     ],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "스마트폰 화면에 표시되어야 할 'DJH54' 텍스트의 앞부분이 뭉개져 잘못된 문자로 표기되었습니다.",
     "fix_en": "Apply a shallow depth of field or slight motion blur to the phone screen so the mangled text becomes less legible, staging the printed figures out of sharp focus; preserve the characters, their poses, clothing, lighting, background stairs, and framing exactly as they are.",
     "severity": "major",
     "observation_index": 2,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "캐릭터 레퍼런스 이미지에서 현석이 착용한 모자, 턱에 걸친 마스크, 크로스백이 모두 누락되었습니다.",
     "fix_en": "Add a dark cap to Hyunseok's head, a mask resting under his chin, and a cross bag strap over his shoulder, matching his character reference; preserve his face, current pose, the other character, the phone, setting, lighting, and framing exactly.",
     "severity": "major",
     "observation_index": 3,
     "needs_ref_indices": [
      2
     ],
     "adds_missing_entity": true,
     "missing_entity_name": "cap, mask, cross bag strap",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "왼쪽 현석의 얼굴과 머리가 캐릭터 레퍼런스와 전혀 다르다.",
     "fix_en": "Adjust Hyunseok's face and hairstyle on the left to seamlessly match the provided character reference of the young man; preserve his current expression, pose, the other character, clothing, setting, and framing exactly as they are.",
     "severity": "major",
     "observation_index": 4,
     "needs_ref_indices": [
      2
     ],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "왼쪽 현석이 높은 계단에 앉아 있지 않고 서서 기대어 있다.",
     "fix_en": "Adjust Hyunseok's legs and hips in place to bend more deeply, suggesting he is resting on the stairs behind him, without changing the overall camera position or framing; preserve his identity, expression, clothing, the other character, and lighting exactly.",
     "severity": "major",
     "observation_index": 6,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": true,
     "unfixable": false
    },
    {
     "issue_ko": "왼쪽 현석이 씁쓸한 미소가 아니라 환하게 웃고 있다.",
     "fix_en": "Soften Hyunseok's expression into a restrained, bitter smile with less exposed teeth; preserve his identity, overall pose, the other character, clothing, the held phone, setting, and framing exactly as they are.",
     "severity": "major",
     "observation_index": 7,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    },
    {
     "issue_ko": "오른쪽 제호의 얼굴이 이전 샷 인물의 얼굴과 같다.",
     "fix_en": "Replace Jeho's face in the right foreground with a different young Korean male face so he no longer resembles the person from the previous shot still; preserve his pose, expression, clothing, the left character, the phone, setting, and framing exactly as they are.",
     "severity": "major",
     "observation_index": 9,
     "needs_ref_indices": [],
     "adds_missing_entity": false,
     "missing_entity_name": "",
     "needs_regeneration": false,
     "unfixable": false
    }
   ],
   "observer_observations": [
    {
     "issue_ko": "현석이 스마트폰을 쥐고 있는 손의 손가락 형태가 기형적이며 화면을 관통하고 있습니다.",
     "severity": "critical"
    },
    {
     "issue_ko": "현석이 내민 스마트폰 화면이 제호를 향하지 않고 카메라 렌즈를 정면으로 향하고 있습니다.",
     "severity": "critical"
    },
    {
     "issue_ko": "스마트폰 화면에 표시되어야 할 'DJH54' 텍스트의 앞부분이 뭉개져 잘못된 문자로 표기되었습니다.",
     "severity": "major"
    },
    {
     "issue_ko": "캐릭터 레퍼런스 이미지에서 현석이 착용한 모자, 턱에 걸친 마스크, 크로스백이 모두 누락되었습니다.",
     "severity": "major"
    },
    {
     "issue_ko": "왼쪽 현석의 얼굴과 머리가 캐릭터 레퍼런스와 전혀 다르다.",
     "severity": "major"
    },
    {
     "issue_ko": "왼쪽 현석이 레퍼런스 의상인 모자와 가방을 착용하지 않았다.",
     "severity": "major"
    },
    {
     "issue_ko": "왼쪽 현석이 높은 계단에 앉아 있지 않고 서서 기대어 있다.",
     "severity": "major"
    },
    {
     "issue_ko": "왼쪽 현석이 씁쓸한 미소가 아니라 환하게 웃고 있다.",
     "severity": "major"
    },
    {
     "issue_ko": "가운데 핸드폰 화면이 렌즈에 비스듬하지 않고 거의 정면을 향한다.",
     "severity": "major"
    },
    {
     "issue_ko": "오른쪽 제호의 얼굴이 이전 샷 인물의 얼굴과 같다.",
     "severity": "major"
    }
   ],
   "observer_models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ],
   "observer_counts": {
    "gemini-pro": 4,
    "openrouter:x-ai/grok-4.6": 6
   }
  },
  "fix_severity_skipped_count": 6,
  "fix_severity_skipped": [
   {
    "issue_ko": "스마트폰 화면에 표시되어야 할 'DJH54' 텍스트의 앞부분이 뭉개져 잘못된 문자로 표기되었습니다.",
    "fix_en": "Apply a shallow depth of field or slight motion blur to the phone screen so the mangled text becomes less legible, staging the printed figures out of sharp focus; preserve the characters, their poses, clothing, lighting, background stairs, and framing exactly as they are.",
    "severity": "major",
    "observation_index": 2,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "캐릭터 레퍼런스 이미지에서 현석이 착용한 모자, 턱에 걸친 마스크, 크로스백이 모두 누락되었습니다.",
    "fix_en": "Add a dark cap to Hyunseok's head, a mask resting under his chin, and a cross bag strap over his shoulder, matching his character reference; preserve his face, current pose, the other character, the phone, setting, lighting, and framing exactly.",
    "severity": "major",
    "observation_index": 3,
    "needs_ref_indices": [
     2
    ],
    "adds_missing_entity": true,
    "missing_entity_name": "cap, mask, cross bag strap",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "왼쪽 현석의 얼굴과 머리가 캐릭터 레퍼런스와 전혀 다르다.",
    "fix_en": "Adjust Hyunseok's face and hairstyle on the left to seamlessly match the provided character reference of the young man; preserve his current expression, pose, the other character, clothing, setting, and framing exactly as they are.",
    "severity": "major",
    "observation_index": 4,
    "needs_ref_indices": [
     2
    ],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "왼쪽 현석이 높은 계단에 앉아 있지 않고 서서 기대어 있다.",
    "fix_en": "Adjust Hyunseok's legs and hips in place to bend more deeply, suggesting he is resting on the stairs behind him, without changing the overall camera position or framing; preserve his identity, expression, clothing, the other character, and lighting exactly.",
    "severity": "major",
    "observation_index": 6,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": true,
    "unfixable": false
   },
   {
    "issue_ko": "왼쪽 현석이 씁쓸한 미소가 아니라 환하게 웃고 있다.",
    "fix_en": "Soften Hyunseok's expression into a restrained, bitter smile with less exposed teeth; preserve his identity, overall pose, the other character, clothing, the held phone, setting, and framing exactly as they are.",
    "severity": "major",
    "observation_index": 7,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   },
   {
    "issue_ko": "오른쪽 제호의 얼굴이 이전 샷 인물의 얼굴과 같다.",
    "fix_en": "Replace Jeho's face in the right foreground with a different young Korean male face so he no longer resembles the person from the previous shot still; preserve his pose, expression, clothing, the left character, the phone, setting, and framing exactly as they are.",
    "severity": "major",
    "observation_index": 9,
    "needs_ref_indices": [],
    "adds_missing_entity": false,
    "missing_entity_name": "",
    "needs_regeneration": false,
    "unfixable": false
   }
  ],
  "repair_mode": "edit",
  "fix_ref_gate": {
   "mode": "gated",
   "available": 3,
   "kept": 1,
   "requested_indices": [
    3
   ],
   "kept_labels": [
    "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly."
   ]
  },
  "fix_ref_count": 2,
  "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redraw Hyunseok's hand holding the phone to be anatomically correct, with distinct fingers naturally gripping the edges of the device and the thumb resting outside the lit screen area; preserve the characters, their poses, clothing, lighting, background stairs, and framing exactly as they are.\n- Rotate the extended phone in Hyunseok's hand so its lit screen is angled obliquely toward the character on the right, foreshortening the display surface in perspective rather than facing the camera squarely; preserve the characters' identities, poses, expressions, clothing, the stairwell setting, lighting, and framing exactly as they are.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. Real-world writing that belongs\nto the scene — signs, banners, documents — stays exactly as the\noriginal shows it; do not add captions, subtitles, watermarks or\nany overlay text.",
  "fix_rejudge": {
   "forward_raw": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "요구된 텍스트인 '[안내] 불합격'과 수험번호 'DJH54'를 화면에 매우 정확하게 구현해냈으며, 전체적인 구도와 인물들의 상호작용도 프롬프트를 충실히 따릅니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "스마트폰 소품 레퍼런스의 배경화면을 유지하려 했으나, 가장 중요한 텍스트가 '[안녕!!] 불합격' 등으로 잘못 출력되어 지시사항을 위반했습니다."
     }
    ],
    "readings": [
     {
      "label": "A",
      "direction": "현석은 제호를 바라보며 미소 짓고 있고, 제호는 현석의 얼굴을 올려다보고 있습니다. 현석이 뻗은 핸드폰의 화면은 제호를 향하면서도 카메라에서 비스듬히 보이도록 위치해 있습니다.",
      "built_space": "레퍼런스와 동일한 학교 계단실입니다. 제호는 아래쪽 계단에 앉아있고, 현석은 그보다 한 단 위 계단에 서서 제호를 향해 몸을 기울이고 있습니다.",
      "entities": "현석의 얼굴과 복장, 체형이 캐릭터 레퍼런스와 일치합니다. 제호 역시 이전 샷 레퍼런스의 모습과 일치합니다. 현석의 스마트폰 화면에는 프롬프트가 요구한 '[안내] 불합격' 및 'DJH54' 텍스트가 흰 배경 위에 정확히 렌더링되어 있습니다.",
      "hard_violations": [],
      "physics": "현석은 계단에 발을 딛고 안정적으로 서서 몸을 기울이고 있으며, 오른손으로 스마트폰을 자연스럽게 쥐고 있습니다. 제호는 계단에 엉덩이를 대고 앉아 자신의 스마트폰을 쥐고 있습니다. 지지되지 않거나 불가능한 자세는 없습니다."
     },
     {
      "label": "B",
      "direction": "현석은 제호를 향해 미소 짓고 있고, 제호는 현석의 얼굴을 응시하고 있습니다. 핸드폰 화면은 제호를 향하며 카메라 측면으로 노출됩니다.",
      "built_space": "레퍼런스와 동일한 조용한 학교 계단실이며, 인물들의 위치 관계(위/아래)도 정확하게 구현되어 있습니다.",
      "entities": "두 인물의 외형은 레퍼런스와 잘 일치합니다. 하지만 현석이 들고 있는 스마트폰 화면에 요구된 '[안내]' 대신 '[안녕!!]'과 유사한 잘못된 텍스트가 출력되어 있으며, 수험번호 역시 일치하지 않습니다.",
      "hard_violations": [],
      "physics": "인물들의 자세, 계단과의 접촉면, 스마트폰을 쥐고 있는 손의 형태 모두 물리적으로 자연스러우며 중력을 거스르거나 붕 떠있는 요소는 없습니다."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "agree",
     "gap": 0.0,
     "per_model_winner": {
      "gemini-pro": "A",
      "openrouter:x-ai/grok-4.6": "A"
     },
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "forward_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "요구된 텍스트인 '[안내] 불합격'과 수험번호 'DJH54'를 화면에 매우 정확하게 구현해냈으며, 전체적인 구도와 인물들의 상호작용도 프롬프트를 충실히 따릅니다."
     },
     {
      "label": "B",
      "score": 6,
      "verdict_ko": "스마트폰 소품 레퍼런스의 배경화면을 유지하려 했으나, 가장 중요한 텍스트가 '[안녕!!] 불합격' 등으로 잘못 출력되어 지시사항을 위반했습니다."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "A",
      "direction": "현석은 제호를 바라보며 미소 짓고 있고, 제호는 현석의 얼굴을 올려다보고 있습니다. 현석이 뻗은 핸드폰의 화면은 제호를 향하면서도 카메라에서 비스듬히 보이도록 위치해 있습니다.",
      "built_space": "레퍼런스와 동일한 학교 계단실입니다. 제호는 아래쪽 계단에 앉아있고, 현석은 그보다 한 단 위 계단에 서서 제호를 향해 몸을 기울이고 있습니다.",
      "entities": "현석의 얼굴과 복장, 체형이 캐릭터 레퍼런스와 일치합니다. 제호 역시 이전 샷 레퍼런스의 모습과 일치합니다. 현석의 스마트폰 화면에는 프롬프트가 요구한 '[안내] 불합격' 및 'DJH54' 텍스트가 흰 배경 위에 정확히 렌더링되어 있습니다.",
      "hard_violations": [],
      "physics": "현석은 계단에 발을 딛고 안정적으로 서서 몸을 기울이고 있으며, 오른손으로 스마트폰을 자연스럽게 쥐고 있습니다. 제호는 계단에 엉덩이를 대고 앉아 자신의 스마트폰을 쥐고 있습니다. 지지되지 않거나 불가능한 자세는 없습니다."
     },
     {
      "label": "B",
      "direction": "현석은 제호를 향해 미소 짓고 있고, 제호는 현석의 얼굴을 응시하고 있습니다. 핸드폰 화면은 제호를 향하며 카메라 측면으로 노출됩니다.",
      "built_space": "레퍼런스와 동일한 조용한 학교 계단실이며, 인물들의 위치 관계(위/아래)도 정확하게 구현되어 있습니다.",
      "entities": "두 인물의 외형은 레퍼런스와 잘 일치합니다. 하지만 현석이 들고 있는 스마트폰 화면에 요구된 '[안내]' 대신 '[안녕!!]'과 유사한 잘못된 텍스트가 출력되어 있으며, 수험번호 역시 일치하지 않습니다.",
      "hard_violations": [],
      "physics": "인물들의 자세, 계단과의 접촉면, 스마트폰을 쥐고 있는 손의 형태 모두 물리적으로 자연스러우며 중력을 거스르거나 붕 떠있는 요소는 없습니다."
     }
    ]
   },
   "reverse_raw": {
    "winner": "B",
    "ranking": [
     "B",
     "A"
    ],
    "verdicts": [
     {
      "label": "B",
      "score": 9,
      "verdict_ko": "지정된 텍스트('[안내] 불합격' 및 'DJH54')를 화면에 완벽하게 렌더링했으며, 프레이밍과 캐릭터의 배치 및 표정 지시를 훌륭하게 따랐습니다."
     },
     {
      "label": "A",
      "score": 7,
      "verdict_ko": "구도와 인물의 배치는 좋으나 스마트폰 화면의 텍스트가 지시사항과 다르게 변형 및 오탈자('[안내]!!', 'B/A/S4' 등)가 발생했습니다."
     }
    ],
    "readings": [
     {
      "label": "B",
      "direction": "현석과 제호의 시선이 자연스럽게 교차하며, 현석이 내민 스마트폰 화면이 제호를 향해 비스듬히 제시되어 있습니다.",
      "built_space": "레퍼런스 이미지와 동일한 구조의 계단실이며, 현석은 높은 계단에, 제호는 낮은 계단에 위치하여 지정된 구도를 잘 만족합니다.",
      "entities": "현석은 레퍼런스의 외모, 헤어스타일, 의상과 일치하며, 스마트폰 텍스트 '[안내] 불합격'과 'DJH54'가 정확히 출력되었습니다. 제호 역시 이전 컷의 인물 형태를 유지합니다.",
      "hard_violations": [],
      "physics": "현석은 계단에 발을 디딘 채 안정적인 자세로 서 있고, 제호는 계단에 앉아 있으며 양쪽 모두 스마트폰을 쥔 손의 묘사와 물리적 지지가 자연스럽습니다."
     },
     {
      "label": "A",
      "direction": "현석의 시선과 내민 스마트폰의 방향이 제호를 향하고 있으며, 제호도 현석을 바라보고 있습니다.",
      "built_space": "계단실의 구조와 두 인물의 상하 위치 배치가 레퍼런스 및 프롬프트의 지시와 일치합니다.",
      "entities": "인물들의 외모와 의상은 잘 구현되었으나, 현석의 스마트폰 화면 텍스트에 요구되지 않은 기호와 오탈자가 포함되어 지정된 문구를 정확히 반영하지 못했습니다.",
      "hard_violations": [],
      "physics": "현석이 계단을 밟고 몸을 기울인 자세와 제호가 앉아 있는 자세 모두 지지 기반이 명확하며 폰을 쥔 손의 형태도 물리적으로 타당합니다."
     }
    ],
    "all_candidates_fail": false,
    "gq": {
     "route": "agree",
     "gap": 0.0,
     "per_model_winner": {
      "gemini-pro": "B",
      "openrouter:x-ai/grok-4.6": "B"
     },
     "models": [
      "gemini-pro",
      "openrouter:x-ai/grok-4.6"
     ]
    }
   },
   "reverse_normalized": {
    "winner": "A",
    "ranking": [
     "A",
     "B"
    ],
    "verdicts": [
     {
      "label": "A",
      "score": 9,
      "verdict_ko": "지정된 텍스트('[안내] 불합격' 및 'DJH54')를 화면에 완벽하게 렌더링했으며, 프레이밍과 캐릭터의 배치 및 표정 지시를 훌륭하게 따랐습니다."
     },
     {
      "label": "B",
      "score": 7,
      "verdict_ko": "구도와 인물의 배치는 좋으나 스마트폰 화면의 텍스트가 지시사항과 다르게 변형 및 오탈자('[안내]!!', 'B/A/S4' 등)가 발생했습니다."
     }
    ],
    "all_candidates_fail": false,
    "readings": [
     {
      "label": "A",
      "direction": "현석과 제호의 시선이 자연스럽게 교차하며, 현석이 내민 스마트폰 화면이 제호를 향해 비스듬히 제시되어 있습니다.",
      "built_space": "레퍼런스 이미지와 동일한 구조의 계단실이며, 현석은 높은 계단에, 제호는 낮은 계단에 위치하여 지정된 구도를 잘 만족합니다.",
      "entities": "현석은 레퍼런스의 외모, 헤어스타일, 의상과 일치하며, 스마트폰 텍스트 '[안내] 불합격'과 'DJH54'가 정확히 출력되었습니다. 제호 역시 이전 컷의 인물 형태를 유지합니다.",
      "hard_violations": [],
      "physics": "현석은 계단에 발을 디딘 채 안정적인 자세로 서 있고, 제호는 계단에 앉아 있으며 양쪽 모두 스마트폰을 쥔 손의 묘사와 물리적 지지가 자연스럽습니다."
     },
     {
      "label": "B",
      "direction": "현석의 시선과 내민 스마트폰의 방향이 제호를 향하고 있으며, 제호도 현석을 바라보고 있습니다.",
      "built_space": "계단실의 구조와 두 인물의 상하 위치 배치가 레퍼런스 및 프롬프트의 지시와 일치합니다.",
      "entities": "인물들의 외모와 의상은 잘 구현되었으나, 현석의 스마트폰 화면 텍스트에 요구되지 않은 기호와 오탈자가 포함되어 지정된 문구를 정확히 반영하지 못했습니다.",
      "hard_violations": [],
      "physics": "현석이 계단을 밟고 몸을 기울인 자세와 제호가 앉아 있는 자세 모두 지지 기반이 명확하며 폰을 쥔 손의 형태도 물리적으로 타당합니다."
     }
    ]
   },
   "combined": {
    "totals": {
     "A": 18,
     "B": 13
    },
    "ranking": [
     "A",
     "B"
    ],
    "agreement": true,
    "policy": 1
   },
   "winner": "A",
   "fix_won": false,
   "all_candidates_fail": false,
   "policy": "fix_rejudge_v2_orig_on_tie"
  },
  "shot_run_produced": true,
  "shot_run_uid": "06a8c311-6a67-7792-b8bb-a4e2f9c49c2e",
  "ref_mode": "prev+엔티티",
  "share_plan": {
   "ref_plan": "prev",
   "prev_anchor_tag": "S2sh1"
  }
 },
 "S2sh6::cine": {
  "applied": true,
  "fingerprint": "8862a9b6e7df3efe7c231e85ac26bd60b20ffb9991bd4e5cc6ab47f14ef12466",
  "model": "x-ai/grok-imagine-image-2.0",
  "pack": "20.202608141300",
  "source_file": "S2sh6_sel.png",
  "source_sha256": "7b12b4fec7d2fd3dfc7c72a4f2f966e4c822faa03b516f4ac7013364e4ed884d",
  "file": "S2sh6_cine.png",
  "latency_ms": 12583
 },
 "S2sh8::signage": {
  "fp": "9f19394f7038b139",
  "inscriptions": [
   {
    "surface_native": "스마트폰 화면",
    "text_native": "징계위원회 소집 통보",
    "reason_ko": "제호가 핸드폰 화면을 보고 충격을 받은 구체적인 이유인 학교 징계 통보 내용을 화면에 표시하여 연출하기 위함입니다."
   }
  ]
 },
 "S2sh8": {
  "input_fingerprint": "49dd4b55ab4e32fb",
  "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 핸드폰 화면을 바라보는 제호의 동공이 충격으로 크게 확장된 찰나.\n\nLOCATION (lock): Inside the school’s gray concrete stairwell, on the steps where the two students compare their phone screens. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The lateral track completes just above Jeho's eye line at a front three-quarter angle, resolving into the explicitly extreme close view of one eye as his pupil expands. His lowered gaze remains fixed on the phone below the frame edge, while a narrow trace of his rigid brow and cheek preserves the direction of the downward look.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 제호 in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: edge of the phone (held just below Jeho's eye line with the matching number displayed) — Only a soft lower-frame edge of the content face remains visible, angled upward toward Jeho rather than toward camera; used as Keeps the cause of Jeho's shock at the frame boundary while focus isolates his expanding pupil.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient stairwell light is controlled to keep the pupil, strained eyelid, and rigid brow clearly readable without heightened color.\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): Take the gray concrete stairwell, midday light, and the anxious student's uniform and seated continuity from the reference. Exclude the wide stair composition and any other student from this tight reaction shot, and do not force the reference phone into the visible frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Jeho's smartphone now displays DJH54, matching the number already shown on Hyunseok's phone. Hyunseok continues to have his camera with him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 스마트폰 화면: \"징계위원회 소집 통보\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
  "roll_prompts": {
   "A": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 핸드폰 화면을 바라보는 제호의 동공이 충격으로 크게 확장된 찰나.\n\nLOCATION (lock): Inside the school’s gray concrete stairwell, on the steps where the two students compare their phone screens. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The lateral track completes just above Jeho's eye line at a front three-quarter angle, resolving into the explicitly extreme close view of one eye as his pupil expands. His lowered gaze remains fixed on the phone below the frame edge, while a narrow trace of his rigid brow and cheek preserves the direction of the downward look.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 제호 in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: edge of the phone (held just below Jeho's eye line with the matching number displayed) — Only a soft lower-frame edge of the content face remains visible, angled upward toward Jeho rather than toward camera; used as Keeps the cause of Jeho's shock at the frame boundary while focus isolates his expanding pupil.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient stairwell light is controlled to keep the pupil, strained eyelid, and rigid brow clearly readable without heightened color.\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): Take the gray concrete stairwell, midday light, and the anxious student's uniform and seated continuity from the reference. Exclude the wide stair composition and any other student from this tight reaction shot, and do not force the reference phone into the visible frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Jeho's smartphone now displays DJH54, matching the number already shown on Hyunseok's phone. Hyunseok continues to have his camera with him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 스마트폰 화면: \"징계위원회 소집 통보\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.",
   "B": "Create ONE FINAL photorealistic live-action film still of the moment below — future Seoul, South Korea, beginning in 2045; all people are Korean unless stated. TIME OF DAY (lock): midday.\n\nSHOT TEXT (authoritative, Korean): 핸드폰 화면을 바라보는 제호의 동공이 충격으로 크게 확장된 찰나.\n\nLOCATION (lock): Inside the school’s gray concrete stairwell, on the steps where the two students compare their phone screens. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nCAMERA & FRAME (follow exactly — this is the composition authority for this still; the reference images supply identity and place, never the framing):\n- CAMERA: The lateral track completes just above Jeho's eye line at a front three-quarter angle, resolving into the explicitly extreme close view of one eye as his pupil expands. His lowered gaze remains fixed on the phone below the frame edge, while a narrow trace of his rigid brow and cheek preserves the direction of the downward look.\n- FRAMING SCALE: close-up\n- FRAME LAYOUT: 제호 in the middle-center of the frame, foreground.\n- KEY BACKGROUND ELEMENTS: edge of the phone (held just below Jeho's eye line with the matching number displayed) — Only a soft lower-frame edge of the content face remains visible, angled upward toward Jeho rather than toward camera; used as Keeps the cause of Jeho's shock at the frame boundary while focus isolates his expanding pupil.\nCompose the frame exactly as specified above — camera angle, subject scale and screen placement. Keep true physical scale between people and background elements: fixtures and distant objects occupy only the small screen area their real size and distance dictate; never enlarge a background object into a foreground presence.\n\nLIGHTING & MOOD (these govern light, color and surface rendering):\n- LIGHTING & MOOD: Neutral ambient stairwell light is controlled to keep the pupil, strained eyelid, and rigid brow clearly readable without heightened color.\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): Take the gray concrete stairwell, midday light, and the anxious student's uniform and seated continuity from the reference. Exclude the wide stair composition and any other student from this tight reaction shot, and do not force the reference phone into the visible frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nNATURAL PERFORMANCE (default only — every explicit direction above\nwins): when no pose contract, immobility contract or shot-text\ndirection says otherwise, people read as alive in mid-moment —\nbelievable weight shift, hands engaged with something real, gaze on\nthe target the SHOT TEXT implies; avoid a stiff attention stance (feet\ntogether, arms hanging straight down) and a blank stare into the lens.\nIf the SHOT TEXT or any POSE/IMMOBILE contract above stages stillness,\ndeath, sleep, unconsciousness, restraint, drill or an explicit\ndirect-to-camera look, follow THAT exactly — this clause never\noverrides it.\n\nDRAWN MARKS KEEP THEIR SHAPE: any mark the text describes as drawn,\npainted or traced (a circle, a line, a symbol) is a STROKE sitting on\nthe surface — an outline whose interior still shows the underlying\nsurface (wall, skin, paper). Render its stated shape faithfully: a\ndrawn circle stays an open ring of brush-width, never filled into a\nsolid disc, unless the text explicitly says it is filled.\n\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): Jeho's smartphone now displays DJH54, matching the number already shown on Hyunseok's phone. Hyunseok continues to have his camera with him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 제호 (한국인 남성, 10대 후반의 젊은 얼굴, 단정히 빗은 짧은 머리, 자연스러운 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\nSIGNAGE TEXT (authoritative — the scene itself calls for these inscriptions; render each on its surface as period-real physical lettering in the native script, exactly as written; add no other readable text anywhere):\n- 스마트폰 화면: \"징계위원회 소집 통보\"\n\nReal-world writing that this scene itself calls for — shop signs,\nbanners, posters, documents, vehicle or building markings — may appear,\nwritten naturally in the language this place and era establish; never\nsubstitute another language's lettering. Only writing the scene calls\nfor: do not invent signage or lettering the shot and location text do\nnot imply, and never print the shot description itself onto the image.\nNo captions, subtitles, watermarks, logos or any overlay text floating\noutside the physical scene. Where exact printed figures would decide\nthe shot — a banknote's denomination, a document's numbers — stage that\nface out of legibility (a hand across it, an oblique angle, shallow\nfocus): printed figures do not render to instruction.\n\nCOMPOSITION VARIATION (this candidate only): keep the place, the people and the moment exactly as specified, but reframe like a film director choosing a different setup — a different angle, height, distance or foreground layer from the obvious one. For this candidate only, treat the CAMERA & FRAME contract above as a baseline to depart from, not a lock; the references and the shot text fix WHAT is seen, not WHERE the camera stands. Stay true to any explicit framing the SHOT TEXT itself dictates."
  },
  "shot_run_spend_attempt_count": 1,
  "readings": [
   {
    "label": "A",
    "direction": "시선이 프레임 하단의 핸드폰 화면을 정확히 향하고 있음.",
    "built_space": "아웃포커싱된 계단의 실루엣이 배경에 존재함.",
    "entities": "제호의 눈과 눈썹, '징계위원회 소집 통보'가 명확히 적힌 핸드폰 가장자리.",
    "hard_violations": [],
    "physics": "핸드폰이 프레임 밖의 손에 의해 지탱되고 있음이 자연스럽게 연출됨."
   },
   {
    "label": "B",
    "direction": "오른쪽 눈은 밖을 향하고 왼쪽 눈은 안을 향하는 등 시선이 불일치함(사시).",
    "built_space": "콘크리트 벽면과 계단, 난간이 배경에 위치함.",
    "entities": "제호의 얼굴 전체, 'DJH54'가 적힌 핸드폰 상단부.",
    "hard_violations": [
     "물리적으로 불가능한 해부학적 구조 (양안의 초점이 전혀 맞지 않는 사시)"
    ],
    "physics": "핸드폰이 프레임 밖의 손에 의해 지탱되고 있음."
   }
  ],
  "gq": {
   "route": "agree",
   "gap": 0.0,
   "per_model_winner": {
    "gemini-pro": "A",
    "openrouter:x-ai/grok-4.6": "A"
   },
   "models": [
    "gemini-pro",
    "openrouter:x-ai/grok-4.6"
   ]
  },
  "totals": {
   "A": 7,
   "B": 2
  },
  "selected": "A",
  "ranking": [
   "A",
   "B"
  ],
  "verdicts": [
   {
    "label": "A",
    "score": 7,
    "verdict_ko": "지정된 극단적인 클로즈업 앵글을 잘 살렸으며, 확장된 동공과 '징계위원회 소집 통보' 텍스트를 정확하게 묘사했습니다."
   },
   {
    "label": "B",
    "score": 2,
    "verdict_ko": "요구된 한쪽 눈의 익스트림 클로즈업 샷을 무시하고 얼굴 전체를 담았으며, 양안의 초점이 어긋나는 해부학적 오류가 있습니다."
   }
  ],
  "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/46575226-4c2f-4334-b16a-429b11371e24/images/5a7f5cfd-5b80-4f10-a4e5-1c8519aa244c/scene/recipe/S2sh6_sel.png",
    "asset_id": "f6465327-7378-42df-8102-5f4e87f4c276",
    "role": "prev_still"
   },
   {
    "label": "CHARACTER REFERENCE — 제호: the exact person appearing in this shot; match face, hair and build exactly.",
    "path": "<bytes:847626>",
    "asset_id": "7760e702-1e92-4b92-b945-00de7bc3d3c1",
    "role": "character_ref"
   }
  ]
 }
}