{ "S1sh3": { "input_fingerprint": "3e4ac32421b62160", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 눈을 부릅뜨고 입을 벌린 채, 어두운 뒤편을 향해 고개를 돌린 한국인 남성의 겁에 질린 얼굴.\n\nLOCATION (lock): A dense forest with thick undergrowth obstructing the ground-level path. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The fleeing man is missing one shoe, and the large bite wound on his thigh continues to bleed.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 숲속의 한국인 남성 (한국인 남성 얼굴, 성인 연령대, 자연스러운 검은 머리, 평범한 이목구비 윤곽타입 출력을 위한 JSON입니다. 시작은 반드시 { 로 끝은 반드시 } 로 마무리하세요.) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 8, "A": 6, "C": 5 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 8, "verdict_ko": "스토리보드의 카메라 구도와 인물 배치를 정확히 따르며, 눈을 부릅뜨고 입을 벌린 겁에 질린 표정을 완벽하게 구현했습니다." }, { "label": "A", "score": 6, "verdict_ko": "구도와 의상은 기준에 부합하나, 입을 벌린 정도와 겁에 질린 표정의 강도가 프롬프트 요구에 비해 다소 약합니다." }, { "label": "C", "score": 5, "verdict_ko": "표정 연기는 훌륭하나, 허벅지 상처를 억지로 화면에 넣기 위해 스토리보드의 구도를 무시하고 해부학적으로 부자연스러운 자세를 만들었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S1sh3.png" }, { "label": "CHARACTER REFERENCE — 숲속의 한국인 남성: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [] }, "fix_skipped": true, "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S1sh5": { "input_fingerprint": "ee93863f3118f109", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 바닥에 넘어진 한국인 남성 위로 검은 인영이 체중을 실어 거칠게 짓누르고, 남자의 상체가 바닥에 억눌려 있는 순간.\n\nLOCATION (lock): A dense forest with thick undergrowth obstructing the ground-level path. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The pinned man remains one-shoed, with the large bite wound on his thigh still bleeding.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 숲속의 한국인 남성 (한국인 남성 얼굴, 성인 연령대, 자연스러운 검은 머리, 평범한 이목구비 윤곽타입 출력을 위한 JSON입니다. 시작은 반드시 { 로 끝은 반드시 } 로 마무리하세요.); 검은 인영 (인간형 머리 윤곽, 식별되지 않는 얼굴, 어둠에 묻힌 머리 세부) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "스토리보드의 억누르는 구도를 완벽히 재현했으며, 검은 인영의 식별 불가능한 얼굴, 허벅지 상처 등 모든 지시 사항을 충실히 반영했습니다." }, { "label": "B", "score": 5, "verdict_ko": "행동과 구도는 일치하나, 어둠에 묻혀야 할 검은 인영의 머리와 피부 세부가 눈에 띄게 노출되어 감점 요소가 됩니다." }, { "label": "C", "score": 3, "verdict_ko": "남성을 거칠게 짓누른다는 핵심 텍스트와 스토리보드의 구도를 전혀 따르지 않고 단순히 마주 보는 형태로 연출되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S1sh5.png" }, { "label": "CHARACTER REFERENCE — 숲속의 한국인 남성: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "쓰러진 남자의 왼발이 프레임 밖으로 잘려 있어, 지문에 명시된 '신발을 한 짝만 신은(one-shoed)' 상태를 확인할 수 없습니다.", "fix_en": "Widen the camera framing to include the man's left foot, clearly showing that it is missing a shoe." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Widen the camera framing to include the man's left foot, clearly showing that it is missing a shoe.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S2sh1": { "input_fingerprint": "bc2dcb609860a8d5", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, dim streetlight.\n\nSHOT TEXT (authoritative, Korean): 희미한 가로등 불빛 아래, 전봇대에 등을 기댄 한국인 여성과 그 위에 겹쳐 서 있는 한국인 남성의 밀착된 상태.\n\nLOCATION (lock): A narrow urban back alley lined with utility poles, including one where two figures are entangled. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 도시 뒷골목의 한국인 여성 (한국인 여성, 젊은 성인 얼굴, 일반적인 동아시아계 이목구비, 자연스러운 검은 머리); 도시 뒷골목의 한국인 남성 (활성화 상태) (한국인 남성, 젊은 성인 얼굴, 일반적인 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "캐릭터들의 얼굴과 지정된 의상(여성의 트렌치코트, 남성의 비니와 가죽 재킷)을 완벽하게 재현했으나, 배경에 금지된 텍스트가 포함되었습니다." }, { "label": "A", "score": 5, "verdict_ko": "남성의 의상과 구도는 적절하나, 여성의 의상이 레퍼런스의 베이지색 트렌치코트와 완전히 다르게 생성되었습니다." }, { "label": "B", "score": 3, "verdict_ko": "두 인물 모두 지정된 레퍼런스 의상(비니, 트렌치코트 등)을 전혀 착용하지 않아 캐릭터 재현도가 매우 낮습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S2sh1.png" }, { "label": "CHARACTER REFERENCE — 도시 뒷골목의 한국인 여성: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 도시 뒷골목의 한국인 남성 (활성화 상태): the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "골목길의 간판, 포스터, 플라스틱 상자 등에 텍스트(소주, 포차 등)가 포함되어 있어 '텍스트 없음(No text)' 규칙을 위반했습니다.", "fix_en": "Remove all text from the signs, posters, windows, and crates in the alleyway to leave them blank." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text from the signs, posters, windows, and crates in the alleyway to leave them blank.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S2sh4": { "input_fingerprint": "875f2f912fdb3e9a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, dim streetlight.\n\nSHOT TEXT (authoritative, Korean): 아스팔트 바닥에 고개가 꺾인 채 늘어져 있는 여성의 상체.\n\nLOCATION (lock): A narrow urban back alley lined with utility poles, including one where two figures are entangled. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the dim streetlamp illumination, asphalt texture, utility pole, and cramped nighttime alley look from the reference. Exclude the standing man and the woman's earlier upright position; show only the collapsed woman required in this shot.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): The woman is collapsed motionless on the asphalt, with her upper torso against the pavement and her head bent sharply to one side. Her arms and legs remain slack and unbraced.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The collapsed woman is motionless, and the torn wound on her neck remains visible.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 도시 뒷골목의 한국인 여성 (한국인 여성, 젊은 성인 얼굴, 일반적인 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 6, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지시된 바닥에 엎드려 고개가 꺾인 자세와 목의 상처를 잘 표현했으며, 텍스트 금지 조건도 충실히 준수했습니다." }, { "label": "B", "score": 6, "verdict_ko": "쓰러진 자세와 목의 상처 묘사는 훌륭하나, 프롬프트에서 금지한 텍스트가 좌측 파란색 상자에 나타났습니다." }, { "label": "C", "score": 3, "verdict_ko": "고개가 꺾이지 않고 똑바로 위를 향해 누워 있어 자세 지침에 어긋나며, 필수 요소인 목의 상처도 누락되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S2sh4.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S2sh1_sel.png" }, { "label": "CHARACTER REFERENCE — 도시 뒷골목의 한국인 여성: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "화면 우측에 놓인 손의 손가락 형태가 심하게 뭉개져 해부학적으로 비정상적입니다.", "fix_en": "Redraw the hand on the right side of the frame with anatomically correct, distinct fingers resting slackly." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redraw the hand on the right side of the frame with anatomically correct, distinct fingers resting slackly.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+prev+엔티티" }, "S3sh2": { "input_fingerprint": "95fdce7bc4c7d53d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 형사 무리 한가운데 꼿꼿이 서서 손가락으로 바닥의 시신 쪽을 가리키고 있는 혜수의 정면.\n\nLOCATION (lock): Cordoned-off crime scenes at a forest trail and an urban back alley, with police lines, forensic activity, and body recovery shown as local news footage. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The attacked victims remain dead with their injuries intact at the cordoned scenes until the police recover their bodies.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 3, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "혜수의 레퍼런스 의상(네이비 재킷)과 외모를 정확히 반영했으며, 스토리보드에 지정된 형사들의 배치와 동작을 가장 충실하게 구현했습니다." }, { "label": "B", "score": 3, "verdict_ko": "혜수가 지정된 의상 대신 검은색 정장을 입고 있어 캐릭터 레퍼런스 지침을 위반했으며, 우측 가장자리에 불필요한 인물의 신체 일부가 노출되었습니다." }, { "label": "C", "score": 5, "verdict_ko": "혜수의 복장과 동작은 일치하나, 샷 텍스트에 명시되지 않은 카메라맨과 제복 경찰 등 다수의 불필요한 인물이 과도하게 추가되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S3sh2.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "폴리스 라인(POLICE LINE)과 주변 벽면, 기둥 등에 영문 및 한글 텍스트가 명확하게 노출되어 있어 '텍스트 없음' 지시를 위반했습니다.", "fix_en": "Remove all visible text, letters, and numbers from the police tape, walls, poles, and posters." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all visible text, letters, and numbers from the police tape, walls, poles, and posters.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S4sh1": { "input_fingerprint": "4a32b23b5175498f", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 낮의 맑은 하늘 아래, 오래된 변두리 다세대 빌라 옥상 위의 옥탑방 지붕이 내려다보이는 넓은 부감 쇼트.\n\nLOCATION (lock): The rooftop level of an old low-rise residential villa on the outskirts of a city, with a small rooftop dwelling seen from above. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 4, "B": 4, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 4, "verdict_ko": "스토리보드의 카메라 구도를 따르지 않고 레퍼런스 사진의 프레이밍과 흐린 조명을 그대로 복사한 채 하늘색만 바꾸어 지시를 위반했습니다." }, { "label": "B", "score": 4, "verdict_ko": "후보 A와 동일하게 참조 사진의 타이트한 앵글을 답습하여 넓은 부감 쇼트 지시사항을 이행하지 못했습니다." }, { "label": "C", "score": 4, "verdict_ko": "다른 후보들과 마찬가지로 스토리보드 기반의 카메라 셋업을 무시하고 레퍼런스 사진을 단순 합성하는 데 그쳤습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S4sh1.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 '맑은 하늘'이라고 명시되어 있고 레퍼런스의 조명을 복사하지 말라는 지시가 있었으나, 화면 상단에 보이는 하늘과 전반적인 분위기가 레퍼런스 사진처럼 흐린 날씨로 묘사되었습니다.", "fix_en": "Change the visible sky to a clear blue and adjust the lighting to reflect bright, sunny daylight." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the visible sky to a clear blue and adjust the lighting to reflect bright, sunny daylight.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed" }, "S5sh1": { "input_fingerprint": "e7a918a6106824b4", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, thin sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 작은 창문으로 들어오는 가느다란 햇살 아래, 앞치마를 두른 강민숙(민숙)이 굳은 표정으로 거실 TV 화면에 시선을 고정한 모습.\n\nLOCATION (lock): A modest rooftop apartment interior with a small living-room window, a compact kitchen sink and stove, a dining table, a television, and adjoining bedrooms. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Minsuk already bears the small circular mark on her wrist; it is a long-standing physical feature even when not shown in this framing.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 여성, 40대 초반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 5, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "위치 레퍼런스를 정확히 따르고 손목의 둥근 자국(Carried State)을 유일하게 반영했으나, TV 화면에 금지된 텍스트가 생성된 점이 감점 요소입니다." }, { "label": "C", "score": 5, "verdict_ko": "공간 배경과 조명 연출은 지시사항에 부합하지만, 필수 요소인 손목의 둥근 자국이 완전히 누락되었습니다." }, { "label": "A", "score": 3, "verdict_ko": "손목 자국이 누락되었으며, 원본 레퍼런스에 없는 주방을 배경에 임의로 추가하여 공간 제약(Location lock)을 심각하게 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S5sh1.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "TV 화면에 뉴스 자막 등의 텍스트가 포함되어 있습니다 (프롬프트에서 모든 텍스트 금지).", "fix_en": "Remove all text, graphics, and captions from the TV screen display." }, { "issue_ko": "캐릭터의 머리가 뒤로 묶여 있어, 풀어진 짧은 단발머리인 레퍼런스 이미지와 일치하지 않습니다.", "fix_en": "Change the character's hair to an untied, short bob matching the character reference." }, { "issue_ko": "손목에 '작고 둥근 자국'이 명확히 표현되지 않고 불분명한 점들로 묘사되어 있습니다.", "fix_en": "Replace the blurry marks on the character's exposed wrist with one clear, small circular mark." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text, graphics, and captions from the TV screen display.\n- Change the character's hair to an untied, short bob matching the character reference.\n- Replace the blurry marks on the character's exposed wrist with one clear, small circular mark.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S5sh7": { "input_fingerprint": "df6bfd061f2f8162", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, thin sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 수리영이 강민숙(민숙)의 어깨 가까이에 코를 댄 채 옅은 미소를 짓는 옆얼굴.\n\nLOCATION (lock): A modest rooftop apartment interior with a small living-room window, a compact kitchen sink and stove, a dining table, a television, and adjoining bedrooms. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The small circular mark remains on Minsuk's wrist while Suriyoung leans close to her.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 여성, 40대 초반 얼굴, 성숙한 동아시아계 이목구비, 자연스러운 검은 머리); 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 5, "B": 7, "C": 4 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "A", "score": 5, "verdict_ko": "지정된 장소는 정확히 구현했으나, 두 인물의 복장이 레퍼런스와 완전히 다릅니다." }, { "label": "B", "score": 7, "verdict_ko": "장소 구조, 인물 복장, 손목의 표식 및 요구된 포즈를 모두 충실하게 재현했습니다." }, { "label": "C", "score": 4, "verdict_ko": "인물 복장은 비교적 정확하나, 기준 장소의 공간 구조를 임의로 크게 변형했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S5sh7.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "민숙이 베이지색 스웨터와 갈색 앞치마 대신 단색의 갈색 스웨터만 입고 있습니다.", "fix_en": "Change Minsuk's clothing to a beige sweater with a brown apron worn over it." }, { "issue_ko": "수리영의 목에 있던 녹색 넥커치프가 사라졌습니다.", "fix_en": "Add the green neckerchief around Suriyoung's neck under her collar." }, { "issue_ko": "머그잔을 잡고 있는 민숙의 손가락들이 뭉개져 잔의 손잡이와 형태가 합쳐져 있습니다.", "fix_en": "Redraw Minsuk's hand to show clearly defined fingers naturally gripping the mug handle." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change Minsuk's clothing to a beige sweater with a brown apron worn over it.\n- Add the green neckerchief around Suriyoung's neck under her collar.\n- Redraw Minsuk's hand to show clearly defined fingers naturally gripping the mug handle.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S5sh8": { "input_fingerprint": "3a0d398c70f10e25", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, thin sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 강민숙(민숙)의 손목 피부 위에 선명하게 자리 잡은 작은 원형 표식 클로즈업.\n\nLOCATION (lock): A modest rooftop apartment interior with a small living-room window, a compact kitchen sink and stove, a dining table, a television, and adjoining bedrooms. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The clearly defined small circular mark remains fixed on Minsuk's wrist.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 여성, 40대 초반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 8, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 8, "verdict_ko": "장소, 인물의 외모와 의상, 손목에 있는 작은 원형 표식까지 프롬프트와 레퍼런스의 요구사항을 가장 정확하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "인물과 장소의 묘사는 훌륭하나, 손목의 표식이 완전한 원형이 아닌 'C' 자 형태로 그려져 아쉽습니다." }, { "label": "B", "score": 3, "verdict_ko": "인물의 얼굴이 보이지 않고 지정된 의상과 다른 옷을 입고 있으며, 배경 장소의 구조 또한 레퍼런스와 다르게 변형되었습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S5sh8.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 붉은 원형 표식: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "표식이 있는 팔 외에도 이를 잡고 있는 두 개의 손과 화면 우측에 인물의 얼굴이 함께 등장하여, 프롬프트에 없는 추가 인물(또는 여분의 신체 부위)이 묘사되었습니다.", "fix_en": "Remove the face on the right and the extra hand supporting the wrist from below, leaving only one hand holding the marked arm." }, { "issue_ko": "로케이션 레퍼런스 사진의 방 한가운데에 있는 식탁과 의자들이 배경에서 누락되었습니다.", "fix_en": "Add the dining table and chairs back into the center of the room in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the face on the right and the extra hand supporting the wrist from below, leaving only one hand holding the marked arm.\n- Add the dining table and chairs back into the center of the room in the background.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S6sh1": { "input_fingerprint": "7de11237394b02b4", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 이어폰을 낀 수리영이 해안가 도로에서 자전거 페달을 밟아 내리는 중간 찰나, 머리카락과 옷깃이 뒤로 휘날리며 고개를 살짝 위로 들고 있는 상태.\n\nLOCATION (lock): A coastal road running alongside the sea, suitable for bicycle travel. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung is traveling with her bicycle and is still wearing her earphones.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 6, "A": 5 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "캐릭터의 이목구비와 의상(어두운 셔츠, 갈색 부츠)을 레퍼런스와 가장 정확히 일치시켰으며 제시된 동작과 배경을 훌륭하게 구현했습니다." }, { "label": "C", "score": 6, "verdict_ko": "요구된 동작과 배경은 잘 표현되었으나, 캐릭터의 셔츠 색상과 신발이 레퍼런스와 다릅니다." }, { "label": "A", "score": 5, "verdict_ko": "상황 묘사는 적절하나 캐릭터의 얼굴 생김새가 레퍼런스와 다소 차이가 나며 의상 디테일도 일치하지 않습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S6sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터 레퍼런스 이미지에 있는 목 부분의 짙은 녹색 반다나(스카프)가 누락되었습니다.", "fix_en": "Add the dark green bandana around the character's neck to match the reference." }, { "issue_ko": "캐릭터가 메고 있어야 할 검은색 크로스백과 어깨 스트랩이 누락되었습니다.", "fix_en": "Add the black cross-body shoulder bag and its strap across the character's torso." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the dark green bandana around the character's neck to match the reference.\n- Add the black cross-body shoulder bag and its strap across the character's torso.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S7sh1": { "input_fingerprint": "5f3a148081ef0726", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 날카로운 이빨을 드러낸 검은색 로트와일러가 철망 안에서 네 발로 뻣뻣하게 서 있는 위협적인 자태.\n\nLOCATION (lock): A dog shelter with secure kennels for multiple animals and practical areas for veterinary injections, grooming, ointment treatment, and washing. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 4, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "공간의 구조(Location lock)를 정확히 유지하며 개를 기존 철망(견사) 안에 배치했으나, 네 발로 서 있지 않고 앉아 있는 자세가 아쉽습니다." }, { "label": "B", "score": 4, "verdict_ko": "개가 네 발로 서서 이빨을 드러내는 모습은 잘 표현되었으나, 복도 중앙에 원본에 없는 철망 벽을 임의로 생성하여 공간의 구조적 일관성을 훼손했습니다." }, { "label": "A", "score": 3, "verdict_ko": "개는 네 발로 서 있으나 견사 안이 아닌 복도에 위치하며, 화면 전체에 공간감과 무관한 2D 철망 무늬가 겹쳐져 있어 부자연스럽습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L06B01.png" } ], "critique": { "issues": [ { "issue_ko": "로트와일러가 네 발로 뻣뻣하게 서 있어야 한다는 지시와 달리 앉아 있는 자세입니다.", "fix_en": "Change the dog's pose so it is standing stiffly on all four legs." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the dog's pose so it is standing stiffly on all four legs.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트만 (배경 전용)" }, "S7sh3": { "input_fingerprint": "48b83318084e99d1", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 로트와일러가 바닥에 웅크려 엎드린 채 고개를 푹 숙인 얌전한 상태.\n\nLOCATION (lock): A dog shelter with secure kennels for multiple animals and practical areas for veterinary injections, grooming, ointment treatment, and washing. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "A": 4, "C": 3 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "로트와일러가 바닥에 납작 엎드려 고개를 숙인 얌전한 상태를 프롬프트 지시대로 완벽하게 구현했습니다." }, { "label": "A", "score": 4, "verdict_ko": "고개를 숙이고 있으나 바닥에 엎드린(lying flat) 상태가 아니라 뒷다리로 앉아 있어 자세 묘사가 아쉽습니다." }, { "label": "C", "score": 3, "verdict_ko": "엎드린 자세가 아니며 그릇을 향해 몸을 기울이고 있어 프롬프트가 요구한 특정 동작과 일치하지 않습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L06B01.png" } ], "critique": { "issues": [ { "issue_ko": "로트와일러의 앞발이 3개로 잘못 생성되었습니다.", "fix_en": "Remove the extra front paw so the Rottweiler has exactly two anatomically correct front legs resting on the floor." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the extra front paw so the Rottweiler has exactly two anatomically correct front legs resting on the floor.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트만 (배경 전용)" }, "S7sh5": { "input_fingerprint": "56316d68ce515b55", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 스마트폰을 귀에 바짝 댄 채 입술을 살짝 깨문 모습.\n\nLOCATION (lock): A dog shelter with secure kennels for multiple animals and practical areas for veterinary injections, grooming, ointment treatment, and washing. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the animal-shelter wire mesh, utilitarian surfaces, daylight, and nearby caged-dog environment from the reference. Exclude the Rottweiler as a dominant foreground subject and exclude its crouched pose; frame the woman making the phone call instead.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung has her smartphone with her as she tries unsuccessfully to contact Minsuk.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "지시된 복장(스카프)과 행동(입술 깨물기)을 정확히 유지했으며, 배제해야 할 개(로트와일러)도 완벽히 제외했습니다." }, { "label": "A", "score": 5, "verdict_ko": "캐릭터의 복장과 행동은 잘 묘사되었으나, 프롬프트에서 제외하라고 명시된 로트와일러가 배경에 포함되었습니다." }, { "label": "B", "score": 4, "verdict_ko": "로트와일러는 제외되었으나, 이전 샷에서 반드시 이어져야 하는 캐릭터의 녹색 스카프가 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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L06B01.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S7sh3_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "기준 이미지에 있는 캐릭터의 검은색 크로스백 가방 끈이 누락되었습니다.", "fix_en": "Add the black crossbody bag strap diagonally across the character's chest." }, { "issue_ko": "스마트폰 화면(기능면)이 귀가 아닌 바깥(카메라) 쪽을 향하고 있습니다.", "fix_en": "Flip the smartphone so the screen faces the ear and the back faces the camera." }, { "issue_ko": "전화를 쥔 오른손이 앞뒤가 뒤집혀 손등이 뺨에 닿아 있는 해부학적 오류가 있습니다.", "fix_en": "Redraw the right hand with the palm facing the ear and fingers wrapping around the back of the phone." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the black crossbody bag strap diagonally across the character's chest.\n- Flip the smartphone so the screen faces the ear and the back faces the camera.\n- Redraw the right hand with the palm facing the ear and fingers wrapping around the back of the phone.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+prev+엔티티" }, "S8sh2": { "input_fingerprint": "73fb46d9bb7717fd", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, neon-lit streets and supermarket interior.\n\nSHOT TEXT (authoritative, Korean): 마트 안 상품 진열대 사이에서 미간을 살짝 찌푸리고 두리번거리는 각도로 얼굴을 돌린 수리영.\n\nLOCATION (lock): A busy commercial district beside a harbor, crowded with neon signs, taxis, street vendors, and food stalls, leading into a large supermarket with stocked display aisles. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung's bicycle remains parked outside the mart, and she continues to carry her smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지정된 인물만 등장해야 한다는 엄격한 제한과 장소 구조를 잘 준수했으나, 캐릭터의 크로스백이 누락되고 지시된 고개 각도가 다소 밋밋합니다." }, { "label": "B", "score": 5, "verdict_ko": "캐릭터의 의상 디테일과 표정 연기는 훌륭하지만, 프롬프트에 명시되지 않은 다수의 배경 인물을 추가하여 중요한 배제 조건을 위반했습니다." }, { "label": "C", "score": 3, "verdict_ko": "캐릭터가 잘못된 의상(앞치마)을 입고 있으며, 장소의 구조 변경, 임의의 텍스트 추가, 불필요한 인물 등장 등 다수의 지침을 크게 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L08B01.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S8sh2.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "텍스트의 '얼굴을 돌린' 묘사와 레이아웃 스케치에 제시된 자세(고개를 왼쪽으로 돌리고 걷는 포즈)가 반영되지 않고, 정면을 향해 꼿꼿이 서 있습니다.", "fix_en": "Turn her head to the left and change her stance to a walking pose, matching the action described in the text and the layout sketch." }, { "issue_ko": "캐릭터 레퍼런스 이미지에 있는 검은색 크로스백이 누락되었습니다.", "fix_en": "Add the black crossbody bag and its strap across her chest exactly as shown in the character reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Turn her head to the left and change her stance to a walking pose, matching the action described in the text and the layout sketch.\n- Add the black crossbody bag and its strap across her chest exactly as shown in the character reference.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S8sh4": { "input_fingerprint": "2652e867934070f1", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, neon-lit streets and supermarket interior.\n\nSHOT TEXT (authoritative, Korean): 마트 밖 번화가 길가에 서서 스마트폰 화면을 빤히 내려다보는 수리영.\n\nLOCATION (lock): A busy commercial district beside a harbor, crowded with neon signs, taxis, street vendors, and food stalls, leading into a large supermarket with stocked display aisles. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Her bicycle remains outside the mart while Suriyoung continues carrying and checking her smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "A": 5, "C": 3 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "주인공의 복장(가방, 스카프) 재현이 훌륭하며, 불필요한 인물 추가를 최소화해 프롬프트에 가장 충실합니다." }, { "label": "A", "score": 5, "verdict_ko": "장소 묘사는 좋으나 텍스트에 없는 군중과 노점상을 다수 추가했고 인물의 크로스백이 누락되었습니다." }, { "label": "C", "score": 3, "verdict_ko": "인물의 스카프와 가방이 누락되었으며, 배치된 자전거 앞바퀴의 형태가 심하게 왜곡되었습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L08B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S8sh4.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 명시되지 않은 배경 인물들이 길거리에 추가되었습니다.", "fix_en": "Remove all background people from the sidewalk and street so that only Suriyoung is in the frame." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all background people from the sidewalk and street so that only Suriyoung is in the frame.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S9sh1": { "input_fingerprint": "d499dd3c73b30769", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 어두운 밤 해안도로를 배경으로 자전거 핸들을 꽉 쥐고 허리를 숙인 채 한쪽 무릎을 굽혀 힘차게 페달을 밟고 있는 수리영의 측면 자세, 머리카락과 옷깃이 뒤로 휘날리고 있음.\n\nLOCATION (lock): A coastal road beside the dark sea, used as a bicycle route back toward the residential district. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung continues home on the same bicycle with her smartphone still in her possession.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 5, "A": 4 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 측면 구도를 정확히 따랐으며, 스마트폰 소지 지시도 주머니에 자연스럽게 반영되었습니다." }, { "label": "B", "score": 5, "verdict_ko": "측면 구도와 배경은 일치하나, 허리 뒤에 스마트폰이 부자연스럽게 떠 있는 오류가 있습니다." }, { "label": "A", "score": 4, "verdict_ko": "스토리보드에 명시된 측면 앵글을 무시하고 정면 대각선 구도를 사용하여 카메라 지침을 크게 위반했습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S9sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터 레퍼런스에 있는 검은색 크로스백과 어깨 끈이 누락되었습니다.", "fix_en": "Add the black crossbody bag and its strap across the character's chest to match the reference." }, { "issue_ko": "캐릭터 레퍼런스에서 목에 두르고 있는 초록색 반다나가 없습니다.", "fix_en": "Add the green bandana around the character's neck exactly as shown in the reference." }, { "issue_ko": "캐릭터의 왼쪽 발이 페달을 밟고 있지 않고 자전거 크랭크 위 허공에 떠 있습니다.", "fix_en": "Reposition the left foot so it is firmly pressing down on a bicycle pedal attached to the crankset." }, { "issue_ko": "왼쪽 발의 신발이 레퍼런스의 갈색 부츠가 아닌 하얀 밑창의 운동화로 잘못 생성되었습니다.", "fix_en": "Change the shoe on the left foot to a brown boot, matching the character reference and the right foot." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the black crossbody bag and its strap across the character's chest to match the reference.\n- Add the green bandana around the character's neck exactly as shown in the reference.\n- Reposition the left foot so it is firmly pressing down on a bicycle pedal attached to the crankset.\n- Change the shoe on the left foot to a brown boot, matching the character reference and the right foot.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S10sh3": { "input_fingerprint": "e80985e01390f201", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light casting long shadows.\n\nSHOT TEXT (authoritative, Korean): 자물쇠가 있는 옥탑방 현관 철문이 안쪽으로 아주 미세한 틈을 벌린 채 멈춘 구도.\n\nLOCATION (lock): An exposed rooftop around a small rooftop apartment, with a water tank, clothesline, drying platform covered with herbs, and a metal entrance door. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The iron entrance door has been unlocked and left slightly open by the intruder.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 8, "C": 4, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 8, "verdict_ko": "문이 안쪽으로 미세하게 열린 상태와 우측 경첩 구조를 정확히 구현했으며, 야간 조명 및 옥상 소품 배치도 지시에 부합합니다." }, { "label": "C", "score": 4, "verdict_ko": "프레이밍과 배경 요소는 무난하나, 현관문이 완전히 닫혀 있어 '미세한 틈을 벌린' 핵심 묘사를 누락했습니다." }, { "label": "A", "score": 3, "verdict_ko": "문이 바깥쪽으로 넓게 열려 있어 '안쪽으로 미세한 틈'이라는 지시를 위반했으며 레퍼런스의 문의 경첩 위치도 반대로 왜곡되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S10sh3.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" } ], "critique": { "issues": [ { "issue_ko": "문이 열려 있어야 한다는 설정과 다르게 자물쇠가 걸쇠에 굳게 채워져 있습니다.", "fix_en": "Show the padlock hanging open and the latch unhooked so the door can be slightly open." }, { "issue_ko": "지시문에 없는 CCTV 카메라가 건물 외벽 우측 상단에 임의로 추가되었습니다.", "fix_en": "Remove the CCTV camera from the brick wall." }, { "issue_ko": "허브를 말리는 채반들이 허공에 떠서 벽을 뚫고 겹쳐져 있습니다.", "fix_en": "Rest the herb drying trays completely flat on top of the concrete railing." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Show the padlock hanging open and the latch unhooked so the door can be slightly open.\n- Remove the CCTV camera from the brick wall.\n- Rest the herb drying trays completely flat on top of the concrete railing.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed" }, "S10sh4": { "input_fingerprint": "2fd7fea979382b6c", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light casting long shadows.\n\nSHOT TEXT (authoritative, Korean): 창문 유리에 짙게 드리워진 그림자 속, 남성의 실루엣이 여성의 어깨를 거칠게 물어뜯는 듯한 역동적인 자세로 멈춘 찰나.\n\nLOCATION (lock): An exposed rooftop around a small rooftop apartment, with a water tank, clothesline, drying platform covered with herbs, and a metal entrance door. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Minsuk still bears the small circular wrist mark, and the attack is creating the severe bite wound on her shoulder.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 9, "B": 6, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 9, "verdict_ko": "스토리보드의 구도를 완벽히 따르면서 창문에 비친 그림자 손목에 둥근 자국까지 정확히 표현한 가장 훌륭한 결과물입니다." }, { "label": "B", "score": 6, "verdict_ko": "구도와 장소는 잘 재현했으나, 손목의 자국이 누락되었고 유리창의 오돌토돌한 질감이 다소 부자연스럽습니다." }, { "label": "C", "score": 4, "verdict_ko": "창문에 비친 도시 도로망의 반사가 물리적으로 어색하며, 손목 자국과 빨랫줄의 옷 등 주요 디테일이 누락되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S10sh4.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" } ], "critique": { "issues": [ { "issue_ko": "여성의 들어 올린 왼팔 그림자 내부에 둥근 손목 자국이 선으로 묘사되어 있습니다 (단일한 그림자 영역에는 피부 표면의 무늬가 투영될 수 없습니다).", "fix_en": "Remove the dark circular line drawn inside the shadow of the raised left wrist, making the shadow's interior a solid, uniform tone." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the dark circular line drawn inside the shadow of the raised left wrist, making the shadow's interior a solid, uniform tone.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed" }, "S11sh3": { "input_fingerprint": "2a35b408af570214", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 수리영의 시선 너머, 옥상 바닥 위로 반쯤 활짝 열린 채 내부의 어둠을 드러내고 있는 현관 철문.\n\nLOCATION (lock): A narrow residential alley leading into the yard of a multi-unit villa, with a secluded bicycle parking spot and an exterior stairway rising to a rooftop apartment and its metal door. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung's bicycle remains hastily parked in the rear yard, and the iron entrance door remains half open from the intrusion. She still has her smartphone.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 3, "C": 4 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "스토리보드의 앵글과 지정된 구조물의 디테일(흰색 창틀, 단일 물탱크), 반쯤 열린 철문의 상태를 가장 정확히 재현했습니다." }, { "label": "B", "score": 3, "verdict_ko": "지정된 옥상 구조와 달리 왼쪽에 물탱크가 두 개로 중복 생성되는 오류가 발생했습니다." }, { "label": "C", "score": 4, "verdict_ko": "전반적인 구도는 무난하나, 레퍼런스 사진에 고정된 창틀 색상(흰색)을 임의로 어둡게 변경했습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S11sh3.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" } ], "critique": { "issues": [ { "issue_ko": "레퍼런스 사진에서 옥상 건물의 창문과 현관문은 90도 모서리를 기준으로 서로 수직인 다른 벽면에 위치하지만, 생성된 이미지에서는 한 평면 벽에 나란히 있습니다.", "fix_en": "Rebuild the rooftop structure so the window and the metal door are on adjacent, perpendicular walls meeting at a corner, exactly matching the reference photo's layout." }, { "issue_ko": "레퍼런스 사진의 옥상 건물 창문은 3칸으로 나뉘어 있으나, 생성된 이미지의 창문은 4칸으로 되어 있습니다.", "fix_en": "Change the rooftop window to have exactly three glass panels instead of four to match the reference photo." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rebuild the rooftop structure so the window and the metal door are on adjacent, perpendicular walls meeting at a corner, exactly matching the reference photo's layout.\n- Change the rooftop window to have exactly three glass panels instead of four to match the reference photo.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed" }, "S12sh6": { "input_fingerprint": "6ade0da59394399f", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, dim floor-lamp and television light.\n\nSHOT TEXT (authoritative, Korean): 살점이 덜렁거리고 거칠게 물어뜯긴 어깨 상처에서 피가 흐르는 강민숙(민숙)이 고개를 푹 떨구고 바닥에 앉아 있는 참혹한 모습.\n\nLOCATION (lock): A compact rooftop apartment with a living room, dining table, main bedroom, and a separate young woman’s bedroom with a bed, wall, window, and curtain. The interior contains a cooled cooking pot, a television, and two teacups. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): Min-sook is seated motionless on the floor behind the curtain, with her torso slumped and her head hanging forward. One rigid hand grips a crumpled photograph, and her remaining limbs stay slack in the seated position.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Minsuk remains dead, seated motionless on the floor behind the curtain, with her torso slumped and her head hanging forward, her shoulder and collarbone torn open and bleeding; the room retains the bloodstains, pooled blood, and bloody footprints. Because she cannot grasp it later, one rigid hand already grips the crumpled photograph, and her wrist still bears the circular mark.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 여성, 40대 초반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 3, "B": 7, "C": 2 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "프롬프트에 명시된 어깨의 참혹한 상처와 출혈이 전혀 묘사되지 않았습니다." }, { "label": "B", "score": 7, "verdict_ko": "어깨의 뜯긴 상처와 출혈을 프롬프트대로 정확히 구현했으며, 축 늘어진 자세와 사진을 쥔 손의 디테일도 잘 살렸습니다." }, { "label": "C", "score": 2, "verdict_ko": "어깨 상처가 누락되었고, 무릎을 세운 자세는 근육의 힘이 빠진 상태라는 지시와 어긋납니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B06.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S12sh6.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "손목에 명시된 둥근 자국이 없습니다.", "fix_en": "Add the circular mark to her visible left wrist." }, { "issue_ko": "손에 쥔 사진이 구겨져 있지 않고 반듯하게 펴져 있습니다.", "fix_en": "Change the flat photograph in her hand into a heavily crumpled piece of paper." }, { "issue_ko": "축 늘어져야 할 죽은 사람의 손목이 꺾인 채 사진을 지탱하고 있어 중력 원칙에 위배됩니다.", "fix_en": "Relax the left wrist so the hand lies completely flat and limp on the floor while gripping the object." }, { "issue_ko": "레퍼런스에 있던 벽면 붉은 원 안의 피 묻은 사진이 누락되었습니다.", "fix_en": "Restore the bloody photograph pinned inside the red circle on the wall, matching the location reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the circular mark to her visible left wrist.\n- Change the flat photograph in her hand into a heavily crumpled piece of paper.\n- Relax the left wrist so the hand lies completely flat and limp on the floor while gripping the object.\n- Restore the bloody photograph pinned inside the red circle on the wall, matching the location reference.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S12sh9": { "input_fingerprint": "b6ac562d79d32f74", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, dim floor-lamp and television light.\n\nSHOT TEXT (authoritative, Korean): 피투성이가 된 강민숙(민숙)의 굳은 손안에 사진 한 장이 구겨진 채 쥐어져 있는 클로즈업.\n\nLOCATION (lock): A compact rooftop apartment with a living room, dining table, main bedroom, and a separate young woman’s bedroom with a bed, wall, window, and curtain. The interior contains a cooled cooking pot, a television, and two teacups. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the dim room lighting, bloodied condition, and the dead woman's fixed seated position and clothing from the reference. Exclude the wider view of her torn shoulder and surrounding room from the tight frame; concentrate on her rigid hand and the crumpled photograph.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): Min-sook is seated motionless on the floor behind the curtain, with her torso slumped and her head hanging forward. One rigid hand grips a crumpled photograph, and her remaining limbs stay slack in the seated position.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The dead Minsuk remains seated behind the curtain with one rigid, bloodied hand gripping the crumpled photograph. Her circular wrist mark is present and turns red as Suriyoung removes the photograph.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 강민숙 (한국인 여성, 40대 초반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 5, "B": 3, "C": 7 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "A", "score": 5, "verdict_ko": "타이트한 클로즈업 지시를 따르지 않았고, 팔이 허공에 떠 있어 중력 규칙을 위반했습니다." }, { "label": "B", "score": 3, "verdict_ko": "클로즈업 및 배경 배제 지시를 완전히 무시한 넓은 화각이며, 손의 자세도 이전 샷과 불일치합니다." }, { "label": "C", "score": 7, "verdict_ko": "클로즈업 프레이밍과 구겨진 사진, 손의 중력 표현은 훌륭하나 지문에 없는 타인의 손이 등장해 감점되었습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B06.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S12sh6_sel.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 금월 식당과 두 어린 여자아이가 찍힌 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 명시되지 않은 두 번째 인물(사진을 당기는 깨끗한 손)이 화면에 추가되었습니다.", "fix_en": "Remove the clean right hand and sleeve pulling the photograph from the right side of the frame." }, { "issue_ko": "이전 샷 레퍼런스에서 사진은 바닥에 늘어진 오른손에 있었으나, 이빨 자국이 있는 왼손이 무릎 위에서 사진을 쥐고 있어 자세와 위치가 일치하지 않습니다.", "fix_en": "Change the framing to show the crumpled photograph in the right hand resting on the floor, leaving the arm with the circular mark empty on her lap." }, { "issue_ko": "사망한 인물의 손목과 손이 지지대에 완전히 축 늘어지지 않고 공중으로 들린 채 물건을 쥐고 있어 중력 규칙을 위반했습니다.", "fix_en": "Lower the bloodied wrist and hand so they lie completely flat and limp against a supporting surface, without holding the object up in the air." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the clean right hand and sleeve pulling the photograph from the right side of the frame.\n- Change the framing to show the crumpled photograph in the right hand resting on the floor, leaving the arm with the circular mark empty on her lap.\n- Lower the bloodied wrist and hand so they lie completely flat and limp against a supporting surface, without holding the object up in the air.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+prev+엔티티" }, "S13sh3": { "input_fingerprint": "81402aae42089943", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 초점을 잃은 흐릿한 눈동자로 위쪽 허공을 응시하며 팔을 늘어뜨리고 멍하게 서 있는 수리영의 측면.\n\nLOCATION (lock): The exterior of a rooftop apartment and the steep stairway descending to the multi-unit villa below, with street access for an arriving police vehicle. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung still has the smartphone she used to call Hyesu; Minsuk's body and the bloodied room remain inside until their unexplained disappearance before the officers enter.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 카메라 구도와 인물 배치를 정확히 따랐으며, 지정된 옥탑방 건축물의 구조와 형태를 가장 충실하게 재현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "스토리보드에 제시된 카메라 구도와 배경 배치를 벗어났으며, 기준 이미지의 옥탑 건물 형태가 정확히 반영되지 않았습니다." }, { "label": "B", "score": 3, "verdict_ko": "인물과 배경 건물 간의 원근감과 크기 비례가 완전히 어긋나는 물리적으로 불가능한 스케일 오류(Hard Violation)가 있습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S13sh3.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영의 머리가 캐릭터 레퍼런스(푼 머리)와 다르게 스토리보드를 따라 포니테일로 묶여 있습니다.", "fix_en": "Change Suriyoung's hair to be worn loose and down, matching the character reference exactly." }, { "issue_ko": "수리영이 캐릭터 레퍼런스에 지정된 의상(갈색 재킷, 스카프, 가방 등)이 아닌 스토리보드에 그려진 반팔 티셔츠를 입고 있습니다.", "fix_en": "Dress Suriyoung in the brown field jacket, green scarf, and crossbody bag strap shown in her character reference." }, { "issue_ko": "경찰차 측면과 번호판에 문자와 숫자가 포함되어 있어 '텍스트 없음' 지침을 위반했습니다.", "fix_en": "Remove all text, letters, and numbers from the police car's side livery and license plate." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change Suriyoung's hair to be worn loose and down, matching the character reference exactly.\n- Dress Suriyoung in the brown field jacket, green scarf, and crossbody bag strap shown in her character reference.\n- Remove all text, letters, and numbers from the police car's side livery and license plate.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed+엔티티" }, "S13sh4": { "input_fingerprint": "2b0f1d54f1980106", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 두 눈을 동그랗게 뜨고 입을 반쯤 벌린 채 충격으로 굳어 있는 혜수의 얼굴 클로즈업.\n\nLOCATION (lock): The exterior of a rooftop apartment and the steep stairway descending to the multi-unit villa below, with street access for an arriving police vehicle. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the nighttime exterior lighting, worn residential-building surroundings, and the distressed young woman's position nearby from the reference. Exclude the young woman as the main subject and exclude her vacant stare; frame the arriving female officer's shocked face in close-up.\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\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "얼굴 클로즈업 프레이밍과 충격받은 표정을 완벽하게 구현했으며, 기준 이미지의 배경 구조물과 인물 디자인을 가장 정확하게 반영했습니다." }, { "label": "B", "score": 5, "verdict_ko": "얼굴 클로즈업과 표정 지시는 잘 따랐으나, 배경 구조물이 기준 이미지에 잠금 설정된 장소와 일치하지 않습니다." }, { "label": "C", "score": 4, "verdict_ko": "지정된 얼굴 클로즈업보다 샷이 넓게 잡혔으며, 샷 텍스트에 명시되지 않은 인물이 배경에 포함되어 프레이밍과 인물 지시를 어겼습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S13sh4.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S13sh3_sel.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터 레퍼런스 이미지에는 없는 경찰 마크 패치가 재킷 소매에 임의로 추가되었습니다.", "fix_en": "Remove the police insignia patch from the left sleeve so the jacket is completely plain, matching the character reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the police insignia patch from the left sleeve so the jacket is completely plain, matching the character reference.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed+prev+엔티티" }, "S14sh2": { "input_fingerprint": "2f50c7537acf496b", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 핏자국이나 시신 없이 완벽하게 정돈된 방안과 깨끗한 침대, 열려 있는 커튼 쪽의 빈 공간이 보이는 정적 구도.\n\nLOCATION (lock): The same compact rooftop apartment, including its entrance, clean dining area, and an unnaturally immaculate bedroom with a bed, curtain, backpack, and doll. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The room is now unnaturally clean: Minsuk's body, all blood, and the red wall mark are gone, and the curtain area is empty. Suriyoung's backpack and doll remain in the room but have been shifted slightly from their earlier positions.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 6, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "찢어진 벽지와 바닥재 등 원본 공간의 고정 요소를 유지하면서 침대를 배치해 커튼 쪽 빈 공간을 명확히 확보, 지문의 의도를 가장 잘 구현했습니다." }, { "label": "A", "score": 6, "verdict_ko": "장소의 물리적 특징과 정돈된 상태를 잘 살렸으나, 커튼 쪽 빈 공간이 좁고 지문에 없는 스탠드 조명이 추가되어 아쉽습니다." }, { "label": "B", "score": 4, "verdict_ko": "방은 정돈되었으나, 원본 사진의 뜯어진 벽지를 깨끗하게 바꿔버려 장소 일치(Location lock) 조건을 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B04.png" } ], "critique": { "issues": [] }, "fix_skipped": true, "ref_mode": "플레이트만 (배경 전용)" }, "S14sh3": { "input_fingerprint": "40389629c2900738", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혈색 없이 창백해진 얼굴로 미간을 잔뜩 찌푸린 수리영이 빈 방 한가운데 우두커니 서 있는 정면.\n\nLOCATION (lock): The same compact rooftop apartment, including its entrance, clean dining area, and an unnaturally immaculate bedroom with a bed, curtain, backpack, and doll. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the spotless room layout, clean bed, open curtain, and nighttime interior lighting from the reference. Exclude the previous empty composition as the sole focus and add only the pale young woman standing motionless in the middle of the room; do not restore any corpse or blood.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The clean room remains devoid of Minsuk's corpse, bloodstains, and the red wall mark; the backpack and doll remain subtly displaced. Suriyoung still carries her smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 3, "B": 7, "C": 5 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "제거해야 할 벽의 붉은 자국이 그대로 남아있으며 캐릭터의 스카프와 스마트폰이 누락되었습니다." }, { "label": "B", "score": 7, "verdict_ko": "깨끗해진 방의 상태, 캐릭터의 복장(스카프 포함), 스마트폰 파지 및 찌푸린 표정까지 지시사항을 모두 정확히 구현했습니다." }, { "label": "C", "score": 5, "verdict_ko": "방의 상태와 스마트폰 파지는 지시를 잘 따랐으나 레퍼런스에 있는 캐릭터의 스카프가 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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B04.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S14sh2_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영의 미간이 잔뜩 찌푸려져 있지 않고 평온한 표정임.", "fix_en": "Make Suriyoung deeply furrow her brow to show a heavy frown." }, { "issue_ko": "수리영의 얼굴이 혈색 없이 창백하지 않음.", "fix_en": "Make Suriyoung's face noticeably pale and bloodless." }, { "issue_ko": "레퍼런스 이미지의 카메라 구도를 그대로 복사하여 금지 사항을 위반함.", "fix_en": "Change the camera framing, angle, or distance so it does not identically copy the reference photograph's composition." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Make Suriyoung deeply furrow her brow to show a heavy frown.\n- Make Suriyoung's face noticeably pale and bloodless.\n- Change the camera framing, angle, or distance so it does not identically copy the reference photograph's composition.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+prev+엔티티" }, "S15sh2": { "input_fingerprint": "3a902ba3ed6e451a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 혜수의 손이 차가운 캔 음료수를 수리영의 가슴 앞쪽 허공으로 불쑥 내민 정지 순간.\n\nLOCATION (lock): A coastal bus-stop area with seating oriented toward the sea. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung still has her smartphone after repeatedly trying to reach Minsuk, while Hyesu is handing her the purchased canned drink.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 자연스러운 검은 머리); 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 팔짱 낀 포즈와 해안가 배경, 소지품(스마트폰, 캔) 및 의상 조건을 가장 충실하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "배경과 인물 의상은 지침에 부합하나, 스토리보드에 명시된 수리영의 팔짱 낀 포즈를 누락했습니다." }, { "label": "B", "score": 3, "verdict_ko": "배경이 바다가 아닌 도로를 향해 있고, 금지된 텍스트가 포함되었으며 인물의 의상도 참조와 다릅니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S15sh2.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "혜수가 캔을 쥐고 있는 오른손의 손가락 개수가 정상보다 많게(캔 앞면을 감싼 손가락이 5개) 그려져 해부학적 오류가 있습니다.", "fix_en": "Correct the anatomy of Hyesu's right hand so that exactly four fingers are wrapping around the front of the can." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Correct the anatomy of Hyesu's right hand so that exactly four fingers are wrapping around the front of the can.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S15sh5": { "input_fingerprint": "4be32b338f714fec", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 거리를 둔 채 한쪽 다리를 들고 달리는 자세인 수리영의 뒷모습 너머, 입가에 양손을 나팔 모양으로 모으고 입을 벌린 채 서 있는 혜수의 상체가 원경으로 보이는 구도.\n\nLOCATION (lock): A coastal bus-stop area with seating oriented toward the sea. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung runs off with her smartphone still in her possession.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 자연스러운 검은 머리); 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "수리영의 지정된 의상(재킷, 청바지, 가방)과 스마트폰, 달리는 포즈 및 멀리서 외치는 혜수의 모습까지 모든 요소를 정확히 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "구도와 혜수의 행동은 맞지만, 전경의 수리영 의상이 레퍼런스와 전혀 다른 반팔 운동복으로 묘사되었습니다." }, { "label": "C", "score": 4, "verdict_ko": "배경과 구도는 적절하나, 수리영의 의상이 캐릭터 레퍼런스를 무시하고 임의의 검은색 운동복으로 변경되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S15sh5.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 혜수의 '상체'가 원경으로 보인다고 명시되어 있으나, 하체를 포함한 전신이 모두 드러나 있음.", "fix_en": "Obscure or frame Hyesu so that only her upper body is visible in the background." }, { "issue_ko": "뒤로 꺾여 들린 수리영의 왼쪽 발(신발) 형태가 뒤틀려 해부학적으로 불가능한 모양새를 하고 있음.", "fix_en": "Correct the anatomy of Suriyoung's lifted left foot so the boot naturally faces the proper direction for a bent running leg." }, { "issue_ko": "스마트폰을 쥐고 있는 수리영의 오른손 손가락들이 뭉개져 기형적으로 렌더링됨.", "fix_en": "Fix Suriyoung's right hand to show anatomically correct fingers cleanly gripping the smartphone." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Obscure or frame Hyesu so that only her upper body is visible in the background.\n- Correct the anatomy of Suriyoung's lifted left foot so the boot naturally faces the proper direction for a bent running leg.\n- Fix Suriyoung's right hand to show anatomically correct fingers cleanly gripping the smartphone.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S16sh1": { "input_fingerprint": "357c83cbaa3494a0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 경찰서 사무실 안, 입술을 꽉 다문 채 스마트폰을 귀에 바짝 밀착시키고 선 혜수의 굳은 옆얼굴.\n\nLOCATION (lock): A police office with work desks, telephones, and computer stations, largely empty except for the officer on night duty. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Hyesu continues carrying her smartphone and is still using Minsuk's saved number to call her without a response.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 6 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "캐릭터의 외모(단발머리)와 꽉 다문 입술 등 인물 묘사는 정확하나, 장소 레퍼런스의 카메라 구도를 그대로 복사하지 말라는 제한을 위반했습니다." }, { "label": "B", "score": 5, "verdict_ko": "레이아웃 스케치의 구도는 잘 따랐으나, 캐릭터 레퍼런스(단발머리)를 무시하고 포니테일로 묘사했으며 입술을 꽉 다문 표정도 아닙니다." }, { "label": "C", "score": 6, "verdict_ko": "캐릭터의 외모와 표정은 기준에 부합하지만, 스마트폰 뒷면의 디자인이 원본과 다르게 변형되었고 장소 구도 제한도 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L11B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S16sh1.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스마트폰을 쥐고 있는 오른손의 구조가 해부학적으로 뒤틀려 있어, 손바닥 측면과 손가락 안쪽이 카메라를 향하고 있습니다.", "fix_en": "Redraw the right hand holding the smartphone so the back of the hand faces the camera, with the fingers naturally wrapping around the far edge of the device." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redraw the right hand holding the smartphone so the back of the hand faces the camera, with the fingers naturally wrapping around the far edge of the device.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S17sh1": { "input_fingerprint": "1fa79340ce0842e9", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dawn, rooftop room light on.\n\nSHOT TEXT (authoritative, Korean): 여명이 밝아오는 푸르스름한 새벽빛 아래, 불 켜진 옥탑방 창문을 향해 고개를 돌린 검은 고양이가 시멘트 담벼락 위에 네 발로 서 있는 실루엣.\n\nLOCATION (lock): The exterior rooftop area around a small rooftop apartment, bordered by a wall that a cat can cross. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop room's light remains switched on through dawn.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "A": 5, "C": 4 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "창문을 향해 고개를 돌린 고양이의 실루엣과 푸르스름한 새벽빛, 지정된 건축물 구조를 완벽하게 재현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "고양이가 창문이 아닌 반대 방향을 바라보고 있으며, 주변 배경의 디테일이 구조 레퍼런스와 다소 어긋납니다." }, { "label": "C", "score": 4, "verdict_ko": "고양이가 창문을 향하지 않고 카메라를 응시하고 있어 샷의 핵심 행동 지문을 명백히 위반했습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S17sh1.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_rooftop_residence_sel.png" } ], "critique": { "issues": [ { "issue_ko": "고양이의 뒷다리 하나가 누락되어 '네 발로 서 있는' 지시가 지켜지지 않았습니다.", "fix_en": "Add the missing fourth leg (back left leg) to the cat's silhouette so it stands on all fours." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the missing fourth leg (back left leg) to the cat's silhouette so it stands on all fours.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed" }, "S18sh1": { "input_fingerprint": "f77880be086b0410", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 밝은 낮, 안방 침대 모서리에 이마를 기댄 채 두 눈을 감고 앉아 있는 수리영의 상체.\n\nLOCATION (lock): A compact rooftop apartment containing a main bedroom, a meticulously ordered second bedroom with a curtained window, a bathroom with a mirror, and a living and dining area displaying a family photograph. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The house remains vacant, and Suriyoung's room is still unnaturally clean with no corpse or blood present. Her doll-adorned backpack remains in the house until she takes it out later in the scene.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 6, "A": 4 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "침대 모서리에 이마를 기댄 자세와 감은 눈을 정확히 연출했으며, 인물의 헤어스타일도 레퍼런스와 가장 잘 일치합니다." }, { "label": "B", "score": 6, "verdict_ko": "지정된 자세와 배경은 충실히 구현했으나, 인물의 머리를 묶고 있어 캐릭터 레퍼런스와 다소 차이가 있습니다." }, { "label": "A", "score": 4, "verdict_ko": "이마가 아닌 뺨을 침대에 기대고 있어 샷 텍스트가 명시한 핵심 자세(이마를 기댄 채)를 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B03.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S18sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [] }, "fix_skipped": true, "ref_mode": "플레이트+콘티+엔티티" }, "S18sh8": { "input_fingerprint": "0553aa67e6f07e6c", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 거울에 반사된 욕실 타일 벽면 한가운데, 붉은색 원형 표식이 선명하게 떠 있는 시각적 찰나.\n\nLOCATION (lock): A compact rooftop apartment containing a main bedroom, a meticulously ordered second bedroom with a curtained window, a bathroom with a mirror, and a living and dining area displaying a family photograph. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung now has her doll-adorned backpack with her, along with her smartphone. The red circle on the bathroom wall is only a momentary apparition and leaves no lasting mark.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 6, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지정된 욕실 구조와 거울 반사면을 정확하게 재현했으며, 붉은색 원형 표식을 공간 내에 나타난 입체적인 환영으로 적절히 연출했습니다." }, { "label": "B", "score": 6, "verdict_ko": "공간 구조는 레퍼런스와 완벽히 일치하나, 거울 속 붉은 표식이 공간 안의 환영이라기보다는 이미지 위에 그려진 2D 주석처럼 보입니다." }, { "label": "C", "score": 4, "verdict_ko": "이른 아침의 채광은 훌륭하나, 거울에 반사된 맞은편 벽면에 존재하지 않는 개방된 문이 나타나는 등 공간 구조가 왜곡되었습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B01.png" } ], "critique": { "issues": [ { "issue_ko": "거울 속에 반사된 벽면이 텍스트의 명시('욕실 타일 벽면')와 달리 타일 질감이 없는 매끄러운 표면입니다.", "fix_en": "Add visible tile lines and grout to the wall surface reflected inside the mirror." }, { "issue_ko": "붉은색 표식이 반사된 벽에 있는 것이 아니라 거울 표면(또는 그 앞)에 떠 있는 발광체로 묘사되어, 거울 밖 오른쪽 타일 벽에 붉은 빛을 반사하고 있습니다.", "fix_en": "Place the red circular mark flatly onto the reflected wall inside the mirror, and remove the red ambient light it casts onto the physical tile wall to the right of the mirror." }, { "issue_ko": "왼쪽 문의 패널 무늬가 세로형인 참고 사진과 달리 가로형으로 다르게 생성되었습니다.", "fix_en": "Change the panels on the left white door to vertical rectangles to match the door design in the reference photograph." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add visible tile lines and grout to the wall surface reflected inside the mirror.\n- Place the red circular mark flatly onto the reflected wall inside the mirror, and remove the red ambient light it casts onto the physical tile wall to the right of the mirror.\n- Change the panels on the left white door to vertical rectangles to match the door design in the reference photograph.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트만 (배경 전용)" }, "S18sh11": { "input_fingerprint": "1ec4da3d3e8d6b0c", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 식탁 위에 놓인 사진 액자 속 평면 이미지 안에서, (강민숙(민숙))의 웃는 얼굴 형태가 기괴하게 뭉개져 흘러내린 기괴한 상태.\n\nLOCATION (lock): A compact rooftop apartment containing a main bedroom, a meticulously ordered second bedroom with a curtained window, a bathroom with a mirror, and a living and dining area displaying a family photograph. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung continues to carry the doll-adorned backpack and her smartphone. The framed photograph itself remains normal after the momentary distortion vanishes.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "지정된 로케이션(창틀, 질감, 의자 등)을 정확히 재현했으며, 물리적 액자가 아닌 2D 이미지 내부의 얼굴 왜곡만을 완벽하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "사진 속 얼굴이 흘러내리는 묘사는 적절하나, 레퍼런스 장소에 없는 탁상조명과 커튼을 임의로 추가해 로케이션 일치도가 떨어집니다." }, { "label": "B", "score": 4, "verdict_ko": "레퍼런스와 완전히 다른 커튼 배경을 사용했고, 2D 이미지가 아닌 물리적 액자 프레임 자체를 녹아내리게 연출하여 지시를 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B01.png" } ], "critique": { "issues": [] }, "fix_skipped": true, "ref_mode": "플레이트만 (배경 전용)" }, "S19sh1": { "input_fingerprint": "836f3de546f37612", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, supermarket interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 이른 아침 한산한 마트 안, 마트 남자 직원을 마주보고 꼿꼿이 서 있는 수리영의 정면.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with customer aisles, staff areas, and an exterior wall where surveillance-camera coverage can be checked. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung's bicycle remains parked outside the mart, while her doll-adorned backpack and smartphone stay with her.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 4, "B": 3, "C": 7 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "A", "score": 4, "verdict_ko": "레이아웃 스케치를 따르느라 텍스트가 명시한 '수리영의 정면' 지시를 위반하고 뒷모습을 연출했습니다." }, { "label": "B", "score": 3, "verdict_ko": "직원을 마주보지 않는 구도이며, 금지된 영문 텍스트가 포함되었고 배낭을 앞으로 멘 어색한 묘사가 있습니다." }, { "label": "C", "score": 7, "verdict_ko": "'수리영의 정면', 직원을 마주보는 앵글, 지정된 소지품(스마트폰, 인형 달린 배낭)과 장소 요건을 완벽하게 구현했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L08B04.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S19sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영이 쥐고 있는 스마트폰의 화면(기능면)이 지문 지시 없이 카메라를 향해 노출되어 있습니다.", "fix_en": "Turn the smartphone in her hand so its back faces the camera, hiding the screen." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Turn the smartphone in her hand so its back faces the camera, hiding the screen.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S19sh4": { "input_fingerprint": "b2f1e3a0beaabd3d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, supermarket interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 마트 밖에서 스마트폰을 귀에 댄 채 시선을 허공에 둔 수리영의 옆얼굴.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with customer aisles, staff areas, and an exterior wall where surveillance-camera coverage can be checked. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung continues carrying her backpack and smartphone while calling Hyesu; her bicycle remains parked at the mart.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "캐릭터 의상, 스마트폰 소품, 마트 배경 및 지정된 포즈를 참조 이미지에 맞게 충실히 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "캐릭터의 가방이 배경의 파란 상자와 하나로 합쳐지는 심각한 형태 오류가 발생했습니다." }, { "label": "C", "score": 5, "verdict_ko": "지시된 재킷 대신 회색 후드티를 입고 있어 캐릭터 의상 참조를 명백히 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L08B03.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S19sh4.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "레이아웃 스케치는 인물이 우측에서 왼쪽을 바라보는 포즈를 지시하나, 이미지에서는 인물이 중앙에서 오른쪽을 바라보고 있습니다.", "fix_en": "Move the character to the right side of the frame and have her face left, strictly following the pose and placement of the layout sketch." }, { "issue_ko": "스마트폰의 후면 카메라 렌즈가 기기의 아래쪽(턱 부근)에 위치해 있어 전화기가 위아래로 뒤집혀 있습니다.", "fix_en": "Flip the smartphone vertically so the rear camera lenses are at the top, correcting its real-world orientation." }, { "issue_ko": "로케이션 레퍼런스의 좌측에 있는 주택가 골목과 붉은 벽돌집들이 셔터가 내려진 회색 상가 건물로 잘못 변형되었습니다.", "fix_en": "Revert the background on the left side of the mart to the narrow alleyway and brick residential buildings seen in the location reference." }, { "issue_ko": "캐릭터 레퍼런스에서 착용하고 있는 초록색 무늬의 목수건과 회색 안쪽 셔츠가 누락되었습니다.", "fix_en": "Add the green patterned neckerchief and grey inner shirt visible under the brown jacket to match the character reference exactly." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Move the character to the right side of the frame and have her face left, strictly following the pose and placement of the layout sketch.\n- Flip the smartphone vertically so the rear camera lenses are at the top, correcting its real-world orientation.\n- Revert the background on the left side of the mart to the narrow alleyway and brick residential buildings seen in the location reference.\n- Add the green patterned neckerchief and grey inner shirt visible under the brown jacket to match the character reference exactly.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S20sh3": { "input_fingerprint": "7bb64a668de68d93", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, CCTV monitor glow and otherwise unspecified interior lighting.\n\nSHOT TEXT (authoritative, Korean): 혜수와 나란히 서서 모니터를 노려보며 미간을 심하게 찌푸린 수리영의 불안한 얼굴.\n\nLOCATION (lock): A supermarket administrative and security office with a surveillance workstation and monitor for reviewing recorded camera footage. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung still has her doll-adorned backpack and smartphone with her during the CCTV review.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 자연스러운 검은 머리); 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 6, "B": 5, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 6, "verdict_ko": "모니터를 노려보는 행동과 장소 구조를 잘 구현했으나, 지시를 어기고 레퍼런스 사진의 배경 인물까지 그대로 포함함." }, { "label": "B", "score": 5, "verdict_ko": "인물들의 표정과 시선 처리는 정확하지만, 고정 요건인 장소의 건축 구조와 모니터 배치를 크게 왜곡함." }, { "label": "C", "score": 4, "verdict_ko": "장소와 인물 외형은 완벽하게 일치하나, 수리영이 모니터가 아닌 뒤를 돌아보고 있어 핵심 행동 묘사에 실패함." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L12B01.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S20sh3.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지문에 명시되지 않은 인물이 배경 우측 출입구 쪽에 그대로 남아있습니다 (지문 외 인물 추가 금지 위반).", "fix_en": "Remove the unrecognized figure standing in the background doorway on the far right." }, { "issue_ko": "지문에서 수리영이 '미간을 심하게 찌푸린' 얼굴이어야 한다고 명시했으나, 미간이 매끄럽고 단순한 근심 어린 표정에 가깝습니다.", "fix_en": "Deeply crease the space between Suriyoung's eyebrows to create a severe, heavy frown." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the unrecognized figure standing in the background doorway on the far right.\n- Deeply crease the space between Suriyoung's eyebrows to create a severe, heavy frown.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S21sh5": { "input_fingerprint": "1357c85293c0ce9d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 책상 위 사진들을 향해 시선을 던진 채 두 눈을 부릅뜬 수리영의 굳은 표정.\n\nLOCATION (lock): A busy police station office with clustered desks, computer monitors, paperwork, and enlarged crime-scene photographs spread across a detective’s desk. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The enlarged crime-scene photographs remain scattered across Hyesu's desk. Suriyoung continues carrying her backpack and smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 3, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "레이아웃이 지시한 자세(가방 끈과 스마트폰을 쥔 양손)와 '두 눈을 부릅뜬 굳은 표정'을 가장 충실히 재현했으며, 인물 배제 규칙도 잘 지켰습니다." }, { "label": "B", "score": 3, "verdict_ko": "자세와 표정은 비교적 양호하나, 프롬프트에 없는 여러 명의 경찰서 배경 인물을 임의로 추가하여 엄격한 인물 배제 규칙을 위반했습니다." }, { "label": "C", "score": 5, "verdict_ko": "캐릭터 레퍼런스의 재킷 형태는 가장 잘 일치하나, 지정된 양손의 자세를 전혀 따르지 않았고 '두 눈을 부릅뜬' 감정 표현도 부족합니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L11B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S21sh5.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "모니터 뒤 파티션에 붙은 종이에 금지된 텍스트(문자)가 생성됨.", "fix_en": "Remove all text from the paper pinned to the cubicle partition behind the monitor, leaving it blank." }, { "issue_ko": "캐릭터가 레퍼런스 사진의 녹색 스카프를 착용하지 않았으며, 재킷의 디자인이 다름.", "fix_en": "Add the green neckerchief and change the character's jacket to the brown button-up field jacket shown in the character reference." }, { "issue_ko": "왼쪽 벽에 걸린 시계에 숫자와 바늘이 없는 백지 상태로 묘사됨.", "fix_en": "Add clock hands and numbers to the blank wall clock on the left wall to match the location reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text from the paper pinned to the cubicle partition behind the monitor, leaving it blank.\n- Add the green neckerchief and change the character's jacket to the brown button-up field jacket shown in the character reference.\n- Add clock hands and numbers to the blank wall clock on the left wall to match the location reference.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S21sh6": { "input_fingerprint": "52ddccc8b248e973", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 책상 위 컴퓨터 모니터 화면에 바다 위 좌표가 붉게 표시된 해상 지도의 멈춘 상태.\n\nLOCATION (lock): A busy police station office with clustered desks, computer monitors, paperwork, and enlarged crime-scene photographs spread across a detective’s desk. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The location-result window remains open on Kim's monitor, showing Minsuk's last coordinates over the sea. The crime-scene photographs remain spread across the nearby desk.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "해상 지도 위 바다에 붉은 좌표가 정확히 표시되었고, 빈 사무실과 책상 위 사건 사진 등 요구사항을 가장 충실히 구현했습니다." }, { "label": "B", "score": 5, "verdict_ko": "기준 장소의 분위기는 잘 살렸으나, 모니터 상의 붉은 좌표가 바다가 아닌 육지(내륙)에 표시되어 지시사항을 어겼습니다." }, { "label": "C", "score": 4, "verdict_ko": "바다 위 좌표는 맞혔으나, 금지된 텍스트(영어 문구)가 화면과 사진에 노골적으로 생성되었고 지도가 다소 비현실적입니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L11B02.png" } ], "critique": { "issues": [ { "issue_ko": "'텍스트 없음' 규칙에도 불구하고 모니터 화면의 지도 팝업창에 'popped state'라는 영어 텍스트와 책상 위 서류들에 의미 없는 글자들이 생성되었습니다.", "fix_en": "Remove the text 'popped state' from the monitor's pop-up box and all gibberish text from the documents on the desk." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the text 'popped state' from the monitor's pop-up box and all gibberish text from the documents on the desk.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트만 (배경 전용)" }, "S21sh10": { "input_fingerprint": "001cc84a536ca333", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 반으로 접힌 메모지를 쥔 수리영의 손이 바지 주머니 틈새에 반쯤 끼워진 정적 구도.\n\nLOCATION (lock): A busy police station office with clustered desks, computer monitors, paperwork, and enlarged crime-scene photographs spread across a detective’s desk. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the police-office workstation look, daylight, desk clutter, and the young woman's same clothing from the reference. Exclude her face, the enlarged crime photographs, and the computer screen from the insert; show only her hand slipping the folded note into her trouser pocket.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The folded note containing Minsuk's coordinates is placed in Suriyoung's pocket and remains with her. She also retains her backpack and smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "지시된 대로 화면에서 컴퓨터 모니터를 제외하고 텍스트 없는 메모지를 주머니에 넣는 정적 구도를 가장 잘 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "메모지를 넣는 동작은 좋으나, 프레임에서 배제해야 할 컴퓨터 화면과 범죄 사진이 배경에 뚜렷하게 나타납니다." }, { "label": "B", "score": 3, "verdict_ko": "금지된 텍스트가 메모지에 적혀 있으며, 배제해야 할 컴퓨터 화면의 일부가 포함되어 프롬프트 지침을 크게 위반했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L11B02.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S21sh5_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지시문에서 삽입 샷(insert)에서 확대된 범죄 현장 사진을 제외하라고 명시했으나, 배경의 책상 위에 사진들이 흐릿하게 보입니다.", "fix_en": "Remove the scattered crime scene photographs from the desk in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the scattered crime scene photographs from the desk in the background.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+prev+엔티티" }, "S22sh1": { "input_fingerprint": "a4a99fad8f709541", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): late morning.\n\nSHOT TEXT (authoritative, Korean): 밝은 낮, 야외에서 자전거 안장에 앉아 한쪽 무릎을 굽혀 페달을 밟아 내리는 주행 중간 동작으로, 머리카락이 바람에 뒤로 흩날리는 수리영의 측면 모습.\n\nLOCATION (lock): The exterior parking area of a police station, with a bicycle parked for immediate departure. The shot takes place here — the attached STRUCTURE LOOK photograph shows the fixed structure of this exact place — the structure's shape, materials, colors and openings are LOCKED to it. Build the remaining surroundings strictly from the location text above and the shot text, and take staging and camera from the attached STORYBOARD SKETCH.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung rides away on her bicycle with the folded coordinate note in her pocket, while retaining her backpack and smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "C": 4, "B": 3 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지정된 구조물 레퍼런스와 주행 동작을 정확히 재현했으며, 주머니 속 쪽지와 배낭 등 소지품 지시를 가장 잘 따랐습니다." }, { "label": "C", "score": 4, "verdict_ko": "경찰서라는 텍스트에 치중하여 형태와 재질이 고정된 건물 레퍼런스를 완전히 다른 건물로 변형하는 위반을 범했습니다." }, { "label": "B", "score": 3, "verdict_ko": "자전거 프레임이 인체의 다리를 관통하는 치명적인 물리적 오류가 있으며 스마트폰 위치도 부자연스럽습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S22sh1.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text is the sole authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/background_chain/seed_bg_coastal_road_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "우측 가장자리에 있는 표지판에 문자가 포함되어 있습니다 (텍스트 금지 지시 위반).", "fix_en": "Remove all text from the signpost on the right edge of the frame." }, { "issue_ko": "자전거의 오른쪽 페달이 없어 인물의 신발이 페달 대(크랭크 암)를 직접 밟고 있습니다.", "fix_en": "Add a bicycle pedal between the character's right boot and the crank arm." }, { "issue_ko": "캐릭터 레퍼런스 이미지에 있는 초록색 목 스카프가 누락되었습니다.", "fix_en": "Add the green neck scarf from the character reference to the character." }, { "issue_ko": "캐릭터 레퍼런스 이미지에 있는 검은색 미니 크로스백과 스트랩이 누락되었습니다.", "fix_en": "Add the small black cross-body bag and its strap across the character's chest." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text from the signpost on the right edge of the frame.\n- Add a bicycle pedal between the character's right boot and the crank arm.\n- Add the green neck scarf from the character reference to the character.\n- Add the small black cross-body bag and its strap across the character's chest.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(structure_plate): 스케치+seed+엔티티" }, "S23sh1": { "input_fingerprint": "a1c1a8883eed5851", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, monitor glow and otherwise unspecified interior lighting.\n\nSHOT TEXT (authoritative, Korean): 어두운 PC방 모니터 불빛에 비친 수리영의 무표정한 얼굴 클로즈업.\n\nLOCATION (lock): A public computer café with individual computer stations, monitors, and access to a printer. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung retains the folded coordinate note in her pocket and keeps her backpack and smartphone with her while checking the coordinates.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 4, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "지시된 무표정한 얼굴과 레퍼런스의 의상을 잘 재현했으며, 금지된 텍스트나 추가 인물 없이 조건을 가장 잘 충족했습니다." }, { "label": "B", "score": 4, "verdict_ko": "얼굴 묘사는 무난하나 레퍼런스의 의상이 누락되었고, 배경에 프롬프트에 없는 다른 인물이 등장하여 감점되었습니다." }, { "label": "A", "score": 3, "verdict_ko": "클로즈업 구도는 적절하지만, 배경 모니터에 금지된 텍스트가 명확히 출력되었고 배경에 다른 인물들이 포함되어 규칙을 위반했습니다." } ], "refs": [ { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐리드 스테이트와 레퍼런스에 있는 검은색 크로스백 가방 끈이 캐릭터의 어깨와 가슴 부위에 없습니다.", "fix_en": "Add the black bag strap diagonally across her right shoulder and chest over the jacket." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the black bag strap diagonally across her right shoulder and chest over the jacket.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S23sh2": { "input_fingerprint": "4cc2d55a2be48f49", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, monitor glow and otherwise unspecified interior lighting.\n\nSHOT TEXT (authoritative, Korean): 모니터 화면 안 해상 지도 위에 붉은 점이 찍혀 있는 좌표 지점의 정지 컷.\n\nLOCATION (lock): A public computer café with individual computer stations, monitors, and access to a printer. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The folded coordinate note remains in Suriyoung's pocket, and the newly printed maps are folded and placed inside her doll-adorned backpack.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 4, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "자막이나 텍스트를 추가하지 말라는 지시를 유일하게 준수했으며, 해상 지도와 붉은 점이 잘 묘사되었습니다." }, { "label": "A", "score": 4, "verdict_ko": "화면 하단에 지문 텍스트가 그대로 자막처럼 출력되어 '텍스트 금지' 규칙을 위반했습니다." }, { "label": "B", "score": 3, "verdict_ko": "모니터 화면 한가운데에 지문 텍스트가 삽입되어 '텍스트 금지' 규칙을 심각하게 위반했습니다." } ], "refs": [], "critique": { "issues": [ { "issue_ko": "텍스트 생성 금지 지시를 어기고 모니터 화면 안 지도 위에 글자와 숫자들(예: '붉은 점')이 생성됨.", "fix_en": "Remove all text, numbers, and characters from the map displayed on the monitor." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text, numbers, and characters from the map displayed on the monitor.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트만 (배경 전용)" }, "S24sh1": { "input_fingerprint": "84d8b9341c5a60d1", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, thunderstorm and heavy rain.\n\nSHOT TEXT (authoritative, Korean): 짙은 먹구름과 세찬 비가 쏟아지는 해안가 도로 위, 빗물에 젖은 수리영이 자전거 안장에 앉아 한쪽 페달을 힘껏 밟아 누른 채 핸들을 꽉 쥔 전신 실루엣, 옷자락과 빗방울이 뒤로 흩날리는 찰나.\n\nLOCATION (lock): An exposed coastal route beside the sea, traversed by bicycle during heavy rain. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung rides through the rain with the printed maps secured in her backpack and the coordinate note still in her pocket; she also retains her smartphone.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 구도와 자전거 형태(바구니 포함)를 정확히 따르며, 캐릭터 의상과 날씨 지시를 충실히 구현함." }, { "label": "A", "score": 5, "verdict_ko": "캐릭터와 날씨 표현은 좋으나, 자전거 바구니가 누락되었고 스토리보드와 자전거 프레임 형태가 다름." }, { "label": "B", "score": 3, "verdict_ko": "참조 이미지와 다른 파란색 재킷을 입고 있으며, 우측 표지판에 금지된 텍스트가 포함됨." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S24sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 안장에 앉아있어야 한다는 프롬프트 지시와 달리, 엉덩이를 떼고 페달 위에 서서 자전거를 타고 있습니다.", "fix_en": "Lower the character's hips so she is fully seated resting on the bicycle saddle." }, { "issue_ko": "지도가 백팩 안에 보관되어야 한다는 지시와 달리, 가방 바깥쪽에 정체불명의 흰색 직사각형 물건이 꽂혀 있습니다.", "fix_en": "Remove the white rectangular object sticking out from the side of the backpack." }, { "issue_ko": "자전거 핸들을 잡고 있는 손이 심하게 뭉개져 손잡이의 금속 부분과 하나로 융합되어 있습니다.", "fix_en": "Redraw the hand on the handlebar to show clearly separated, anatomically correct fingers gripping the metal." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Lower the character's hips so she is fully seated resting on the bicycle saddle.\n- Remove the white rectangular object sticking out from the side of the backpack.\n- Redraw the hand on the handlebar to show clearly separated, anatomically correct fingers gripping the metal.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S25sh4": { "input_fingerprint": "ac75837f1fb15472", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior light from an open window.\n\nSHOT TEXT (authoritative, Korean): 열린 방문 너머, 방 안 벽면에 그려진 붉은색 원형 표식과 그 가운데 꽂힌 낡은 사진 구도.\n\nLOCATION (lock): A ransacked rooftop apartment with an open living-room window, overturned dining chairs, scattered household items, and muddy footprints. A bedroom wall bears a red circle with a bloodstained photograph fixed at its center. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The apartment remains ransacked, with intruder footprints, an open living-room window, and overturned dining chairs. A red circle is fixed on the bedroom wall with the bloodied old photograph pinned at its center; Suriyoung still has her backpack, maps, coordinate note, and smartphone.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 4 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "거실의 열린 창문, 뒤집힌 의자, 발자국을 묘사하며 열린 방문 너머 침실의 붉은 표식을 바라보는 구도를 정확히 구현했습니다." }, { "label": "B", "score": 5, "verdict_ko": "열린 방문 너머의 시점을 시도했으나, 거실과 침실의 공간 구분이 모호하고 건축 구조가 하나로 이어져 부자연스럽습니다." }, { "label": "C", "score": 4, "verdict_ko": "화면 중앙을 가로지르는 정체불명의 구조물로 인해 공간이 기형적으로 분할되어 요구된 카메라 구도를 심각하게 훼손했습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B04.png" } ], "critique": { "issues": [ { "issue_ko": "침실 안쪽 벽에 존재하지 않는 단차(기둥 또는 벽의 갈라짐)가 생성되어, 붉은색 원과 사진이 비정상적으로 꺾이거나 나뉘어 물리 법칙과 레퍼런스의 공간 구조에 어긋납니다.", "fix_en": "Remove the architectural glitch splitting the bedroom's back wall; render it as one continuous, flat surface with the red circle and photograph seamlessly intact on it." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the architectural glitch splitting the bedroom's back wall; render it as one continuous, flat surface with the red circle and photograph seamlessly intact on it.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트만 (배경 전용)" }, "S25sh8": { "input_fingerprint": "b90f970e231bb0a6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior light from an open window.\n\nSHOT TEXT (authoritative, Korean): 한 손에 구겨진 사진을 꽉 쥔 채 백팩을 한쪽 어깨에 걸치고 선 수리영의 뒷모습.\n\nLOCATION (lock): A ransacked rooftop apartment with an open living-room window, overturned dining chairs, scattered household items, and muddy footprints. A bedroom wall bears a red circle with a bloodstained photograph fixed at its center. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the ransacked room, open doorway, red circle on the wall, embedded photograph, and dusk lighting from the reference. Exclude the photograph remaining pinned to the wall and exclude an empty room; show the young woman holding the removed photograph and wearing her backpack as she prepares to leave.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung leaves wearing her backpack, with the printed maps inside and the coordinate note still in her pocket, while tightly holding the removed bloodied photograph. The apartment remains ransacked, and the red circle remains on the wall after the photograph is removed.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 5, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "백팩을 한쪽 어깨에 걸치고 구겨진 피 묻은 사진을 꽉 쥔 뒷모습(측후면)을 가장 정확하게 구현했습니다." }, { "label": "B", "score": 5, "verdict_ko": "뒷모습과 구겨진 사진의 표현은 우수하나, 백팩을 양쪽 어깨에 모두 메고 있어 지시어와 어긋납니다." }, { "label": "A", "score": 3, "verdict_ko": "백팩을 한쪽에 멨으나 사진이 전혀 구겨지지 않았고, 백팩 밖으로 지도가 튀어나와 묘사가 부정확합니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B05.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S25sh4_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 금월 식당과 두 어린 여자아이가 찍힌 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "카메라가 인물의 측면(옆모습)을 보여주고 있으며, 샷 텍스트에 명시된 '뒷모습'이 아닙니다.", "fix_en": "Rotate the character so she is facing entirely away from the camera to show her back view." }, { "issue_ko": "인물이 손에 쥔 물건이 제공된 소품 레퍼런스의 낡은 사진이 아니라 피 묻은 천 조각처럼 보입니다.", "fix_en": "Replace the cloth-like object in her hand with a crumpled, bloodstained version of the referenced vintage photograph." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the character so she is facing entirely away from the camera to show her back view.\n- Replace the cloth-like object in her hand with a crumpled, bloodstained version of the referenced vintage photograph.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+prev+엔티티" }, "S26sh2": { "input_fingerprint": "a9ab363cdd40e6ea", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, dense fog after rain.\n\nSHOT TEXT (authoritative, Korean): 안개 낀 바다를 향해 젖은 갯바위 위에 무릎을 세우고 앉은 수리영의 측면.\n\nLOCATION (lock): An empty harbor with dense fog obscuring the boats, a rocky shoreline for sitting, and a mooring point where a single small boat arrives. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung still carries her smartphone, the old photograph, the coordinate note, and the backpack containing the printed maps.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 8, "B": 4, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 8, "verdict_ko": "지정된 소지품(스마트폰, 사진, 메모, 지도 배낭)과 무릎을 세운 자세, 안개 낀 배경을 가장 정확하게 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "스마트폰, 사진, 메모 등 필수 소지품이 완전히 누락되었으며 무릎을 세운 자세의 묘사도 부족합니다." }, { "label": "C", "score": 5, "verdict_ko": "소지품을 배낭 주머니에 비현실적으로 꽂아두었으며, 스토리보드와 달리 배낭을 메고 있어 설정에 어긋납니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S26sh2.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터가 무릎을 세우고 앉지 않고 다리를 아래로 뻗고 있어 스토리보드 및 프롬프트의 자세 지시와 다릅니다.", "fix_en": "Pose the character with her knees drawn up and feet resting flat on the rock to match the storyboard." }, { "issue_ko": "지도가 프롬프트의 지시사항과 다르게 배낭 안이 아닌 배낭 겉면에 올려져 있습니다.", "fix_en": "Place the printed maps completely inside the backpack so they do not rest on the outside." }, { "issue_ko": "캐릭터 레퍼런스 이미지에 명확히 보이는 검은색 소형 크로스백이 누락되었습니다.", "fix_en": "Add the black crossbody bag and strap worn across the character's chest from the reference." }, { "issue_ko": "스마트폰과 종이들을 쥐고 있는 손의 해부학적 구조가 뭉개졌으며, 사물들이 손가락을 관통하고 있습니다.", "fix_en": "Render an anatomically correct hand properly gripping the phone, photograph, and note without clipping." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Pose the character with her knees drawn up and feet resting flat on the rock to match the storyboard.\n- Place the printed maps completely inside the backpack so they do not rest on the outside.\n- Add the black crossbody bag and strap worn across the character's chest from the reference.\n- Render an anatomically correct hand properly gripping the phone, photograph, and note without clipping.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S26sh6": { "input_fingerprint": "a2d2fcecc5766ad4", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, dense fog after rain.\n\nSHOT TEXT (authoritative, Korean): 포구에 닿은 배 바로 앞 바닥에 선 수리영이 위쪽의 인우를 향해 고개를 든 정지 상태.\n\nLOCATION (lock): An empty harbor with dense fog obscuring the boats, a rocky shoreline for sitting, and a mooring point where a single small boat arrives. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung retains her smartphone, the old photograph, the coordinate note, and her map-filled backpack as she prepares to board Inwoo's boat.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 자연스러운 검은 머리); 인우 (한국인 남성, 20대 중반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 6 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "배경의 배를 가리는 짙은 안개와 수리영이 쥔 소지품(스마트폰, 사진, 메모)이 프롬프트 지시사항을 가장 충실히 반영했습니다." }, { "label": "B", "score": 5, "verdict_ko": "배경의 배들이 선명하게 보여 '빈 항구' 설정에 어긋나며, 수리영이 들어야 할 사진과 메모가 누락되었습니다." }, { "label": "C", "score": 6, "verdict_ko": "수리영의 소지품은 일부 표현되었으나, 배경의 배들이 안개에 충분히 가려지지 않아 지시된 분위기와 다소 차이가 있습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S26sh6.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 인우: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인우의 다리가 배의 불투명한 외벽(선체)을 뚫고 앞에 겹쳐져 물리적으로 불가능한 구조를 보입니다.", "fix_en": "Render Inwoo's legs completely behind the solid bulwark of the boat so he is properly standing on the interior deck." }, { "issue_ko": "수리영이 스마트폰과 함께 종이를 한 장만 들고 있어, 프롬프트에 명시된 '좌표 메모(coordinate note)'가 누락되었습니다.", "fix_en": "Add a second distinct piece of paper in Suriyoung's hand to represent the missing coordinate note alongside the photograph." }, { "issue_ko": "수리영의 캐릭터 레퍼런스에 있는 목에 두른 초록색 패턴 스카프가 누락되었습니다.", "fix_en": "Add the green patterned scarf around Suriyoung's bare neck to match her character reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Render Inwoo's legs completely behind the solid bulwark of the boat so he is properly standing on the interior deck.\n- Add a second distinct piece of paper in Suriyoung's hand to represent the missing coordinate note alongside the photograph.\n- Add the green patterned scarf around Suriyoung's bare neck to match her character reference.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S27sh1": { "input_fingerprint": "888078308fe50b48", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 난장판이 된 옥탑방 거실 한가운데 서서 스마트폰을 귀에 댄 혜수의 초조한 얼굴.\n\nLOCATION (lock): A desolate, ransacked rooftop apartment with an unsecured entrance door rattling in the wind. The shot takes place here — the attached PREVIOUS SHOT STILL shows this exact place.\n\nTHIS SHOT CONTINUES THE PREVIOUS SHOT: everything the attached still established — the place, its fixed features and wear, each person's clothing and state — persists. Any person in it who cannot move stays PRECISELY as photographed (body, pose, contact points, held objects); only the camera changes.\n\nPREVIOUS STILL USAGE (follow exactly — what to take from the attached still and what to exclude): Take the same ransacked interior, displaced household objects, open spaces, and bleak dusk lighting from the reference. Exclude the departing young woman, her backpack, and the photograph in her hand; place the female officer alone in the messy living room while checking the phone message.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The rooftop apartment remains ransacked, with the window and displaced furnishings left as Suriyoung found them; the entrance door rattles in the wind. The photograph is no longer pinned to the bedroom wall, although the red circle remains there.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 여성, 40대 중반 얼굴, 성숙한 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "A": 5, "C": 4 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "명시된 침실 벽의 붉은 원, 검은색 스마트폰, 그리고 초조한 표정까지 모든 요구 사항을 가장 정확하게 반영했습니다." }, { "label": "A", "score": 5, "verdict_ko": "프롬프트에 명시된 침실 벽의 붉은 원이 누락되어 이전 샷과의 연속성을 상실했습니다." }, { "label": "C", "score": 4, "verdict_ko": "소품 레퍼런스와 명백히 다른 흰색 스마트폰이 사용되었고 의상에 불필요한 로고가 추가되었습니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L04B05.png" }, { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/scene/recipe/S25sh8_sel.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 스마트폰: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "이전 샷 레퍼런스에 있던 창가 쪽 나무 테이블과 의자가 거실에서 완전히 사라졌습니다.", "fix_en": "Restore the wooden table and chairs to the background exactly as they were positioned in the previous shot still." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Restore the wooden table and chairs to the background exactly as they were positioned in the previous shot still.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+prev+엔티티" }, "S28sh3": { "input_fingerprint": "aaa1cf016760e7ef", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 수리영의 한 손이 '금월'이라는 작은 글씨가 적힌 낡은 사진을 공중에 치켜든 클로즈업.\n\nLOCATION (lock): The open deck of a small fishing boat far out at sea, with an enclosed wheelhouse nearby. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung has the old photograph and one of the printed maps in hand or immediately at hand; the remaining maps stay in her backpack, and the coordinate note and smartphone remain with her.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "샷 텍스트에 따라 배경에 다른 인물을 추가하지 않았고 소품 사진의 원래 형태와 의상을 가장 잘 유지했으나, 명시된 지도가 보이지 않습니다." }, { "label": "B", "score": 5, "verdict_ko": "인물 제한 규칙은 잘 지켰으나, 기준과 달리 사진을 폴라로이드 형태로 변형하였고 의상 색상이 다소 다릅니다." }, { "label": "C", "score": 3, "verdict_ko": "스토리보드의 구도는 훌륭하게 따랐으나, 샷 텍스트에 언급되지 않은 인물(수리영의 복제)을 배경에 임의로 추가하여 치명적인 감점 요소가 되었습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S28sh3.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 금월 식당과 두 어린 여자아이가 찍힌 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지문에 명시된 '인쇄된 지도(printed map)'가 손이나 주변에 보이지 않습니다.", "fix_en": "Add one of the printed maps either in the hand holding the photograph or resting immediately nearby." }, { "issue_ko": "사진을 쥐고 있는 손의 해부학적 구조가 비정상적입니다(손등 방향인데 손가락 바닥 면이 앞으로 넘어와 있음).", "fix_en": "Fix the hand's anatomy so the fingers and thumb hold the photograph with a natural, anatomically correct grip." }, { "issue_ko": "조타실 창문 안쪽에 지문에 없는 사람의 실루엣이 임의로 추가되었습니다.", "fix_en": "Remove the dark silhouette of a person visible inside the wheelhouse window." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add one of the printed maps either in the hand holding the photograph or resting immediately nearby.\n- Fix the hand's anatomy so the fingers and thumb hold the photograph with a natural, anatomically correct grip.\n- Remove the dark silhouette of a person visible inside the wheelhouse window.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S28sh4": { "input_fingerprint": "07c5332706a6b716", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 눈앞의 사진에 시선을 고정한 채 두 눈을 크게 뜨고 얼어붙은 인우의 상체.\n\nLOCATION (lock): The open deck of a small fishing boat far out at sea, with an enclosed wheelhouse nearby. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The old photograph remains held out between Suriyoung and Inwoo, while her backpack, maps, coordinate note, and smartphone remain aboard with her.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 인우 (한국인 남성, 20대 중반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 6, "A": 5 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "프레임 밖 수영의 손이 사진을 든 구도와 인우의 표정, 소품 배치가 지시문과 가장 잘 일치합니다." }, { "label": "C", "score": 6, "verdict_ko": "사진을 든 앞쪽의 손이 여성(수영)이 아닌 굵은 남성의 손으로 묘사되어 설정과 어긋납니다." }, { "label": "A", "score": 5, "verdict_ko": "인우가 직접 사진을 들고 있어 두 사람 사이에 사진이 있다는 상호작용 지시를 누락했습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S28sh4.png" }, { "label": "CHARACTER REFERENCE — 인우: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 언급되지 않은 인물(수리영)의 두 손이 화면 하단에 등장함.", "fix_en": "Redraw the foreground hands holding the photograph so they are clearly Inwoo's own hands reaching up from his body." }, { "issue_ko": "소품(사진 뒷면)에 금지된 텍스트와 필기 흔적이 있음.", "fix_en": "Erase all text and handwriting scribbles from the back of the photograph." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redraw the foreground hands holding the photograph so they are clearly Inwoo's own hands reaching up from his body.\n- Erase all text and handwriting scribbles from the back of the photograph.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치+엔티티" }, "S29sh4": { "input_fingerprint": "6a92a86f551ade6a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영의 검지손가락이 낡은 사진 속 여자아이 형상 표면 위에 맞닿아 있는 정적 구도.\n\nLOCATION (lock): The enclosed wheelhouse of a small fishing boat, containing steering controls and numerous circular forms visible throughout the cabin. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The old photograph remains in Suriyoung's possession as she points to the child in it. The wheelhouse walls already bear the numerous circular marks that become visible moments later; her backpack, maps, note, and phone remain with her.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 8, "B": 6, "A": 4 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 8, "verdict_ko": "해부학적으로 올바른 손 구조로 사진을 가리키는 동작을 완벽히 구현했으며, 텍스트 배제 및 모든 공간적 지시사항을 정확하게 준수했습니다." }, { "label": "B", "score": 6, "verdict_ko": "공간과 소품의 세부 묘사는 뛰어나나, 사진을 잡은 왼손의 해부학적 오류가 뚜렷하고 모니터 화면에 금지된 텍스트가 생성되었습니다." }, { "label": "A", "score": 4, "verdict_ko": "가리키는 오른손의 손가락 구조가 심하게 왜곡되었으며, 소품 사진과 주변 사물의 디테일이 다른 후보작에 비해 떨어집니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L17B02.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S29sh4.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 금월 식당과 두 어린 여자아이가 찍힌 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 사진을 보고 있음에도 사진의 앞면(이미지가 있는 면)이 인물의 눈이 아닌 카메라 쪽을 완전히 향하고 있습니다.", "fix_en": "Rotate the photograph so its functional front side faces the character's eyes, showing its back or angled edge to the camera." }, { "issue_ko": "사진을 잡고 가리키는 오른손의 손가락들이 기형적으로 뭉개져 해부학적으로 어긋나 있습니다.", "fix_en": "Correct the anatomy of the right hand to show distinct, realistic fingers properly holding the photograph and pointing." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the photograph so its functional front side faces the character's eyes, showing its back or angled edge to the camera.\n- Correct the anatomy of the right hand to show distinct, realistic fingers properly holding the photograph and pointing.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S29sh5": { "input_fingerprint": "4b9fc3c03ca9194f", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 두 눈동자가 크게 확장된 채 멍하게 굳어버린 창백한 인우의 얼굴 클로즈업.\n\nLOCATION (lock): The enclosed wheelhouse of a small fishing boat, containing steering controls and numerous circular forms visible throughout the cabin. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): The wheelhouse remains covered with numerous circular marks, including the red circle noticed by Suriyoung. The old photograph and Suriyoung's other belongings remain with her aboard the boat.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 인우 (한국인 남성, 20대 중반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 5, "A": 4 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "창문에 지도가 임의로 추가된 점은 아쉬우나, 확장된 동공, 창백한 안색, 수많은 원형 자국 등 지문과 레이아웃의 핵심 묘사를 가장 잘 구현했습니다." }, { "label": "B", "score": 5, "verdict_ko": "장소 구현은 깔끔하나, 텍스트가 요구한 크게 확장된 동공과 창백한 피부 표현이 부족하며 '수많은 원형 자국'이 누락되었습니다." }, { "label": "A", "score": 4, "verdict_ko": "인물의 동공과 안색 묘사가 부족할 뿐만 아니라, 원형 자국들이 공간에 녹아들지 못하고 2D 그래픽처럼 떠 있어 실사 조건에 크게 위배됩니다." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L17B01.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S29sh5.png" }, { "label": "CHARACTER REFERENCE — 인우: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스케치의 빨간 원들이 실제 공간의 흔적이 아닌 디지털 주석(그래픽)처럼 화면 곳곳에 떠 있습니다.", "fix_en": "Remove the floating red graphic circles and render the circular marks as realistic physical traces physically painted or scratched onto the interior surfaces." }, { "issue_ko": "우측 창문에 위치 참고 사진에 없는 지도가 추가되었으며, 금지된 텍스트가 포함되어 있습니다.", "fix_en": "Remove the map and its text from the right window, leaving the glass bare as shown in the location photograph." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the floating red graphic circles and render the circular marks as realistic physical traces physically painted or scratched onto the interior surfaces.\n- Remove the map and its text from the right window, leaving the glass bare as shown in the location photograph.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S29sh8": { "input_fingerprint": "65a2057a657baa37", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 눈이 반쯤 감긴 수리영의 고개가 꼿꼿이 선 인우의 어깨 위에 힘없이 얹혀 있는 찰나의 정지 상태.\n\nLOCATION (lock): The enclosed wheelhouse of a small fishing boat, containing steering controls and numerous circular forms visible throughout the cabin. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): Su-ri-young’s limp body slumps toward the upright In-woo, with her half-closed-eyed head resting heavily on his shoulder. Her limbs are unbraced as his shoulder supports her collapsing upper body.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-young is now unconscious. Su-ri-young’s limp body slumps toward the upright In-woo, with her half-closed-eyed head resting heavily on his shoulder. Her limbs are unbraced as his shoulder supports her collapsing upper body; the old photograph, her backpack and maps, the coordinate note, and her smartphone remain with her aboard the boat. The wheelhouse's numerous circular marks remain unchanged.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 여성, 20대 초반 얼굴, 젊은 동아시아계 이목구비, 자연스러운 검은 머리); 인우 (한국인 남성, 20대 중반 얼굴, 젊은 동아시아계 이목구비, 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\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "무의식 상태의 늘어진 사지와 중력 지침을 완벽히 구현했으며, 요구된 모든 소지품이 바닥에 정확히 배치됨." }, { "label": "A", "score": 5, "verdict_ko": "배경과 인물 묘사는 우수하나, 기절한 수리영이 지도를 쥐고 있어 근육 무사용 및 중력 지침을 위반함." }, { "label": "B", "score": 4, "verdict_ko": "수리영이 종이를 쥐고 있어 중력 지침에 어긋나며, 인우의 얼굴이 전혀 보이지 않아 인물 확인이 어려움." } ], "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/ecedba66-2474-4dff-a1a4-7d13bac08e1d/episodes/5ea07552-480a-4fdd-a888-b6147ff239f5/images/background_chain/L17B01.png" }, { "label": "LAYOUT SKETCH — a bare thin-line layout guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, pose and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic reference. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/conti_S29sh8.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 인우: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "프롬프트에서 모든 텍스트를 금지했으나 바닥에 놓인 메모지에 글자가 렌더링되었습니다.", "fix_en": "Remove all visible text and writing from the note paper on the floor." }, { "issue_ko": "레이아웃 스케치에 따르면 인우는 의자에 앉아 있어야 하지만 서 있는 자세로 렌더링되었습니다.", "fix_en": "Change In-woo's pose so that he is seated on the chair, exactly matching the layout sketch." }, { "issue_ko": "로케이션 레퍼런스 사진에 존재하는 의자의 금속 프레임 등받이가 누락되었습니다.", "fix_en": "Restore the metal backrest to the chair to match the fixed features of the location photograph." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all visible text and writing from the note paper on the floor.\n- Change In-woo's pose so that he is seated on the chair, exactly matching the layout sketch.\n- Restore the metal backrest to the chair to match the fixed features of the location photograph.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "플레이트+콘티+엔티티" }, "S30sh1": { "input_fingerprint": "ccc5da32b5c72733", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — contemporary South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): sunset, golden light and dense fog.\n\nSHOT TEXT (authoritative, Korean): 황금빛 노을이 덮인 바다 위, 뱃머리가 거대한 안개벽 속으로 반쯤 가려진 채 물살을 가르며 나아가는 소형 어선의 실루엣.\n\nLOCATION (lock): Open sea crossed by a small fishing boat, with a dense bank of fog forming a broad visual boundary across the water. The shot takes place here — the attached STORYBOARD SKETCH fixes the staging, camera and figure placement of this exact place. No location photograph is attached — build the location itself strictly from the location text above and the shot text, inventing nothing beyond them.\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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Suriyoung remains unconscious aboard Inwoo's boat, with her photograph, backpack and maps, coordinate note, and smartphone still aboard in the absence of any shown loss or removal.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 5, "B": 6, "C": 4 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "A", "score": 5, "verdict_ko": "거대한 안개벽을 표현했으나 질감이 연기나 먼지처럼 보이며, 스토리보드에 지정된 좌측 태양의 위치를 누락했습니다." }, { "label": "B", "score": 6, "verdict_ko": "좌측의 태양과 우측의 안개벽이라는 스토리보드의 구도를 가장 충실히 따랐으나, 뱃머리가 안개에 반쯤 가려지는 묘사는 다소 약합니다." }, { "label": "C", "score": 4, "verdict_ko": "태양의 위치가 우측으로 배치되어 레퍼런스 스토리보드의 기본 구도를 명백히 위반했습니다." } ], "refs": [ { "label": "STORYBOARD SKETCH — a bare thin-line staging guide, a REFERENCE ONLY: take from it ONLY the camera framing, figure placement, heading and size/depth order. It carries ZERO visual style — every texture, material, light and all realism come from the text and the photographic references. Never let any line-drawing quality leak into the output.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/ecedba66-2474-4dff-a1a4-7d13bac08e1d/images/5ea07552-480a-4fdd-a888-b6147ff239f5/conti/lane_sketch_S30sh1.png" } ], "critique": { "issues": [ { "issue_ko": "조타실 창문 안쪽에 사람의 실루엣이 보입니다 (인물 등장 불가 지시 위반).", "fix_en": "Remove the human silhouette from inside the boat's cabin." }, { "issue_ko": "뱃머리가 안개벽에 반쯤 가려진 상태여야 하나, 배 전체가 안개 밖에 온전히 노출되어 있습니다.", "fix_en": "Envelop the front half of the boat within the dense wall of fog." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the human silhouette from inside the boat's cabin.\n- Envelop the front half of the boat within the dense wall of fog.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "lane(map_marker): 스케치" } }