{ "S1sh1": { "input_fingerprint": "d4be3369459b0ee6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 우거진 수풀을 헤치며 달려가는 중간 순간, 눈을 크게 뜨고 입을 벌린 채 겁에 질린 금월산 피해 남자의 상체 정면 구도.\n\nLOCATION (lock): A dense forest with thick undergrowth obstructing the path, serving as a rough escape route through the vegetation. 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 has a large, bleeding bite mark on his thigh.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 금월산 피해 남자 (한국인 성인 남성, 자연스러운 얼굴, 짙은색 머리, 성인 남성 얼굴) — 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 상체 정면 구도와 겁에 질린 표정을 정확히 구현했으며, 레퍼런스의 의상과 인상을 가장 충실하게 반영했습니다." }, { "label": "B", "score": 6, "verdict_ko": "구도와 캐릭터 디자인은 지시를 잘 따랐으나, 몸 주변 허공에 떠 있는 나뭇가지들이 다소 부자연스럽습니다." }, { "label": "C", "score": 3, "verdict_ko": "다리 상처를 보여주기 위해 프레임을 넓혀 '상체 구도' 지시를 어겼으며, 의상이 레퍼런스와 다르고 명시되지 않은 막대기를 들고 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S1sh1.png" }, { "label": "CHARACTER REFERENCE — 금월산 피해 남자: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터의 왼손(화면 우측에 보이는 손)에 레퍼런스 이미지에 있는 전술 장갑이 없습니다.", "fix_en": "Add the tactical glove to the character's left hand." }, { "issue_ko": "레퍼런스 이미지에 나타난 얼굴의 핏자국과 상처가 생략되었습니다.", "fix_en": "Add the bloody scratches and injuries to the character's face to match the reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the tactical glove to the character's left hand.\n- Add the bloody scratches and injuries to the character's face to match the 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" }, "S1sh4": { "input_fingerprint": "1e6a64a601adde74", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 정체불명의 습격자(검은 인영·붉은 눈의 남자)가 공중으로 뛰어올라 금월산 피해 남자 위로 짙은 그림자를 드리운 찰나.\n\nLOCATION (lock): A dense forest with thick undergrowth obstructing the path, serving as a rough escape route through the vegetation. 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 dense dusk foliage, forest color palette, and the victim's clothing and position from the reference. Exclude the prior frontal close-up expression and introduce only the airborne attacker and its cast shadow required for this moment.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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 man remains one-shoed with the bleeding bite wound on his thigh as the black figure descends on him.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정체불명의 습격자 (활성 상태) (한국인으로 보이는 성인 남성, 어둠에 가려진 얼굴, 식별하기 어려운 머리 형태, 남성 얼굴 윤곽); 금월산 피해 남자 (한국인 성인 남성, 자연스러운 얼굴, 짙은색 머리, 성인 남성 얼굴) — 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": "지침대로 이전 컷의 정면 표정을 배제하고, 피해자 위로 짙은 그림자가 드리워진 습격자의 도약 장면을 성공적으로 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "배제해야 할 이전 컷의 표정을 그대로 사용했으며, 피해자 위로 짙게 드리워지는 그림자 연출이 부족합니다." }, { "label": "C", "score": 3, "verdict_ko": "스토리보드의 스케치 선이 그대로 노출되고 인물이 콜라주처럼 겹쳐지는 치명적인 규칙 위반이 발생했습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S1sh4.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S1sh1_sel.png" }, { "label": "CHARACTER REFERENCE — 정체불명의 습격자 (활성 상태): the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 금월산 피해 남자: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "습격자의 왼쪽 다리 아래, 피해 남자의 등 뒤 허공에 주인이 없는 정체불명의 신발이 하나 더 생성되어 있습니다.", "fix_en": "Remove the extra, disembodied boot floating in the air behind the victim's waist." }, { "issue_ko": "피해 남자는 '한쪽 신발만 신은 상태(one-shoed)'여야 한다는 지시사항과 달리 양발 모두에 등산화를 신고 있습니다.", "fix_en": "Change the victim's left foot to be entirely barefoot, ensuring he is explicitly one-shoed." }, { "issue_ko": "이전 컷 레퍼런스에서 피해 남자는 한쪽 손에만 장갑을 끼고 있으나, 생성된 이미지에서는 양손 모두에 전술 장갑을 끼고 있습니다.", "fix_en": "Remove the tactical glove from the victim's left hand so he is only wearing one glove, matching the previous shot still." }, { "issue_ko": "텍스트에 습격자가 비인간적인 존재라는 명시가 없으므로 해부학적으로 정상적인 사람의 눈이어야 하나, 눈에서 초자연적인 빛이 나고 있습니다.", "fix_en": "Remove the supernatural glow from the attacker's eyes, rendering them as anatomically normal human eyes." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the extra, disembodied boot floating in the air behind the victim's waist.\n- Change the victim's left foot to be entirely barefoot, ensuring he is explicitly one-shoed.\n- Remove the tactical glove from the victim's left hand so he is only wearing one glove, matching the previous shot still.\n- Remove the supernatural glow from the attacker's eyes, rendering them as anatomically normal human eyes.\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" }, "S2sh1": { "input_fingerprint": "f9492bfb3e0a3354", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, 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 어두운 골목, 콘크리트 전봇대에 정체불명의 습격자(검은 인영·붉은 눈의 남자)와 효장동 피해 여자의 몸이 밀착된 전신 구도.\n\nLOCATION (lock): A narrow urban back alley lined with utility poles, including a pole around which two figures are entangled. The alley is sparsely lit by a dim streetlamp. 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: 정체불명의 습격자 (활성 상태) (한국인으로 보이는 성인 남성, 어둠에 가려진 얼굴, 식별하기 어려운 머리 형태, 남성 얼굴 윤곽); 효장동 피해 여자 (한국인 성인 여성, 자연스러운 얼굴, 짙은색 머리, 성인 여성 얼굴) — 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": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "스토리보드의 구도와 골목 배경 구조를 가장 정확히 구현했으며 습격자의 특징(검은 후드, 붉은 눈)을 잘 살렸으나, 콘크리트 전봇대가 아닌 벽면 배관에 밀착된 점이 다소 아쉽습니다." }, { "label": "B", "score": 4, "verdict_ko": "인물들이 전봇대에 밀착한 점은 지문에 부합하나, 전봇대가 화면 중앙에 위치하도록 배경 구조와 구도를 임의로 변경하여 스토리보드 지침을 크게 어겼습니다." }, { "label": "C", "score": 3, "verdict_ko": "스토리보드에 없는 왼쪽 전봇대를 임의로 생성하고 인물들을 그곳에 배치하여 지정된 공간 및 구도 지침을 완전히 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S2sh1.png" }, { "label": "CHARACTER REFERENCE — 정체불명의 습격자 (활성 상태): the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 효장동 피해 여자: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "습격자의 얼굴이 이목구비 없는 검은 형체로 묘사되었고, 눈이 해부학적 지침을 위반하여 비현실적으로 붉게 빛납니다.", "fix_en": "Render a realistic human face for the attacker in the shadows, replacing the glowing red dots with anatomically normal eyes." }, { "issue_ko": "기준 이미지에서 피해 여성이 착용하고 있는 모자가 벗겨져 전봇대 옆 바닥에 떨어져 있습니다.", "fix_en": "Remove the hat from the ground and place it onto the victim's head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Render a realistic human face for the attacker in the shadows, replacing the glowing red dots with anatomically normal eyes.\n- Remove the hat from the ground and place it onto the victim's head.\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": "fa6d612d07d0fb8d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, dim streetlight.\n\nSHOT TEXT (authoritative, Korean): 효장동 피해 여자가 두 눈을 감은 채 아스팔트 바닥에 축 늘어져 쓰러진 전신 구도.\n\nLOCATION (lock): A narrow urban back alley lined with utility poles, including a pole around which two figures are entangled. The alley is sparsely lit by a dim streetlamp. 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 lighting, concrete utility pole, asphalt texture, and the woman's clothing from the reference. Exclude the attacker and the earlier upright struggle; show only the woman collapsed on the alley floor.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is fully collapsed on the asphalt with her eyes closed and her entire body slack. Her head, torso, both arms, and both legs rest limply against the road surface.\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 woman is fully collapsed on the asphalt with her eyes closed and her entire body slack, with flesh torn from her neck. Her head, torso, both arms, and both legs rest limply against the road surface.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 효장동 피해 여자 (한국인 성인 여성, 자연스러운 얼굴, 짙은색 머리, 성인 여성 얼굴) — 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": 6, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 6, "verdict_ko": "지정된 로케이션과 전신 쓰러진 자세를 잘 구현했으나, 이전 샷의 맨다리와 달리 긴바지로 렌더링된 점이 아쉽습니다." }, { "label": "A", "score": 5, "verdict_ko": "배경은 일치하지만 필수 소품인 모자가 누락되었고, 상처를 보여주기 위해 목이 부자연스럽게 꺾인 해부학적 오류가 있습니다." }, { "label": "B", "score": 4, "verdict_ko": "의상은 이전 샷과 가장 유사하지만, 배경에 없는 낙서가 생기고 전봇대 구조가 달라져 로케이션 고정 지침을 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S2sh1_sel.png" }, { "label": "CHARACTER REFERENCE — 효장동 피해 여자: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "이전 샷에서 머리에 쓰고 있던 모자가 벗겨져 머리 아래에 베개처럼 깔려 있습니다.", "fix_en": "Place the hat correctly onto the woman's head as it was worn in the references." }, { "issue_ko": "바닥에 놓인 오른손의 손가락들이 중력을 무시하고 위로 뻣뻣하게 들려 있습니다.", "fix_en": "Make all fingers of the right hand rest completely flat and limp against the asphalt." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Place the hat correctly onto the woman's head as it was worn in the references.\n- Make all fingers of the right hand rest completely flat and limp against the asphalt.\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" }, "S3sh3": { "input_fingerprint": "159450ae5fe0bccd", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 혜수가 한국인 형사들 한가운데 서서 집게손가락으로 특정 방향을 가리키며 입을 벌린 상체 정지 구도.\n\nLOCATION (lock): Two secured outdoor crime scenes: a forest trail and an urban back alley. Police lines, forensic personnel, investigators, and body-recovery activity define both locations. 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: 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": "스토리보드의 인물 배치와 숲 배경을 완벽히 구현했으며, 지정된 동작과 표정이 자연스럽습니다." }, { "label": "B", "score": 6, "verdict_ko": "가리키는 손가락 형태가 다소 기형적이며, 스케치에 없는 도심 골목이 혼합되어 배경이 어색합니다." }, { "label": "A", "score": 4, "verdict_ko": "스토리보드에 없는 인물이 우측에 추가되었고, 인물들의 시선 방향이 스케치와 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S3sh3.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "배경의 폴리스 라인 테이프에 문자가 표시되어 있습니다 (프롬프트의 'No text' 조건 위반).", "fix_en": "Remove all text and letters from the police line tapes in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text and letters from the police line tapes 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": "lane(map_marker): 스케치+엔티티" }, "S4sh1": { "input_fingerprint": "c4e3e1e3fe42cf51", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 흐린 톤의 낮, 인천 변두리의 낡은 다세대 빌라 옥상 위 평상과 작은 옥탑방 지붕이 내려다보이는 부감 전경.\n\nLOCATION (lock): The rooftop level of an aging low-rise residential villa on the outskirts of a city, with a small rooftop dwelling visible in a wide elevated view. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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": 8, "B": 5, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 8, "verdict_ko": "레퍼런스 구도 복사 금지 지침을 지키며 옥탑방 지붕과 평상을 내려다보는 부감 전경을 새롭게 구현함." }, { "label": "B", "score": 5, "verdict_ko": "지정된 옥탑방과 주변 환경의 결합은 우수하나, 위치 레퍼런스의 카메라 구도를 완벽히 복사하여 부정 지침을 위반함." }, { "label": "C", "score": 3, "verdict_ko": "구조물 레퍼런스의 구도를 그대로 복사했으며, 텍스트에 명시된 필수 피사체인 '평상'이 누락됨." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L03B03.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.png" } ], "critique": { "issues": [ { "issue_ko": "나무 평상이 계단실의 콘크리트 벽을 뚫고 물리적으로 겹쳐 있으며, 오른쪽 다리 부분이 허공에 떠 있습니다.", "fix_en": "Reposition the wooden platform so it rests completely flat on the green rooftop surface, entirely separated from the stairwell wall." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Reposition the wooden platform so it rests completely flat on the green rooftop surface, entirely separated from the stairwell wall.\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+seed만 (배경 전용)", "lane_policy": "ab_select_bypass:bg_only", "plate_select": { "candidates": { "A": "L03B02", "B": "L03B03" }, "assigned": "L03B03", "choice": "Candidate B", "confident": true, "reason_ko": "지문에서 요구하는 옥상 위 평상과 작은 옥탑방이 내려다보이는 부감 전경을 동일하게 보여주므로, 현재 할당된 후보를 유지합니다.", "kept": "L03B03" } }, "S5sh3::ab_conti": { "input_fingerprint": "5ae630adf3ace8bd", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 앞치마를 두른 강민숙이 굳고 어두운 표정으로 거실 TV 화면을 뚫어지게 응시하는 상체.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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): Min-sook already bears the small circular mark on the inside of her wrist; it has been present since Su-ri-yeong's childhood.\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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "지정된 실내 구조와 상체 앵글을 가장 잘 재현했으며, 손목 안쪽의 원형 표식과 굳은 표정까지 프롬프트의 지시를 가장 충실히 따랐습니다." }, { "label": "B", "score": 3, "verdict_ko": "배경에 두 번째 TV가 등장하는 치명적인 오류가 발생했으며, 손목의 표식이 안쪽이 아닌 바깥쪽에 위치해 감점되었습니다." }, { "label": "C", "score": 5, "verdict_ko": "인물의 굳은 표정은 잘 묘사되었으나 주방과 복도의 구조가 레퍼런스와 다르게 왜곡되었으며, 손목 표식의 위치가 바깥쪽으로 잘못 적용되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S5sh3.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "배경 우측에 침실과 침대가 노출되어 있으나, 공간 레퍼런스 사진에 따르면 해당 위치는 막힌 벽이어야 합니다.", "fix_en": "Replace the open doorway and bedroom on the right side of the background with a solid wall and wallpaper matching the location reference." }, { "issue_ko": "손목의 작은 원형 자국이 지시문(손목 안쪽)과 달리 팔목 바깥쪽/측면에 위치해 있습니다.", "fix_en": "Move the small circular mark from the outer side of the arm to the inside (palmar side) of the wrist." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Replace the open doorway and bedroom on the right side of the background with a solid wall and wallpaper matching the location reference.\n- Move the small circular mark from the outer side of the arm to the inside (palmar side) of the wrist.\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." }, "S5sh3::ab_noconti": { "input_fingerprint": "f9a3d376827b8d39", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 앞치마를 두른 강민숙이 굳고 어두운 표정으로 거실 TV 화면을 뚫어지게 응시하는 상체.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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): Min-sook already bears the small circular mark on the inside of her wrist; it has been present since Su-ri-yeong's childhood.\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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "지정된 샷 사이즈와 완벽히 일치하는 공간 구조를 구현했으나, 의상이 레퍼런스와 다르고 흉터가 손목 바깥쪽에 위치해 일부 감점되었습니다." }, { "label": "A", "score": 5, "verdict_ko": "캐릭터 의상과 흉터의 위치는 레퍼런스에 충실하지만, 공간 전체가 좌우 반전되어 장소 우선순위에서 감점되었습니다." }, { "label": "C", "score": 3, "verdict_ko": "TV 뒷면에 화면이 표시되는 물리적으로 불가능한 형태와 텍스트 출력 금지 위반 등 치명적인 오류로 인해 최하위입니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터 레퍼런스에서는 머리를 풀고 있으나, 생성된 이미지에서는 머리를 묶고 있어 'hair exactly' 지시를 위반했습니다.", "fix_en": "Change the character's hairstyle so that her hair is down, matching the reference image." }, { "issue_ko": "지시문에는 손목 안쪽(inside of her wrist)에 둥근 자국이 있어야 한다고 명시되어 있으나, 팔 바깥쪽에 자국이 있습니다.", "fix_en": "Move the small circular mark from the outside of her arm to the inside of her wrist." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the character's hairstyle so that her hair is down, matching the reference image.\n- Move the small circular mark from the outside of her arm to the inside of her wrist.\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." }, "S5sh3::conti_ab_decision": { "fingerprint": "e76fba27b12c1297", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "지문에서 요구한 'TV 화면'이 프레임 안에 포함되어 시선의 대상이 명확하며, 손목 안쪽의 원형 표식도 정확하게 묘사되었습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 지문에서 명시한 TV 화면을 바라보는 구도와 손목 안쪽의 둥근 표식을 모두 명확하게 구현했습니다." } ] } }, "S5sh3": { "input_fingerprint": "5ae630adf3ace8bd", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 앞치마를 두른 강민숙이 굳고 어두운 표정으로 거실 TV 화면을 뚫어지게 응시하는 상체.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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): Min-sook already bears the small circular mark on the inside of her wrist; it has been present since Su-ri-yeong's childhood.\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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "지정된 실내 구조와 상체 앵글을 가장 잘 재현했으며, 손목 안쪽의 원형 표식과 굳은 표정까지 프롬프트의 지시를 가장 충실히 따랐습니다." }, { "label": "B", "score": 3, "verdict_ko": "배경에 두 번째 TV가 등장하는 치명적인 오류가 발생했으며, 손목의 표식이 안쪽이 아닌 바깥쪽에 위치해 감점되었습니다." }, { "label": "C", "score": 5, "verdict_ko": "인물의 굳은 표정은 잘 묘사되었으나 주방과 복도의 구조가 레퍼런스와 다르게 왜곡되었으며, 손목 표식의 위치가 바깥쪽으로 잘못 적용되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S5sh3.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "배경 우측에 침실과 침대가 노출되어 있으나, 공간 레퍼런스 사진에 따르면 해당 위치는 막힌 벽이어야 합니다.", "fix_en": "Replace the open doorway and bedroom on the right side of the background with a solid wall and wallpaper matching the location reference." }, { "issue_ko": "손목의 작은 원형 자국이 지시문(손목 안쪽)과 달리 팔목 바깥쪽/측면에 위치해 있습니다.", "fix_en": "Move the small circular mark from the outer side of the arm to the inside (palmar side) of the wrist." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Replace the open doorway and bedroom on the right side of the background with a solid wall and wallpaper matching the location reference.\n- Move the small circular mark from the outer side of the arm to the inside (palmar side) of the wrist.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "지문에서 요구한 'TV 화면'이 프레임 안에 포함되어 시선의 대상이 명확하며, 손목 안쪽의 원형 표식도 정확하게 묘사되었습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 지문에서 명시한 TV 화면을 바라보는 구도와 손목 안쪽의 둥근 표식을 모두 명확하게 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S5sh3__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S5sh3__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B01", "choice": "Candidate A", "confident": true, "reason_ko": "지문에서 '거실 TV 화면'을 명시하고 있으므로, TV가 배치된 거실 공간을 보여주는 현재 할당된 후보가 가장 적합합니다.", "kept": "L04B01" } }, "S5sh10::ab_conti": { "input_fingerprint": "832974dd878ba689", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 수리영이 강민숙의 어깨에 코를 묻고 눈을 지그시 감은 채 미소 짓는 측면.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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 Min-sook's inner wrist while Su-ri-yeong embraces 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 강민숙 (한국인 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": 4 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "수리영의 표정과 행동 지문은 정확히 구현했으나, 원형 표식이 강민숙이 아닌 수리영의 팔에 있음." }, { "label": "B", "score": 3, "verdict_ko": "수리영의 등 뒤에 소속 불명의 손이 추가로 등장하는 치명적인 신체 구조 오류가 있음." }, { "label": "C", "score": 4, "verdict_ko": "김이 나는 냄비가 가스레인지가 아닌 식탁 위에 배치되어 주요 공간 및 소품 조건을 위반함." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S5sh10.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "작은 원형 자국이 강민숙(베이지색 옷)의 안쪽 손목이 아닌 수리영(초록색 후드)의 팔목 바깥쪽에 그려져 있습니다.", "fix_en": "Remove the circular mark from the wrist of the green-sleeved arm and place it on Min-sook's inner wrist." }, { "issue_ko": "초록색 옷을 입은 수리영의 오른쪽 어깨에 얹혀 있는 오른손의 엄지손가락이 위를 향하고 있어, 해부학적으로 왼손이 달린 것처럼 잘못 표현되었습니다.", "fix_en": "Redraw the foreground hand resting on the green shoulder so the thumb points downward, correctly matching the anatomy of a right hand." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the circular mark from the wrist of the green-sleeved arm and place it on Min-sook's inner wrist.\n- Redraw the foreground hand resting on the green shoulder so the thumb points downward, correctly matching the anatomy of a right hand.\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." }, "S5sh10::ab_noconti": { "input_fingerprint": "aeab77ce4d9a7b2d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 수리영이 강민숙의 어깨에 코를 묻고 눈을 지그시 감은 채 미소 짓는 측면.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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 Min-sook's inner wrist while Su-ri-yeong embraces 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 강민숙 (한국인 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "요청된 수리영의 표정과 측면 구도를 가장 밀도 있게 구현했으나, 원형 자국이 강민숙이 아닌 수리영의 팔에 나타났습니다." }, { "label": "B", "score": 5, "verdict_ko": "행동은 잘 묘사되었으나 구도가 넓어 집중도가 떨어지며, 강민숙이 시계를 차고 있고 자국 위치도 잘못되었습니다." }, { "label": "C", "score": 3, "verdict_ko": "두 인물의 의상이 기준 이미지와 완전히 불일치하며, 원형 자국이 두 사람 모두에게 잘못 생성되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영의 눈이 지그시 감겨 있지 않고 아래를 향해 떠 있습니다.", "fix_en": "Close Su-ri-yeong's eyes completely." }, { "issue_ko": "수리영의 코가 강민숙의 어깨에 묻히지 않고 완전히 드러나 있습니다.", "fix_en": "Adjust Su-ri-yeong's head angle so her nose is buried into Min-sook's shoulder." }, { "issue_ko": "화면 앞쪽의 팔(초록색 소매)이 비정상적으로 꺾여 자기 자신(초록색 옷)의 어깨를 껴안고 있는 해부학적 오류가 있습니다.", "fix_en": "Redraw the foreground arm so the hand wraps around and rests on Min-sook's beige clothing." }, { "issue_ko": "원형 표식이 강민숙의 안쪽 손목이 아닌, 초록색 옷을 입은 팔의 바깥쪽 팔뚝에 위치해 있습니다.", "fix_en": "Remove the circular mark from the foreground forearm and place it on Min-sook's inner wrist." }, { "issue_ko": "가스레인지 위 냄비에서 김이 나야 하는데, 싱크대 위 허공에서 수증기가 피어오르고 있습니다.", "fix_en": "Remove the steam above the sink and generate steam rising directly from the pot on the stove." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Close Su-ri-yeong's eyes completely.\n- Adjust Su-ri-yeong's head angle so her nose is buried into Min-sook's shoulder.\n- Redraw the foreground arm so the hand wraps around and rests on Min-sook's beige clothing.\n- Remove the circular mark from the foreground forearm and place it on Min-sook's inner wrist.\n- Remove the steam above the sink and generate steam rising directly from the pot on the stove.\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." }, "S5sh10::conti_ab_decision": { "fingerprint": "81a2b612aa456a27", "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 강민숙(녹색 옷을 입은 중년 여성)의 안쪽 손목에 원형 자국을 정확히 배치한 반면, 후보 1은 수리영의 손목에 자국을 그려 프롬프트를 위반했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "후보 2는 손가락이 6개이고 팔의 연결 구조가 물리적으로 불가능한 치명적 인체 오류(규정 1 위반)가 있어 후보 1이 승리합니다." } ] } }, "S5sh10": { "input_fingerprint": "aeab77ce4d9a7b2d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 수리영이 강민숙의 어깨에 코를 묻고 눈을 지그시 감은 채 미소 짓는 측면.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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 Min-sook's inner wrist while Su-ri-yeong embraces 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 강민숙 (한국인 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "요청된 수리영의 표정과 측면 구도를 가장 밀도 있게 구현했으나, 원형 자국이 강민숙이 아닌 수리영의 팔에 나타났습니다." }, { "label": "B", "score": 5, "verdict_ko": "행동은 잘 묘사되었으나 구도가 넓어 집중도가 떨어지며, 강민숙이 시계를 차고 있고 자국 위치도 잘못되었습니다." }, { "label": "C", "score": 3, "verdict_ko": "두 인물의 의상이 기준 이미지와 완전히 불일치하며, 원형 자국이 두 사람 모두에게 잘못 생성되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영의 눈이 지그시 감겨 있지 않고 아래를 향해 떠 있습니다.", "fix_en": "Close Su-ri-yeong's eyes completely." }, { "issue_ko": "수리영의 코가 강민숙의 어깨에 묻히지 않고 완전히 드러나 있습니다.", "fix_en": "Adjust Su-ri-yeong's head angle so her nose is buried into Min-sook's shoulder." }, { "issue_ko": "화면 앞쪽의 팔(초록색 소매)이 비정상적으로 꺾여 자기 자신(초록색 옷)의 어깨를 껴안고 있는 해부학적 오류가 있습니다.", "fix_en": "Redraw the foreground arm so the hand wraps around and rests on Min-sook's beige clothing." }, { "issue_ko": "원형 표식이 강민숙의 안쪽 손목이 아닌, 초록색 옷을 입은 팔의 바깥쪽 팔뚝에 위치해 있습니다.", "fix_en": "Remove the circular mark from the foreground forearm and place it on Min-sook's inner wrist." }, { "issue_ko": "가스레인지 위 냄비에서 김이 나야 하는데, 싱크대 위 허공에서 수증기가 피어오르고 있습니다.", "fix_en": "Remove the steam above the sink and generate steam rising directly from the pot on the stove." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Close Su-ri-yeong's eyes completely.\n- Adjust Su-ri-yeong's head angle so her nose is buried into Min-sook's shoulder.\n- Redraw the foreground arm so the hand wraps around and rests on Min-sook's beige clothing.\n- Remove the circular mark from the foreground forearm and place it on Min-sook's inner wrist.\n- Remove the steam above the sink and generate steam rising directly from the pot on the stove.\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.", "conti_ab": { "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 강민숙(녹색 옷을 입은 중년 여성)의 안쪽 손목에 원형 자국을 정확히 배치한 반면, 후보 1은 수리영의 손목에 자국을 그려 프롬프트를 위반했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "후보 2는 손가락이 6개이고 팔의 연결 구조가 물리적으로 불가능한 치명적 인체 오류(규정 1 위반)가 있어 후보 1이 승리합니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S5sh10__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S5sh10__ab_noconti_sel.png" }, "ref_mode": "플레이트+엔티티 (A/B: 무콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B01", "choice": "Candidate A", "confident": true, "reason_ko": "지문이 인물 간의 동작(어깨에 코를 묻고 미소 짓는 모습)만 묘사하고 특정 공간이나 가구를 명시하지 않으므로, 현재 배정된 공간을 유지합니다.", "kept": "L04B01" } }, "S5sh12::ab_conti": { "input_fingerprint": "5366865e63899100", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 강민숙의 손목 안쪽 피부에 선명하게 새겨진 작은 원 형태의 뚜렷한 표식 클로즈업.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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): Min-sook's inner wrist bears the same unusually distinct small 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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "요구된 손목 표식 클로즈업 구도를 완벽히 따랐으며, 레퍼런스의 긴소매 의상과 배경의 가스레인지 위 냄비까지 정확하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "올바른 클로즈업 구도를 보여주나, 레퍼런스에 있는 인물의 긴소매 의상이 누락되었고 배경에 끓고 있는 냄비가 없습니다." }, { "label": "C", "score": 3, "verdict_ko": "표식의 클로즈업이라는 명확한 샷 크기 지시를 어기고 인물의 상반신 전체를 넓게 잡아 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S5sh12.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "가스레인지 위에 놓인 냄비에서 프롬프트가 요구한 김(steam)이 피어오르지 않습니다.", "fix_en": "Add visible steam rising from the pot on the stove in the background." }, { "issue_ko": "화면 우측 상단에 위치한 주먹 쥔 손의 손가락 마디와 손톱 위치가 해부학적으로 심하게 왜곡되어 있습니다.", "fix_en": "Render the fingers of the hand at the top right with anatomically correct joints, nails, and separation." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add visible steam rising from the pot on the stove in the background.\n- Render the fingers of the hand at the top right with anatomically correct joints, nails, and separation.\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." }, "S5sh12::ab_noconti": { "input_fingerprint": "c888d44d49a1d150", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 강민숙의 손목 안쪽 피부에 선명하게 새겨진 작은 원 형태의 뚜렷한 표식 클로즈업.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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): Min-sook's inner wrist bears the same unusually distinct small 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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "피부에 새겨진 뚜렷한 원형 표식을 손목 안쪽 클로즈업으로 정확히 묘사했으며 해부학적 오류가 없습니다." }, { "label": "B", "score": 5, "verdict_ko": "원형 표식이 피부에 새겨진 것이 아니라 얹혀 있는 입체적인 반지(물체)로 묘사되어 형태가 어긋납니다." }, { "label": "A", "score": 3, "verdict_ko": "손가락 개수와 관절 구조가 비정상적으로 렌더링된 해부학적 오류(치명적 위반)가 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "손목의 표식이 온전한 원 형태가 아니라 중간이 끊어진 모양입니다.", "fix_en": "Complete the red mark on the wrist so it forms a full, unbroken circle." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Complete the red mark on the wrist so it forms a full, unbroken circle.\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." }, "S5sh12::conti_ab_decision": { "fingerprint": "7f4d60fac2079a82", "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "요구된 배경 요소(김이 나는 냄비가 올려진 가스레인지, 싱크대 등)를 모두 정확히 포함하여 장소 일관성을 완벽히 구현했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "후보 2는 손을 쥐고 있는 손가락의 관절과 개수가 기형적으로 렌더링된 해부학적 오류가 있어, 신체가 자연스럽게 표현된 후보 1이 적합합니다." } ] } }, "S5sh12": { "input_fingerprint": "5366865e63899100", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, sunlight through a small window.\n\nSHOT TEXT (authoritative, Korean): 강민숙의 손목 안쪽 피부에 선명하게 새겨진 작은 원 형태의 뚜렷한 표식 클로즈업.\n\nLOCATION (lock): A compact rooftop apartment interior with a living room, a small sunlit window, a dining table, a sink, and a stove holding a steaming pot. The modest home also connects to separate bedrooms and an entry area. 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): Min-sook's inner wrist bears the same unusually distinct small 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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "요구된 손목 표식 클로즈업 구도를 완벽히 따랐으며, 레퍼런스의 긴소매 의상과 배경의 가스레인지 위 냄비까지 정확하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "올바른 클로즈업 구도를 보여주나, 레퍼런스에 있는 인물의 긴소매 의상이 누락되었고 배경에 끓고 있는 냄비가 없습니다." }, { "label": "C", "score": 3, "verdict_ko": "표식의 클로즈업이라는 명확한 샷 크기 지시를 어기고 인물의 상반신 전체를 넓게 잡아 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S5sh12.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "가스레인지 위에 놓인 냄비에서 프롬프트가 요구한 김(steam)이 피어오르지 않습니다.", "fix_en": "Add visible steam rising from the pot on the stove in the background." }, { "issue_ko": "화면 우측 상단에 위치한 주먹 쥔 손의 손가락 마디와 손톱 위치가 해부학적으로 심하게 왜곡되어 있습니다.", "fix_en": "Render the fingers of the hand at the top right with anatomically correct joints, nails, and separation." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add visible steam rising from the pot on the stove in the background.\n- Render the fingers of the hand at the top right with anatomically correct joints, nails, and separation.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "요구된 배경 요소(김이 나는 냄비가 올려진 가스레인지, 싱크대 등)를 모두 정확히 포함하여 장소 일관성을 완벽히 구현했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "후보 2는 손을 쥐고 있는 손가락의 관절과 개수가 기형적으로 렌더링된 해부학적 오류가 있어, 신체가 자연스럽게 표현된 후보 1이 적합합니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S5sh12__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S5sh12__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B01", "choice": "Candidate A", "confident": true, "reason_ko": "손목의 표식을 보여주는 클로즈업 샷으로 특정 배경이 명시되지 않았으므로, 기존에 할당된 후보를 유지합니다.", "kept": "L04B01" } }, "S6sh1": { "input_fingerprint": "0e9666238e8a5d70", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 밝은 해안가 도로 위, 수리영이 양쪽 귀에 하얀 이어폰을 꽂은 채 자전거 핸들을 쥐고 정면을 향해 미소 짓는 상체 구도, 주행 바람에 머리카락이 뒤로 날리는 찰나.\n\nLOCATION (lock): A coastal road running beside the sea, broad enough for a bicycle route and open to the shoreline. 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: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 6, "C": 5 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "상체 구도, 묶어서 흩날리는 머리카락, 이어폰 등 디테일과 캐릭터 레퍼런스를 가장 사실적이고 완벽하게 구현했습니다." }, { "label": "B", "score": 6, "verdict_ko": "지시된 주요 요소들은 잘 반영되었으나, 이어폰 줄의 연결이 부자연스럽고 배경 디테일이 스토리보드에 비해 다소 단조롭습니다." }, { "label": "C", "score": 5, "verdict_ko": "구도와 인물은 준수하지만, 자전거 핸들 스템의 구조가 물리적으로 불가능하며 도로 바닥에 왜곡된 문자가 포함되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/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": "도로 바닥에 흰색 텍스트(글자)가 그려져 있어 모든 텍스트 배제 지시를 위반했습니다.", "fix_en": "Remove the painted white text and characters from the surface of the road behind the character." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the painted white text and characters from the surface of the road behind the character.\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" }, "S7sh2::ab_conti": { "input_fingerprint": "d9b8f6cc839666e4", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 턱을 당긴 채 매서운 눈빛으로 로트와일러의 두 눈을 똑바로 응시하는 얼굴 클로즈업.\n\nLOCATION (lock): An animal shelter with secured dog kennels and practical veterinary and grooming work areas. The space accommodates injections, clipping, ointment treatment, and bathing. The shot takes place here — the attached LOCATION STRUCTURE PHOTOGRAPH is the single authority for this exact place — its fixed structure and permanent site details are LOCKED to it. No separate location photograph exists for this place. Build everything else strictly from the location text above and the shot text; the layout sketch (when attached) governs framing and placement only, and the shot text governs time of day, lighting and action.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "개의 두 눈을 똑바로 응시하는 시선 처리가 가장 정확하며, 레퍼런스 이미지의 복잡한 구조물과 배경 디테일을 충실히 재현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "시선이 개의 눈이 아닌 주둥이를 향하고 있으며, 지정된 장소의 주변 배경 구조가 전혀 일치하지 않습니다." }, { "label": "C", "score": 3, "verdict_ko": "시선 교환이 빗나갔으며, 개의 목줄이 털과 융합되는 물리적 오류가 발생했습니다." } ], "refs": [ { "label": "LOCATION STRUCTURE PHOTOGRAPH — the confirmed photograph of this exact place and its fixed structure: it is the SINGLE authority for the location, the structure's shape, proportions, materials, colors, openings and every permanent site detail. Never copy its camera framing, time of day or lighting — the shot text is the authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_animal_shelter_yard_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S7sh2.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영의 시선이 로트와일러의 두 눈이 아닌 코/입 부근을 향하고 있습니다.", "fix_en": "Shift her eye line up so she is looking directly into the dog's eyes." }, { "issue_ko": "수리영이 턱을 당기지 않고 오히려 앞으로 살짝 내밀고 있습니다.", "fix_en": "Lower and tuck her chin inward toward her neck." }, { "issue_ko": "지시된 '매서운 눈빛'과 달리 무표정하고 평온한 얼굴입니다.", "fix_en": "Furrow her eyebrows and tighten her eye muscles to create a fierce, glaring expression." }, { "issue_ko": "화면 중앙 탁자 위의 흰색 병과 배경 건물 간판에 텍스트가 포함되어 있습니다.", "fix_en": "Remove all text from the bottle on the table and the signs in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Shift her eye line up so she is looking directly into the dog's eyes.\n- Lower and tuck her chin inward toward her neck.\n- Furrow her eyebrows and tighten her eye muscles to create a fierce, glaring expression.\n- Remove all text from the bottle on the table and the signs 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." }, "S7sh2::ab_noconti": { "input_fingerprint": "003ce1ec660446c0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 턱을 당긴 채 매서운 눈빛으로 로트와일러의 두 눈을 똑바로 응시하는 얼굴 클로즈업.\n\nLOCATION (lock): An animal shelter with secured dog kennels and practical veterinary and grooming work areas. The space accommodates injections, clipping, ointment treatment, and bathing. The shot takes place here — the attached LOCATION STRUCTURE PHOTOGRAPH is the single authority for this exact place — its fixed structure and permanent site details are LOCKED to it. No separate location photograph exists for this place. Build everything else strictly from the location text above and the shot text; the layout sketch (when attached) governs framing and placement only, and the shot text governs time of day, lighting and action.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 3, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지정된 야외 철창 배경과 얼굴 클로즈업 앵글을 가장 잘 구현했으나, 로트와일러가 두 마리 등장하고 시선이 렌즈를 향하는 점이 아쉽습니다." }, { "label": "B", "score": 3, "verdict_ko": "얼굴 클로즈업은 맞으나 로트와일러가 아예 등장하지 않으며, 지정된 배경 대신 임의의 콘크리트 구조물을 생성해 위치 지침을 위반했습니다." }, { "label": "C", "score": 5, "verdict_ko": "개와 시선을 맞추는 액션은 완벽하나, 얼굴 클로즈업보다 넓은 프레이밍을 사용했고 지정된 야외 배경을 실내로 무단 변경했습니다." } ], "refs": [ { "label": "LOCATION STRUCTURE PHOTOGRAPH — the confirmed photograph of this exact place and its fixed structure: it is the SINGLE authority for the location, the structure's shape, proportions, materials, colors, openings and every permanent site detail. Never copy its camera framing, time of day or lighting — the shot text is the authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_animal_shelter_yard_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영이 로트와일러의 두 눈이 아닌 카메라 렌즈를 정면으로 응시하고 있습니다.", "fix_en": "Redirect Su-ri-young's eyes to look directly into the eyes of the Rottweiler." }, { "issue_ko": "로트와일러 한 마리를 응시해야 하지만 전경에 두 마리가 등장했습니다.", "fix_en": "Remove one of the Rottweilers from the foreground so there is only one dog." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redirect Su-ri-young's eyes to look directly into the eyes of the Rottweiler.\n- Remove one of the Rottweilers from the foreground so there is only one dog.\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." }, "S7sh2::conti_ab_decision": { "fingerprint": "3af3d0ff514a2809", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 2는 배경에 프롬프트에 명시되지 않은 인물들을 포함하여 규칙을 위반한 반면, 후보 1은 단독 인물 제약과 개와의 강렬한 시선 교환을 정확하게 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 로트와일러의 두 눈을 똑바로 응시한다는 지문을 정확히 구현한 반면, 후보 1은 시선이 엉뚱한 곳을 향하고 있습니다." } ] } }, "S7sh2": { "input_fingerprint": "d9b8f6cc839666e4", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 턱을 당긴 채 매서운 눈빛으로 로트와일러의 두 눈을 똑바로 응시하는 얼굴 클로즈업.\n\nLOCATION (lock): An animal shelter with secured dog kennels and practical veterinary and grooming work areas. The space accommodates injections, clipping, ointment treatment, and bathing. The shot takes place here — the attached LOCATION STRUCTURE PHOTOGRAPH is the single authority for this exact place — its fixed structure and permanent site details are LOCKED to it. No separate location photograph exists for this place. Build everything else strictly from the location text above and the shot text; the layout sketch (when attached) governs framing and placement only, and the shot text governs time of day, lighting and action.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "개의 두 눈을 똑바로 응시하는 시선 처리가 가장 정확하며, 레퍼런스 이미지의 복잡한 구조물과 배경 디테일을 충실히 재현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "시선이 개의 눈이 아닌 주둥이를 향하고 있으며, 지정된 장소의 주변 배경 구조가 전혀 일치하지 않습니다." }, { "label": "C", "score": 3, "verdict_ko": "시선 교환이 빗나갔으며, 개의 목줄이 털과 융합되는 물리적 오류가 발생했습니다." } ], "refs": [ { "label": "LOCATION STRUCTURE PHOTOGRAPH — the confirmed photograph of this exact place and its fixed structure: it is the SINGLE authority for the location, the structure's shape, proportions, materials, colors, openings and every permanent site detail. Never copy its camera framing, time of day or lighting — the shot text is the authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_animal_shelter_yard_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S7sh2.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영의 시선이 로트와일러의 두 눈이 아닌 코/입 부근을 향하고 있습니다.", "fix_en": "Shift her eye line up so she is looking directly into the dog's eyes." }, { "issue_ko": "수리영이 턱을 당기지 않고 오히려 앞으로 살짝 내밀고 있습니다.", "fix_en": "Lower and tuck her chin inward toward her neck." }, { "issue_ko": "지시된 '매서운 눈빛'과 달리 무표정하고 평온한 얼굴입니다.", "fix_en": "Furrow her eyebrows and tighten her eye muscles to create a fierce, glaring expression." }, { "issue_ko": "화면 중앙 탁자 위의 흰색 병과 배경 건물 간판에 텍스트가 포함되어 있습니다.", "fix_en": "Remove all text from the bottle on the table and the signs in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Shift her eye line up so she is looking directly into the dog's eyes.\n- Lower and tuck her chin inward toward her neck.\n- Furrow her eyebrows and tighten her eye muscles to create a fierce, glaring expression.\n- Remove all text from the bottle on the table and the signs 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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 2는 배경에 프롬프트에 명시되지 않은 인물들을 포함하여 규칙을 위반한 반면, 후보 1은 단독 인물 제약과 개와의 강렬한 시선 교환을 정확하게 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 로트와일러의 두 눈을 똑바로 응시한다는 지문을 정확히 구현한 반면, 후보 1은 시선이 엉뚱한 곳을 향하고 있습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S7sh2__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S7sh2__ab_noconti_sel.png" }, "ref_mode": "seed-bg+콘티+엔티티 (복잡구조물 A/B: 콘티 승)", "lane_policy": "ab_select_ready" }, "S7sh6::ab_conti": { "input_fingerprint": "7c54dc71fbf82263", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 오른쪽 귀에 스마트폰을 바짝 밀착한 채 미간을 찌푸리고 입을 굳게 다문 측면 얼굴 클로즈업.\n\nLOCATION (lock): An animal shelter with secured dog kennels and practical veterinary and grooming work areas. The space accommodates injections, clipping, ointment treatment, and bathing. The shot takes place here — the attached LOCATION STRUCTURE PHOTOGRAPH is the single authority for this exact place — its fixed structure and permanent site details are LOCKED to it. No separate location photograph exists for this place. Build everything else strictly from the location text above and the shot text; the layout sketch (when attached) governs framing and placement only, and the shot text governs time of day, lighting and action.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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-yeong has her smartphone with her while attempting to contact Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 3, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "프롬프트와 스케치에 맞게 오른쪽 귀에 스마트폰을 밀착한 우측면 클로즈업을 정확히 구현했으며, 렌즈가 밖을 향해 소품 방향도 올바릅니다." }, { "label": "B", "score": 3, "verdict_ko": "스마트폰 화면이 바깥을 향하고 있는 치명적인 소품 방향 오류가 있으며, 지시된 오른쪽 귀와 우측면 구도도 어겼습니다." }, { "label": "C", "score": 5, "verdict_ko": "클로즈업과 소품의 파지 방향은 올바르나, 텍스트가 명시한 오른쪽 귀가 아닌 왼쪽 귀에 전화를 대고 있어 감점되었습니다." } ], "refs": [ { "label": "LOCATION STRUCTURE PHOTOGRAPH — the confirmed photograph of this exact place and its fixed structure: it is the SINGLE authority for the location, the structure's shape, proportions, materials, colors, openings and every permanent site detail. Never copy its camera framing, time of day or lighting — the shot text is the authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_animal_shelter_yard_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S7sh6.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 지정된 '오른쪽 귀'가 아닌 왼쪽 얼굴(턱 부근)에 스마트폰을 대고 있습니다.", "fix_en": "Change the character's pose so she is holding the smartphone to her right ear." }, { "issue_ko": "스마트폰이 '귀에 바짝 밀착'되지 않고 턱 옆으로 내려가 떨어져 있습니다.", "fix_en": "Move the smartphone up so it is pressed tightly against her ear." }, { "issue_ko": "스마트폰의 화면이 얼굴 쪽이 아닌 바깥쪽(카메라 방향)을 향하고 있습니다.", "fix_en": "Flip the smartphone so the screen side faces inward toward her head." }, { "issue_ko": "배경 우측 상단의 간판에 텍스트가 포함되어 있어 모든 텍스트 배제 규칙을 위반했습니다.", "fix_en": "Remove all text from the signboard in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the character's pose so she is holding the smartphone to her right ear.\n- Move the smartphone up so it is pressed tightly against her ear.\n- Flip the smartphone so the screen side faces inward toward her head.\n- Remove all text from the signboard 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." }, "S7sh6::ab_noconti": { "input_fingerprint": "c8fe9bde04ccb733", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 오른쪽 귀에 스마트폰을 바짝 밀착한 채 미간을 찌푸리고 입을 굳게 다문 측면 얼굴 클로즈업.\n\nLOCATION (lock): An animal shelter with secured dog kennels and practical veterinary and grooming work areas. The space accommodates injections, clipping, ointment treatment, and bathing. The shot takes place here — the attached LOCATION STRUCTURE PHOTOGRAPH is the single authority for this exact place — its fixed structure and permanent site details are LOCKED to it. No separate location photograph exists for this place. Build everything else strictly from the location text above and the shot text; the layout sketch (when attached) governs framing and placement only, and the shot text governs time of day, lighting and action.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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-yeong has her smartphone with her while attempting to contact Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 4, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "오른쪽 귀에 전화를 댄 측면 클로즈업과 지정된 야외 철창 배경을 모두 정확하게 구현한 최적의 결과물입니다." }, { "label": "C", "score": 4, "verdict_ko": "오른쪽 귀 방향은 맞으나, 고정된 야외 장소 레퍼런스를 무시하고 임의의 실내 공간으로 배경을 왜곡했습니다." }, { "label": "A", "score": 3, "verdict_ko": "배경은 레퍼런스를 잘 따랐으나, 텍스트에 명시된 오른쪽 귀가 아닌 왼쪽 귀에 스마트폰을 대고 있어 샷 구성을 위반했습니다." } ], "refs": [ { "label": "LOCATION STRUCTURE PHOTOGRAPH — the confirmed photograph of this exact place and its fixed structure: it is the SINGLE authority for the location, the structure's shape, proportions, materials, colors, openings and every permanent site detail. Never copy its camera framing, time of day or lighting — the shot text is the authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_animal_shelter_yard_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스마트폰의 위아래가 뒤집혀 있습니다(후면 카메라 렌즈가 귀 쪽이 아닌 턱 쪽에 위치함).", "fix_en": "Rotate the smartphone 180 degrees so the rear camera lenses are at the top near her ear." }, { "issue_ko": "화면 왼쪽 배경의 벽에 붙은 포스터에 텍스트가 포함되어 있습니다.", "fix_en": "Remove all text from the poster on the wall in the background on the left." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone 180 degrees so the rear camera lenses are at the top near her ear.\n- Remove all text from the poster on the wall in the background on the left.\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." }, "S7sh6::conti_ab_decision": { "fingerprint": "2805b5a9182bc34d", "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 지시문이 요구한 '얼굴 클로즈업' 샷 크기를 더 정확히 구현했으며, 배경에 수의학 작업 공간이 명확하게 묘사되어 있습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "1번 후보가 '얼굴 클로즈업'이라는 정확한 샷 크기를 준수했으며, 배경에 동물 보호소의 진료 및 작업 공간을 제대로 반영했습니다." } ] } }, "S7sh6": { "input_fingerprint": "c8fe9bde04ccb733", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 수리영이 오른쪽 귀에 스마트폰을 바짝 밀착한 채 미간을 찌푸리고 입을 굳게 다문 측면 얼굴 클로즈업.\n\nLOCATION (lock): An animal shelter with secured dog kennels and practical veterinary and grooming work areas. The space accommodates injections, clipping, ointment treatment, and bathing. The shot takes place here — the attached LOCATION STRUCTURE PHOTOGRAPH is the single authority for this exact place — its fixed structure and permanent site details are LOCKED to it. No separate location photograph exists for this place. Build everything else strictly from the location text above and the shot text; the layout sketch (when attached) governs framing and placement only, and the shot text governs time of day, lighting and action.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging 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-yeong has her smartphone with her while attempting to contact Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 4, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "오른쪽 귀에 전화를 댄 측면 클로즈업과 지정된 야외 철창 배경을 모두 정확하게 구현한 최적의 결과물입니다." }, { "label": "C", "score": 4, "verdict_ko": "오른쪽 귀 방향은 맞으나, 고정된 야외 장소 레퍼런스를 무시하고 임의의 실내 공간으로 배경을 왜곡했습니다." }, { "label": "A", "score": 3, "verdict_ko": "배경은 레퍼런스를 잘 따랐으나, 텍스트에 명시된 오른쪽 귀가 아닌 왼쪽 귀에 스마트폰을 대고 있어 샷 구성을 위반했습니다." } ], "refs": [ { "label": "LOCATION STRUCTURE PHOTOGRAPH — the confirmed photograph of this exact place and its fixed structure: it is the SINGLE authority for the location, the structure's shape, proportions, materials, colors, openings and every permanent site detail. Never copy its camera framing, time of day or lighting — the shot text is the authority for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_animal_shelter_yard_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스마트폰의 위아래가 뒤집혀 있습니다(후면 카메라 렌즈가 귀 쪽이 아닌 턱 쪽에 위치함).", "fix_en": "Rotate the smartphone 180 degrees so the rear camera lenses are at the top near her ear." }, { "issue_ko": "화면 왼쪽 배경의 벽에 붙은 포스터에 텍스트가 포함되어 있습니다.", "fix_en": "Remove all text from the poster on the wall in the background on the left." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone 180 degrees so the rear camera lenses are at the top near her ear.\n- Remove all text from the poster on the wall in the background on the left.\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.", "conti_ab": { "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 지시문이 요구한 '얼굴 클로즈업' 샷 크기를 더 정확히 구현했으며, 배경에 수의학 작업 공간이 명확하게 묘사되어 있습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "1번 후보가 '얼굴 클로즈업'이라는 정확한 샷 크기를 준수했으며, 배경에 동물 보호소의 진료 및 작업 공간을 제대로 반영했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S7sh6__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S7sh6__ab_noconti_sel.png" }, "ref_mode": "seed-bg+엔티티 (복잡구조물 A/B: 무콘티 승)", "lane_policy": "ab_select_ready" }, "S8sh1": { "input_fingerprint": "b4fa577ff08b3fe0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, neon-lit streets.\n\nSHOT TEXT (authoritative, Korean): 어스름한 저녁 네온사인이 켜진 길거리 포장마차 사이 도로, 수리영이 자전거 안장에 앉아 한쪽 무릎을 아래로 굽혀 페달을 밟아 누른 채 앞을 예리하게 주시하는 전신 구도.\n\nLOCATION (lock): A busy commercial district beside a harbor, filled with neon signs, taxis, street vendors, and food stalls. It connects to a large supermarket with merchandise aisles, stacked boxes, and bicycle parking outside. 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): Su-ri-yeong continues to carry the smartphone she used to call Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 4, "B": 3, "C": 7 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "A", "score": 4, "verdict_ko": "캐릭터와 배경은 잘 구현되었으나, 자전거의 오른쪽 핸들이 완전히 누락된 구조적 오류가 있어 감점되었습니다." }, { "label": "B", "score": 3, "verdict_ko": "레퍼런스에 없는 어두운 겉옷을 착용하여 복장 규정을 어겼으며, 핸들에 있는 스마트폰이 손과 뭉개져 있습니다." }, { "label": "C", "score": 7, "verdict_ko": "요청된 전신 구도, 캐릭터의 지정된 의상, 배경 요소 및 자전거를 타는 자세를 가장 정확하고 자연스럽게 구현했습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S8sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 수리영만 명시되어 있으나, 인도와 포장마차 주변 등 배경에 다른 사람들이 다수 등장합니다.", "fix_en": "Remove all background people from the sidewalks, stalls, and street, leaving only Su-ri-yeong in the shot." }, { "issue_ko": "수리영이 스마트폰을 계속 들고 있어야 한다는 지침과 달리, 양손으로 자전거 손잡이만 잡고 있으며 스마트폰은 거치대에 장착되어 있습니다.", "fix_en": "Remove the phone mount and make Su-ri-yeong hold the smartphone in one of her hands." } ] }, "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 sidewalks, stalls, and street, leaving only Su-ri-yeong in the shot.\n- Remove the phone mount and make Su-ri-yeong hold the smartphone in one of her hands.\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" }, "S8sh5": { "input_fingerprint": "abb540e110dc0c13", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, neon-lit streets.\n\nSHOT TEXT (authoritative, Korean): 수리영이 눈을 동그랗게 뜨고 혼란스러워하는 표정을 지은 얼굴 클로즈업.\n\nLOCATION (lock): A busy commercial district beside a harbor, filled with neon signs, taxis, street vendors, and food stalls. It connects to a large supermarket with merchandise aisles, stacked boxes, and bicycle parking outside. 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 dusk neon lighting, market-entrance street look, and the rider's clothing from the reference. Exclude the earlier cycling pose, bicycle-dominant framing, and unrelated street vendors from the tight facial composition.\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-yeong still has her smartphone with her while searching 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 6, "A": 4 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "얼굴 클로즈업과 눈을 동그랗게 뜬 표정을 가장 정확히 구현했으며, 스마트폰 방향과 복장 연속성도 올바릅니다." }, { "label": "C", "score": 6, "verdict_ko": "소품과 배경의 연속성은 준수하나, 프레임이 다소 넓어 타이트한 클로즈업 지시와 어긋나며 표정 연기가 약합니다." }, { "label": "A", "score": 4, "verdict_ko": "스마트폰 화면이 카메라를 향해 있어 소품 방향 지침을 위반했으며, 레퍼런스의 백팩 끈이 누락되었습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S8sh1_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "화면 오른쪽(캐릭터의 왼쪽) 어깨에 이전 샷과 레퍼런스에는 없던 백팩 끈이 있습니다.", "fix_en": "Remove the backpack strap from the character's left shoulder (viewer's right)." }, { "issue_ko": "배경 왼쪽 길거리에 샷 텍스트에 명시되지 않은 사람의 실루엣이 있습니다.", "fix_en": "Remove the person's silhouette from the sidewalk in the left background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the backpack strap from the character's left shoulder (viewer's right).\n- Remove the person's silhouette from the sidewalk in the left 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": "prev+엔티티", "lane_policy": "ab_select_bypass:prev" }, "S8sh7": { "input_fingerprint": "b0b228f7750796d2", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, neon-lit streets.\n\nSHOT TEXT (authoritative, Korean): 어두운 마트 밖 거리, 수리영이 한 손에 스마트폰을 쥔 채 액정 화면을 심각하게 내려다보는 정지 구도.\n\nLOCATION (lock): A busy commercial district beside a harbor, filled with neon signs, taxis, street vendors, and food stalls. It connects to a large supermarket with merchandise aisles, stacked boxes, and bicycle parking outside. 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 neon-lit market entrance, darkening street ambience, and the woman's clothing from the reference. Exclude the earlier riding action and keep the bicycle and food-stall activity outside the new phone-focused frame unless minimally visible in the background.\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-yeong retains her smartphone and uses it outside the mart to try Min-sook again.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "스마트폰의 후면이 카메라를 향하게 쥐어 소품의 물리적 방향을 올바르게 구현했으며, 이전 컷의 배경 구조를 정확히 유지했습니다." }, { "label": "B", "score": 4, "verdict_ko": "인물이 화면을 바라보고 있음에도 스마트폰 액정이 카메라 쪽을 향하고 있어 소품 방향 규칙을 위반했습니다." }, { "label": "C", "score": 3, "verdict_ko": "스마트폰 액정이 카메라를 향해 방향 규칙을 어겼으며, 이전 컷의 공간 구조(마트와 도로의 좌우 위치)가 반전되었습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S8sh1_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": "배경 왼쪽 포장마차 주변에 프롬프트에 명시되지 않은 인물들이 추가되어 있습니다.", "fix_en": "Remove the people standing around the food stall on the left side of the street." }, { "issue_ko": "자전거가 배경에 최소한으로 보이는 대신 화면 우측 전경에 크게 자리 잡고 있습니다.", "fix_en": "Remove the bicycle from the bottom right foreground." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the people standing around the food stall on the left side of the street.\n- Remove the bicycle from the bottom right foreground.\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": "prev+엔티티", "lane_policy": "ab_select_bypass:prev" }, "S9sh1": { "input_fingerprint": "2507d58083181ca4", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night.\n\nSHOT TEXT (authoritative, Korean): 어둠이 짙게 깔린 해안도로 위, 상체를 굽힌 채 한쪽 자전거 페달에 체중을 실어 내리밟는 수리영의 mid-action, 옷자락이 주행 반대 방향으로 휘날리는 찰나.\n\nLOCATION (lock): A coastal road beside the dark sea, used as a bicycle route back toward the residential area. 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): Su-ri-yeong is still traveling on her bicycle and retains the smartphone used moments earlier.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 5, "A": 4 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 카메라 구도(후방)와 프레이밍을 유일하게 준수했으며 스마트폰 소지 상태를 반영했으나, 상의 의상이 지정된 레퍼런스와 다소 다릅니다." }, { "label": "B", "score": 5, "verdict_ko": "캐릭터 레퍼런스의 의상(후드티)은 훌륭하게 일치하지만, 최우선 조건인 스토리보드의 후방 앵글을 무시하고 정면을 촬영했습니다." }, { "label": "A", "score": 4, "verdict_ko": "스토리보드의 지정된 앵글을 위반하여 피사체 정면을 묘사했으며, 레퍼런스와 일치하지 않는 의상과 스마트폰 소지 조건 누락 등의 감점 요소가 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/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": "스토리보드에 지시된 주행 방향(화면 우측을 향함)과 반대로 인물이 좌측을 향해 자전거를 타고 있어 전체 구도가 좌우 반전되었습니다.", "fix_en": "Mirror the staging horizontally so the character rides toward the right side of the frame, matching the storyboard's heading." }, { "issue_ko": "오른쪽 팔이 완전히 누락되어 빈 소매만 뒤로 휘날리고 있습니다.", "fix_en": "Render the character's right arm extending forward to grip the right side of the handlebar." }, { "issue_ko": "유지해야 할 스마트폰이 손에 쥐어지거나 주머니에 들어있지 않고 바지 겉면에 물리적으로 불가능하게 접착되어 있습니다.", "fix_en": "Render the smartphone either securely held in one of her hands or realistically placed inside a visible pocket." }, { "issue_ko": "왼쪽 자전거 핸들을 잡고 있는 손의 형태가 뭉개져 금속 부품과 하나로 융합되어 있습니다.", "fix_en": "Render the left hand with distinct, anatomically correct fingers cleanly gripping the brake hood." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Mirror the staging horizontally so the character rides toward the right side of the frame, matching the storyboard's heading.\n- Render the character's right arm extending forward to grip the right side of the handlebar.\n- Render the smartphone either securely held in one of her hands or realistically placed inside a visible pocket.\n- Render the left hand with distinct, anatomically correct fingers cleanly gripping the brake hood.\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::ab_conti": { "input_fingerprint": "034048c8c6a1e849", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 정체불명의 습격자(검은 인영·붉은 눈의 남자)가 굳게 닫힌 옥탑방 철문 현관 앞에 우뚝 선 짙은 실루엣의 뒷모습.\n\nLOCATION (lock): The exposed rooftop of an old residential villa, holding a water tank, clotheslines, a low platform with drying herbs, and a metal entrance door to the rooftop apartment. A window reveals silhouettes inside the dwelling. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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): Dried seokchangpo remains scattered on the rooftop platform, and an advertising leaflet is lodged in the closed iron door's gap.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정체불명의 습격자 (활성 상태) (한국인으로 보이는 성인 남성, 어둠에 가려진 얼굴, 식별하기 어려운 머리 형태, 남성 얼굴 윤곽) — 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": 6, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 6, "verdict_ko": "틈새에 끼워진 전단지와 우뚝 선 인물의 실루엣을 정확히 구현했으나, 구조물 레퍼런스(회색 벽면과 녹색 문)를 반영하지 못해 감점됨." }, { "label": "A", "score": 5, "verdict_ko": "전단지와 인물 배치는 무난하나 디테일이 다소 떨어지며, 동일하게 구조물 레퍼런스의 형태와 재질을 위반함." }, { "label": "B", "score": 4, "verdict_ko": "전단지가 문 틈새가 아닌 정면에 붙어 있어 지침을 위반했으며, 구조물 디자인 역시 레퍼런스와 불일치함." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L03B03.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S10sh3.png" }, { "label": "CHARACTER REFERENCE — 정체불명의 습격자 (활성 상태): the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "옥탑방 외벽(적벽돌)과 문(회색 철문)이 STRUCTURE LOOK 사진(회색 콘크리트 외벽, 녹색 문)과 일치하지 않습니다.", "fix_en": "Change the rooftop structure's exterior walls to grey concrete and the door to green to match the STRUCTURE LOOK photograph." }, { "issue_ko": "남자가 굳게 닫힌 철문 앞이 아닌, 가스 계량기가 있는 빈 벽면을 마주 보고 서 있습니다.", "fix_en": "Shift the man's position to the right so he is standing directly in front of the closed door." }, { "issue_ko": "평상 위에 마른 석창포가 흩어져 있지 않고, 멍석 위에 굵은 장작더미가 쌓여 있습니다.", "fix_en": "Replace the thick firewood and mat on the wooden platform with scattered dried herbs." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the rooftop structure's exterior walls to grey concrete and the door to green to match the STRUCTURE LOOK photograph.\n- Shift the man's position to the right so he is standing directly in front of the closed door.\n- Replace the thick firewood and mat on the wooden platform with scattered dried herbs.\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." }, "S10sh3::ab_noconti": { "input_fingerprint": "ddf2b9ae232099a3", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 정체불명의 습격자(검은 인영·붉은 눈의 남자)가 굳게 닫힌 옥탑방 철문 현관 앞에 우뚝 선 짙은 실루엣의 뒷모습.\n\nLOCATION (lock): The exposed rooftop of an old residential villa, holding a water tank, clotheslines, a low platform with drying herbs, and a metal entrance door to the rooftop apartment. A window reveals silhouettes inside the dwelling. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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): Dried seokchangpo remains scattered on the rooftop platform, and an advertising leaflet is lodged in the closed iron door's gap.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정체불명의 습격자 (활성 상태) (한국인으로 보이는 성인 남성, 어둠에 가려진 얼굴, 식별하기 어려운 머리 형태, 남성 얼굴 윤곽) — 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": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "평상 위의 약초와 철문의 구조 등 공간 요소가 가장 정확하나, 전단지가 문틈이 아닌 표면에 위치한 점이 아쉽습니다." }, { "label": "B", "score": 6, "verdict_ko": "전반적인 지시를 잘 따랐으나, 기준 사진과 달리 철문의 손잡이와 경첩 방향이 반대로 뒤집혀 있습니다." }, { "label": "A", "score": 3, "verdict_ko": "뒷모습을 지정했음에도 후두부에 붉은 눈이 발광하고 있어 물리적으로 불가능한 해부학적 묘사(하드 위반)입니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L03B03.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.png" }, { "label": "CHARACTER REFERENCE — 정체불명의 습격자 (활성 상태): the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "옥탑방 구조물이 'Structure Look' 레퍼런스(회색 콘크리트, 녹색 문)를 무시하고 위치 사진의 붉은 벽돌과 회색 철문을 그대로 복사했습니다.", "fix_en": "Change the rooftop structure's red brick walls, grey metal door, and barred window to match the grey concrete walls, green door, and openings shown in the Structure Look reference photograph." }, { "issue_ko": "광고 전단지가 닫힌 문의 틈새에 끼워져 있지 않고 문 정면 표면에 붙어 있습니다.", "fix_en": "Move the advertising leaflet from the flat surface of the door and wedge it directly into the gap between the door and the door frame." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the rooftop structure's red brick walls, grey metal door, and barred window to match the grey concrete walls, green door, and openings shown in the Structure Look reference photograph.\n- Move the advertising leaflet from the flat surface of the door and wedge it directly into the gap between the door and the door 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." }, "S10sh3::conti_ab_decision": { "fingerprint": "04baa04f8166afe2", "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "2번 후보는 피사체를 완벽한 뒷모습의 짙은 실루엣으로 정확히 연출했으며, 철문에 전단지가 붙어 있는 디테일을 포함하여 지시문에 더 부합합니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 인물이 엉뚱한 벽면이 아닌 지시문대로 철문 현관 바로 앞에 서 있는 구도를 정확히 구현했습니다." } ] } }, "S10sh3": { "input_fingerprint": "034048c8c6a1e849", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 정체불명의 습격자(검은 인영·붉은 눈의 남자)가 굳게 닫힌 옥탑방 철문 현관 앞에 우뚝 선 짙은 실루엣의 뒷모습.\n\nLOCATION (lock): The exposed rooftop of an old residential villa, holding a water tank, clotheslines, a low platform with drying herbs, and a metal entrance door to the rooftop apartment. A window reveals silhouettes inside the dwelling. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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): Dried seokchangpo remains scattered on the rooftop platform, and an advertising leaflet is lodged in the closed iron door's gap.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 정체불명의 습격자 (활성 상태) (한국인으로 보이는 성인 남성, 어둠에 가려진 얼굴, 식별하기 어려운 머리 형태, 남성 얼굴 윤곽) — 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": 6, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 6, "verdict_ko": "틈새에 끼워진 전단지와 우뚝 선 인물의 실루엣을 정확히 구현했으나, 구조물 레퍼런스(회색 벽면과 녹색 문)를 반영하지 못해 감점됨." }, { "label": "A", "score": 5, "verdict_ko": "전단지와 인물 배치는 무난하나 디테일이 다소 떨어지며, 동일하게 구조물 레퍼런스의 형태와 재질을 위반함." }, { "label": "B", "score": 4, "verdict_ko": "전단지가 문 틈새가 아닌 정면에 붙어 있어 지침을 위반했으며, 구조물 디자인 역시 레퍼런스와 불일치함." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L03B03.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S10sh3.png" }, { "label": "CHARACTER REFERENCE — 정체불명의 습격자 (활성 상태): the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "옥탑방 외벽(적벽돌)과 문(회색 철문)이 STRUCTURE LOOK 사진(회색 콘크리트 외벽, 녹색 문)과 일치하지 않습니다.", "fix_en": "Change the rooftop structure's exterior walls to grey concrete and the door to green to match the STRUCTURE LOOK photograph." }, { "issue_ko": "남자가 굳게 닫힌 철문 앞이 아닌, 가스 계량기가 있는 빈 벽면을 마주 보고 서 있습니다.", "fix_en": "Shift the man's position to the right so he is standing directly in front of the closed door." }, { "issue_ko": "평상 위에 마른 석창포가 흩어져 있지 않고, 멍석 위에 굵은 장작더미가 쌓여 있습니다.", "fix_en": "Replace the thick firewood and mat on the wooden platform with scattered dried herbs." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the rooftop structure's exterior walls to grey concrete and the door to green to match the STRUCTURE LOOK photograph.\n- Shift the man's position to the right so he is standing directly in front of the closed door.\n- Replace the thick firewood and mat on the wooden platform with scattered dried herbs.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "2번 후보는 피사체를 완벽한 뒷모습의 짙은 실루엣으로 정확히 연출했으며, 철문에 전단지가 붙어 있는 디테일을 포함하여 지시문에 더 부합합니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 인물이 엉뚱한 벽면이 아닌 지시문대로 철문 현관 바로 앞에 서 있는 구도를 정확히 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S10sh3__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S10sh3__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+seed+엔티티 (복잡구조물 A/B: 콘티 승)", "lane_policy": "ab_select_ready", "plate_select": { "candidates": { "A": "L03B02", "B": "L03B03" }, "assigned": "L03B02", "choice": "Candidate A", "confident": true, "reason_ko": "지문에서 '굳게 닫힌 옥탑방 철문 현관 앞'을 명시하고 있으며, 두 후보 이미지가 동일한 옥탑방 현관문 앞 공간을 보여주므로 현재 할당된 A를 유지합니다.", "kept": "L03B02" } }, "S10sh6": { "input_fingerprint": "b7e31c749dbafbb1", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 희미한 불이 켜진 옥탑방 창문 유리에 두 사람의 상체가 거칠게 겹쳐진 실루엣 그림자.\n\nLOCATION (lock): The exposed rooftop of an old residential villa, holding a water tank, clotheslines, a low platform with drying herbs, and a metal entrance door to the rooftop apartment. A window reveals silhouettes inside the dwelling. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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 forced iron entrance has been left slightly open while the intruder attacks Min-sook inside. Min-sook still bears the small circular wrist 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": 3, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "창문에 거칠게 겹쳐진 상체 실루엣과 약간 열린 철문을 텍스트에 맞는 프레이밍으로 가장 정확하게 구현했습니다." }, { "label": "B", "score": 3, "verdict_ko": "실루엣이 아닌 인물의 형체가 뚜렷하게 보이며, 불필요하게 넓은 와이드 샷으로 텍스트의 의도를 훼손했습니다." }, { "label": "C", "score": 5, "verdict_ko": "창문에 실루엣이 비치지만 두 인물이 분리되어 있어 '거칠게 겹쳐진'이라는 핵심 묘사를 충족하지 못했습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L03B03.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.png" } ], "critique": { "issues": [ { "issue_ko": "옥탑방 외벽이 붉은 벽돌로 되어 있으나, STRUCTURE LOOK 사진의 지침에 따라 회색 콘크리트 벽이어야 합니다.", "fix_en": "Change the exterior walls of the rooftop structure from red brick to plain grey concrete to match the Structure Look photo." }, { "issue_ko": "창문이 벽의 대부분을 차지하는 거대한 유리문 형태로 묘사되었으나, STRUCTURE LOOK 사진에 있는 작고 높은 위치의 창문 형태여야 합니다.", "fix_en": "Replace the massive glass window with a small, higher-placed sliding window matching the Structure Look photo." }, { "issue_ko": "옥탑방 철문이 회색으로 되어 있으나, STRUCTURE LOOK 사진에 따라 녹색 문이어야 합니다.", "fix_en": "Change the color of the slightly open metal entrance door from grey to green." }, { "issue_ko": "프롬프트에 명시된 '약초를 말리는 낮은 평상(low platform)'이 누락되었고 바닥에 멍석만 깔려 있습니다.", "fix_en": "Replace the flat mat on the floor with a low wooden platform holding the drying herbs." }, { "issue_ko": "프롬프트에 요구된 빨랫줄(clotheslines)이 물탱크 주변의 빈 공간에 보이지 않습니다.", "fix_en": "Add clotheslines in the open space next to the blue water tank." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the exterior walls of the rooftop structure from red brick to plain grey concrete to match the Structure Look photo.\n- Replace the massive glass window with a small, higher-placed sliding window matching the Structure Look photo.\n- Change the color of the slightly open metal entrance door from grey to green.\n- Replace the flat mat on the floor with a low wooden platform holding the drying herbs.\n- Add clotheslines in the open space next to the blue water tank.\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+seed만 (배경 전용)", "lane_policy": "ab_select_bypass:bg_only", "plate_select": { "candidates": { "A": "L03B02", "B": "L03B03" }, "assigned": "L03B02", "choice": "Candidate A", "confident": true, "reason_ko": "지문에서 요구하는 '옥탑방 창문'이 동일하게 포함된 옥상 전경 이미지들로, 두 후보 간 명확한 차이나 변경할 만한 근거가 없으므로 기존 할당된 후보를 유지합니다.", "kept": "L03B02" } }, "S11sh4": { "input_fingerprint": "914e9e83a8418a96", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 옥상 바닥에 두 발을 딛고 선 수리영이 불쾌감에 코끝을 찡그리고 미간을 잔뜩 좁힌 얼굴 클로즈업.\n\nLOCATION (lock): A narrow residential alley leading into the courtyard of a low-rise multi-unit villa. A secluded bicycle spot and an exterior stairway connect the courtyard to the rooftop apartment and its half-open metal door. 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): Su-ri-yeong's bicycle remains roughly parked in the rear corner of the courtyard.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 4 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 어깨 너머 시선 구도를 정확히 따랐으며, 찡그린 표정, 묶은 머리와 후드티 의상, 반쯤 열린 문과 자전거까지 모든 지시사항을 충실히 반영했습니다." }, { "label": "A", "score": 5, "verdict_ko": "얼굴 표정과 배경은 지시문과 유사하나, 레퍼런스와 달리 머리를 풀고 있으며 후드티가 아닌 반팔을 입고 있습니다." }, { "label": "B", "score": 4, "verdict_ko": "캐릭터의 의상과 표정은 일치하지만, 정면을 향하고 있어 스토리보드의 카메라 구도를 어겼으며 문이 닫혀 있고 자전거가 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S11sh4.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "배경의 반쯤 열린 철문에 부착된 표지판에 텍스트가 적혀 있습니다.", "fix_en": "Remove the text from the rectangular sign on the metal door." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the text from the rectangular sign on the metal door.\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" }, "S11sh5": { "input_fingerprint": "81cf445b827be7a9", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영의 시야 정면 끝에 놓인 옥탑방 현관 철문이 반쯤 열린 채 캄캄한 실내 공간을 스산하게 드러낸 정지 구도.\n\nLOCATION (lock): A narrow residential alley leading into the courtyard of a low-rise multi-unit villa. A secluded bicycle spot and an exterior stairway connect the courtyard to the rooftop apartment and its half-open metal door. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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-yeong's bicycle remains in the rear courtyard, and the rooftop iron entrance stands half open.\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": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "제시된 시점 구도와 캄캄한 실내를 잘 구현했고 마당에 있어야 할 자전거를 프레임에서 올바르게 배제했으나, 구조물 기준 사진의 회색 외벽 재질은 반영되지 않았습니다." }, { "label": "B", "score": 5, "verdict_ko": "구도와 묘사는 무난하나 구조물 기준 사진의 형태를 전혀 반영하지 않았으며, 후원에 있어야 할 자전거를 옥상에 배치하는 위치 오류를 범했습니다." }, { "label": "C", "score": 3, "verdict_ko": "위치 사진의 계단을 허공에 띄운 채 구조물 사진과 억지로 합성하여 물리적으로 불가능한 구도를 생성한 심각한 위반입니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L12B03.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.png" } ], "critique": { "issues": [ { "issue_ko": "열린 문 안쪽은 캄캄한 반면 바로 옆 창문은 실내 조명이 밝게 켜져 있어, '캄캄한 실내 공간'이라는 프롬프트 지시와 논리적으로 모순됩니다.", "fix_en": "Turn off the light glowing through the window so the entire interior of the rooftop apartment appears completely pitch-black." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Turn off the light glowing through the window so the entire interior of the rooftop apartment appears completely pitch-black.\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+seed만 (배경 전용)", "lane_policy": "ab_select_bypass:bg_only" }, "S12sh5": { "input_fingerprint": "94bff102aae76f48", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, 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): 열린 방문 너머 방바닥에 흥건한 검붉은 혈흔과 핏빛 발자국이 찍혀 있는 하단 전경.\n\nLOCATION (lock): A compact rooftop apartment with a living room, dining table, main bedroom, and a young woman's bedroom behind curtains. The living room contains a cooled pot, an operating television, two teacups, and a dim floor lamp; the bedroom floor and curtains become the focus of the bloody discovery. 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 retains a cooled pot, a switched-on TV, and two teacups with one overturned. Su-ri-yeong's room contains pooled blood and bloody footprints leading to the curtain.\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": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "방문 너머 방 안의 혈흔과 피 묻은 커튼을 지시대로 정확히 배치했으나, 거실 바닥에 찍힌 발자국의 방향들이 다소 비논리적입니다." }, { "label": "B", "score": 3, "verdict_ko": "혈흔이 방 안이 아닌 거실에 위치해 공간 설정과 어긋나며, 필수 소품인 식탁과 찻잔 등이 프레임에서 완전히 누락되었습니다." }, { "label": "C", "score": 5, "verdict_ko": "탁자 위 쓰러진 찻잔 등 소품 디테일과 발자국 동선은 매우 훌륭하나, 가장 중요한 혈흔 웅덩이가 방 안이 아닌 거실에 있어 샷 텍스트를 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B05.png" } ], "critique": { "issues": [ { "issue_ko": "식탁 위에 쓰러져 있는 물건이 프롬프트에서 요구한 찻잔이 아니라 주전자입니다.", "fix_en": "Replace the overturned teapot on the dining table with an overturned teacup." }, { "issue_ko": "전경 바닥의 핏빛 발자국들이 해부학적으로 틀렸습니다 (같은 쪽 발이 연속해서 찍혀 있고 발가락 형태가 비정상적입니다).", "fix_en": "Redraw the bloody footprints to follow a realistic, alternating left-right stepping sequence with anatomically correct five-toed human foot shapes." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Replace the overturned teapot on the dining table with an overturned teacup.\n- Redraw the bloody footprints to follow a realistic, alternating left-right stepping sequence with anatomically correct five-toed human foot shapes.\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)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B06", "choice": "Candidate G", "confident": true, "reason_ko": "지문에서 '열린 방문 너머 방바닥'과 '하단 전경'을 명시하고 있으므로, 우측 전경에 열린 문이 걸려 있고 방바닥이 넓게 잘 보이는 구도인 Candidate G가 가장 적합합니다.", "kept": "L04B06" } }, "S12sh9": { "input_fingerprint": "9afdff6c8b2deb96", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, 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): 젖혀진 커튼 뒤로 강민숙이 방바닥에 앉은 채 고개를 푹 수그리고 굳어있는 전신.\n\nLOCATION (lock): A compact rooftop apartment with a living room, dining table, main bedroom, and a young woman's bedroom behind curtains. The living room contains a cooled pot, an operating television, two teacups, and a dim floor lamp; the bedroom floor and curtains become the focus of the bloody discovery. 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 room layout, open doorway, dark blood on the floor, bloody footprints, and dim interior lighting from the reference. Exclude the earlier empty floor-centered framing and add only the seated body behind the opened curtain as specified.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is seated motionless on the bedroom floor behind the curtain, with her torso held in a fixed seated position and her head dropped forward. Her legs remain on the floor, her torn shoulder and collarbone are exposed, and one hand grips a crumpled photograph.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Min-sook is seated motionless on the bedroom floor behind the curtain, with her torso held in a fixed seated position and her head dropped forward, her torn shoulder and collarbone bleeding, and a crumpled photograph gripped in her hand. The surrounding pool and footprints of blood remain in place.\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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": 9, "B": 4, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 9, "verdict_ko": "레퍼런스의 구도를 복사하지 않고 '젖혀진 커튼 뒤'라는 새로운 시점을 훌륭하게 구현했으며, 찢어진 어깨와 사진 소품, 중력에 순응하는 자세를 정확히 반영했습니다." }, { "label": "B", "score": 4, "verdict_ko": "레퍼런스의 카메라 구도를 그대로 답습하여 금지 지침을 어겼으며, 찢어지고 노출된 어깨 지문이 반영되지 않았습니다." }, { "label": "A", "score": 3, "verdict_ko": "카메라 구도를 그대로 복사했을 뿐만 아니라, 방 안쪽 창문에 있어야 할 커튼을 문가로 임의로 이동시켜 공간의 일관성을 훼손했습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S12sh5_sel.png" }, { "label": "CHARACTER REFERENCE — 강민숙: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "사진의 앞면이 인물의 눈이 아닌 카메라 방향(바깥쪽)을 향하고 있습니다.", "fix_en": "Rotate the crumpled photograph so its functional front side faces the character's eyes, showing only the back to the camera." }, { "issue_ko": "한 손으로 구겨진 사진을 쥐어야 한다는 지시와 달리 두 손으로 사진을 들고 있습니다.", "fix_en": "Have only one hand grip the crumpled photograph, and leave the other hand resting empty on her lap." }, { "issue_ko": "찢어진 어깨와 쇄골이 노출되어야 하나, 피 묻은 옷에 가려 피부가 보이지 않습니다.", "fix_en": "Expose the bare skin of her shoulder and collarbone through the torn clothing." }, { "issue_ko": "거실 배경에 있어야 할 식탁과 찻잔, 냄비가 사라졌습니다.", "fix_en": "Restore the dining table with the teacups and pot to the left side of the living room, matching the reference image." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the crumpled photograph so its functional front side faces the character's eyes, showing only the back to the camera.\n- Have only one hand grip the crumpled photograph, and leave the other hand resting empty on her lap.\n- Expose the bare skin of her shoulder and collarbone through the torn clothing.\n- Restore the dining table with the teacups and pot to the left side of the living room, matching the reference image.\n\nPRESERVE EVERYTHING ELSE EXACTLY as in the original photograph —\nevery person, pose, face, garment, object, set feature, lighting,\ncamera and framing stay identical. No text, captions or\nwatermarks anywhere.", "ref_mode": "prev+엔티티" }, "S12sh16": { "input_fingerprint": "ade16cc1d880493a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, 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): 강민숙의 손목 안쪽 피부에 새겨진 작은 원 형태의 표식이 붉은색으로 짙게 물들어 있는 클로즈업.\n\nLOCATION (lock): A compact rooftop apartment with a living room, dining table, main bedroom, and a young woman's bedroom behind curtains. The living room contains a cooled pot, an operating television, two teacups, and a dim floor lamp; the bedroom floor and curtains become the focus of the bloody discovery. 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 bedroom lighting, the seated body's fixed position, damaged clothing, and ongoing bloody state from the reference. Exclude the full-body composition, curtain-wide view, and unrelated floor details from the wrist close-up; emphasize only the circle mark turning red.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is seated motionless on the bedroom floor behind the curtain, with her torso held in a fixed seated position and her head dropped forward. Her legs remain on the floor, her torn shoulder and collarbone are exposed, and one hand grips a crumpled photograph.\n\nIMMOBILE BODIES OBEY GRAVITY: a person who is dead or unconscious\nexerts NO muscular effort. Every part of their body — head, torso,\narms, hands, fingers, legs — rests fully on whatever supports it\n(floor, wall, furniture, their own lap) and hangs or slumps with\ngravity. NEVER show any part of an immobile person's body lifted,\nraised, held up in the air, or posed as if presenting something:\nan object in their grip stays clenched in a hand that itself lies\nfallen on a support — the hand does not hold the object up. If the\ncanonical pose leaves a body part unspecified, resolve it as the\nmost gravity-compliant, fully-supported position.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Min-sook remains seated motionless on the bedroom floor behind the curtain amid the blood, with her torso held in a fixed seated position and her head dropped forward, and one hand grips a crumpled photograph. Her wrist mark is now red, and a red circle has appeared on the wall above the bed.\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 중년 여성, 40세, 성숙한 얼굴, 자연스러운 짙은색 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": "의식이 없는 상태임에도 팔을 공중에 들고 있어 중력 지침(Immobile bodies obey gravity)을 심각하게 위반했습니다." }, { "label": "B", "score": 7, "verdict_ko": "팔을 바닥에 눕혀 중력 규칙을 정확히 따랐으며, 손목의 붉은 표식과 구겨진 사진, 배경의 거실 디테일까지 훌륭하게 재현했습니다." }, { "label": "C", "score": 2, "verdict_ko": "팔이 지지대 없이 허공에 떠 있어 중력 지침을 위반했으며, 쥐고 있어야 할 사진 소품도 프레임 내에서 누락되었습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S12sh9_sel.png" } ], "critique": { "issues": [ { "issue_ko": "지시문에서 배제하라고 한 인물의 반신과 배경이 모두 포함되어 있으며, 요구된 손목 클로즈업 구도가 아닙니다.", "fix_en": "Crop the image to a tight close-up of the inner wrist, completely removing the body, floor, and background room." }, { "issue_ko": "손목의 표식이 '작은 원'이 아닌 크고 깊게 패인 입체적인 상처로 묘사되었습니다.", "fix_en": "Change the large, deep wound on the wrist into a small, flat circular mark dyed red." }, { "issue_ko": "레퍼런스에서 무릎 부근에 있던 왼팔이 카메라 앞쪽으로 길게 뻗어 있어 '자세 유지' 지침을 위반했습니다.", "fix_en": "Move the left arm back so the hand rests near the lap, perfectly matching the pose in the reference." }, { "issue_ko": "오른손에 쥔 사진의 앞면(기능적 측면)이 캐릭터의 시선이 아닌 카메라를 향하고 있습니다.", "fix_en": "Flip the photograph in the right hand so the printed side faces the character and only the blank back faces the camera." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Crop the image to a tight close-up of the inner wrist, completely removing the body, floor, and background room.\n- Change the large, deep wound on the wrist into a small, flat circular mark dyed red.\n- Move the left arm back so the hand rests near the lap, perfectly matching the pose in the reference.\n- Flip the photograph in the right hand so the printed side faces the character and only the blank back faces the camera.\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": "prev+엔티티", "locked_char_refs_excluded": [ "강민숙(C04)" ] }, "S13sh3::ab_conti": { "input_fingerprint": "551efa157f39f6de", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 초점을 잃은 수리영이 계단 아래 시멘트 바닥에 두 팔을 늘어뜨리고 멍하니 서 있는 전신.\n\nLOCATION (lock): The exterior of an old multi-unit villa, including the rooftop apartment, its descending access stairs, and the ground-level area where an emergency vehicle arrives. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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): Until the later disappearance is revealed, Min-sook's seated corpse, the blood, her red wrist mark, and the red circle remain in Su-ri-yeong's room behind 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 4, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "유일하게 원본 사진 발코니의 불필요한 인물을 제거하여 엄격한 인물 제한 규칙을 완벽히 준수했으며, 전신 샷의 요구사항을 충실히 반영했습니다." }, { "label": "C", "score": 4, "verdict_ko": "레이아웃의 측면 포즈와 팔을 늘어뜨린 디테일을 가장 잘 구현했으나, 발코니에 지시되지 않은 인물이 그대로 남아있어 하드 위반(Extra body)에 해당합니다." }, { "label": "A", "score": 3, "verdict_ko": "레이아웃의 포즈를 무시한 채 정면을 향해 서 있으며, 발코니에 허용되지 않은 외부 인물이 포함되어 프롬프트 충실도가 가장 낮습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L11B02.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S13sh3.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 정면을 보고 있으나, 레이아웃 스케치에 지정된 대로 오른쪽을 향해 서 있는 측면 모습(프로필)이어야 합니다.", "fix_en": "Rotate the character's body and head to face the right side of the frame in a side profile, matching the layout sketch." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the character's body and head to face the right side of the frame in a side profile, matching the layout sketch.\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." }, "S13sh3::ab_noconti": { "input_fingerprint": "8e14a78ebfb76ec0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 초점을 잃은 수리영이 계단 아래 시멘트 바닥에 두 팔을 늘어뜨리고 멍하니 서 있는 전신.\n\nLOCATION (lock): The exterior of an old multi-unit villa, including the rooftop apartment, its descending access stairs, and the ground-level area where an emergency vehicle arrives. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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): Until the later disappearance is revealed, Min-sook's seated corpse, the blood, her red wrist mark, and the red circle remain in Su-ri-yeong's room behind 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 6, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "계단 아래 멍하니 서 있는 전신 구도와 장소, 청바지 의상은 잘 구현되었으나 상의가 후드티 대신 재킷 형태로 다소 변형되었습니다." }, { "label": "C", "score": 6, "verdict_ko": "요구된 샷의 구도와 인물의 자세는 정확하나, 레퍼런스와 달리 상하의 의상 색상과 종류가 일치하지 않습니다." }, { "label": "A", "score": 3, "verdict_ko": "주인공의 의상과 자세는 훌륭하지만, 장소 레퍼런스 사진 우측 상단 발코니에 있던 지문에 없는 인물까지 그대로 생성된 치명적 위반이 있습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L11B02.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터가 레퍼런스 이미지에 있는 백팩을 메고 있지 않으며, 풀오버 후드티 대신 집업 재킷을 입고 있습니다.", "fix_en": "Change her zip-up jacket to a pullover hoodie and add the olive green backpack to her right shoulder 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- Change her zip-up jacket to a pullover hoodie and add the olive green backpack to her right shoulder 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." }, "S13sh3::conti_ab_decision": { "fingerprint": "6494f476c42af687", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 장소 설정에 명시된 구급차 도착 상황을 배경에 충실히 반영하였으며, 초점을 잃고 서 있는 인물의 전신을 영화적으로 더 자연스럽게 연출했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "배경에 구급/경찰 차량의 경광등이 명확히 묘사되어 장소의 설정(응급 차량 도착)을 충실히 반영했으며, 초점을 잃고 멍하니 서 있는 인물의 상태도 자연스럽게 연출되었습니다." } ] } }, "S13sh3": { "input_fingerprint": "551efa157f39f6de", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 초점을 잃은 수리영이 계단 아래 시멘트 바닥에 두 팔을 늘어뜨리고 멍하니 서 있는 전신.\n\nLOCATION (lock): The exterior of an old multi-unit villa, including the rooftop apartment, its descending access stairs, and the ground-level area where an emergency vehicle arrives. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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): Until the later disappearance is revealed, Min-sook's seated corpse, the blood, her red wrist mark, and the red circle remain in Su-ri-yeong's room behind 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 4, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "유일하게 원본 사진 발코니의 불필요한 인물을 제거하여 엄격한 인물 제한 규칙을 완벽히 준수했으며, 전신 샷의 요구사항을 충실히 반영했습니다." }, { "label": "C", "score": 4, "verdict_ko": "레이아웃의 측면 포즈와 팔을 늘어뜨린 디테일을 가장 잘 구현했으나, 발코니에 지시되지 않은 인물이 그대로 남아있어 하드 위반(Extra body)에 해당합니다." }, { "label": "A", "score": 3, "verdict_ko": "레이아웃의 포즈를 무시한 채 정면을 향해 서 있으며, 발코니에 허용되지 않은 외부 인물이 포함되어 프롬프트 충실도가 가장 낮습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L11B02.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S13sh3.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 정면을 보고 있으나, 레이아웃 스케치에 지정된 대로 오른쪽을 향해 서 있는 측면 모습(프로필)이어야 합니다.", "fix_en": "Rotate the character's body and head to face the right side of the frame in a side profile, matching the layout sketch." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the character's body and head to face the right side of the frame in a side profile, matching the layout sketch.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 장소 설정에 명시된 구급차 도착 상황을 배경에 충실히 반영하였으며, 초점을 잃고 서 있는 인물의 전신을 영화적으로 더 자연스럽게 연출했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "배경에 구급/경찰 차량의 경광등이 명확히 묘사되어 장소의 설정(응급 차량 도착)을 충실히 반영했으며, 초점을 잃고 멍하니 서 있는 인물의 상태도 자연스럽게 연출되었습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S13sh3__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S13sh3__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+seed+엔티티 (복잡구조물 A/B: 콘티 승)", "lane_policy": "ab_select_ready", "plate_select": { "candidates": { "A": "L11B01", "B": "L11B02" }, "assigned": "L11B02", "choice": "Candidate B", "confident": true, "reason_ko": "지문에서 '계단 아래 시멘트 바닥'을 명시하고 있으므로, 외부 계단과 그 아래 시멘트 바닥 공간이 잘 드러나는 Candidate B가 적합합니다.", "kept": "L11B02" } }, "S13sh4": { "input_fingerprint": "354626bff6fa579b", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 수리영의 양어깨를 두 손으로 강하게 움켜쥔 채 굳은 표정으로 마주 보는 정면 구도.\n\nLOCATION (lock): The exterior of an old multi-unit villa, including the rooftop apartment, its descending access stairs, and the ground-level area where an emergency vehicle arrives. 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 courtyard, concrete ground, stair-base placement, and the stunned woman's clothing and standing position from the reference. Exclude the earlier solitary pose and add only the arriving woman gripping both shoulders.\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): Min-sook's corpse and the bloody, red-marked room remain upstairs while Hye-su confronts the stunned Su-ri-yeong outside.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "텍스트와 스토리보드에서 요구한 정면(오버 더 숄더) 구도를 정확히 구현했으며, 어깨를 쥔 동작과 복장 모두 훌륭하게 일치합니다." }, { "label": "B", "score": 4, "verdict_ko": "가장 중요한 프레이밍 지시인 '정면 구도'를 무시하고 측면(프로필) 구도로 연출하여 감점 요소가 큽니다." }, { "label": "C", "score": 3, "verdict_ko": "요구된 정면 구도를 위반한 측면 구도이며, 혜수의 핵심 의상인 모자까지 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S13sh4.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S13sh3_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "이전 컷에 있던 오른쪽 벽돌 건물 정면에 위치한 우측 상단으로 올라가는 큰 계단이 사라졌고, 대신 왼쪽 배경에 없던 계단이 잘못 생성되었습니다.", "fix_en": "Remove the stairs in the left background and restore the prominent concrete staircase ascending from left to right along the front of the right-hand brick building." }, { "issue_ko": "혜수가 수리영을 마주 보고 앞쪽에 서 있음에도 불구하고, 수리영의 어깨를 잡은 두 손은 마치 등 뒤에서 손을 얹은 것처럼 손등과 엄지 방향이 해부학적으로 반대로 그려졌습니다.", "fix_en": "Redraw Hye-su's hands to correctly show her gripping the shoulders from the front, with her fingers wrapping over the top and pointing down Su-ri-yeong's back." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the stairs in the left background and restore the prominent concrete staircase ascending from left to right along the front of the right-hand brick building.\n- Redraw Hye-su's hands to correctly show her gripping the shoulders from the front, with her fingers wrapping over the top and pointing down Su-ri-yeong's back.\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" }, "S13sh6": { "input_fingerprint": "65d7f960f72d8ac6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 두 눈을 크게 뜨고 입을 반쯤 벌린 채 충격받은 얼굴 클로즈업.\n\nLOCATION (lock): The exterior of an old multi-unit villa, including the rooftop apartment, its descending access stairs, and the ground-level area where an emergency vehicle arrives. 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 night courtyard lighting and exterior material palette from the reference. Exclude the stunned young woman, the full-body staging, and the stair-base action from this facial close-up; retain only the older woman's shocked expression and clothing.\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 last established upstairs state is still Min-sook's corpse seated behind the curtain amid blood and the red wall circle.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": "지시된 클로즈업 앵글과 충격받은 표정을 자연스럽게 연출했으며, 인물 설정과 배경 디테일도 훌륭하게 일치합니다." }, { "label": "B", "score": 5, "verdict_ko": "지시사항은 잘 반영했으나, 정면을 향한 시선과 다소 과장된 눈동자가 영화 스틸컷으로서의 자연스러움을 떨어뜨립니다." }, { "label": "C", "score": 3, "verdict_ko": "오른쪽 뺨에 파란색 발광 얼룩(마커/아티팩트 누출)이 남아 있어 심각한 오류에 해당합니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/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 레퍼런스에 있는 회색 모자가 없습니다.", "fix_en": "Add the grey cap from the character reference to her head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the grey cap from the character reference to her head.\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": "prev+엔티티", "lane_policy": "ab_select_bypass:prev" }, "S14sh5": { "input_fingerprint": "92977fc059a7185a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영의 얼굴이 창백하게 질린 채 커튼 쪽을 멍하니 응시하는 측면 클로즈업.\n\nLOCATION (lock): A compact rooftop apartment with a metal-door entry, a living and dining area, and a small bedroom. The bedroom is now unnaturally clean and orderly, with a bed, curtain, backpack, and doll but no visible blood, body, or red wall mark. 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 bedroom layout, curtain position, and dim nighttime look from the reference. Exclude the body, blood, footprints, wounds, and all other crime-state elements because the room is now unnaturally clean; do not carry over the earlier victim pose.\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-yeong's room is now excessively clean: Min-sook's body, every blood trace, and the red circle above the bed have vanished. The backpack and doll remain in the room but are slightly displaced.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 8, "A": 4, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 8, "verdict_ko": "지정된 수리영의 복장(후드티)과 방의 원래 구조를 유지하며, 커튼을 바라보는 측면 클로즈업을 정확하게 구현했습니다." }, { "label": "A", "score": 4, "verdict_ko": "이전 샷 인물의 복장을 잘못 가져왔으며, 커튼을 향한 시선 방향이 샷 텍스트와 불일치합니다." }, { "label": "B", "score": 3, "verdict_ko": "캐릭터 복장 오류가 발생했으며, 배경인 방의 좌우 공간 구조가 레퍼런스와 완전히 반대로 뒤집혔습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S12sh9_sel.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": "prev+엔티티" }, "S14sh7::ab_conti": { "input_fingerprint": "f1a5829aa4720e3f", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 김형사가 미간을 잔뜩 구긴 채 손가락으로 수리영을 가리키며 입을 비스듬히 벌린 정지 동작.\n\nLOCATION (lock): A compact rooftop apartment with a metal-door entry, a living and dining area, and a small bedroom. The bedroom is now unnaturally clean and orderly, with a bed, curtain, backpack, and doll but no visible blood, body, or red wall mark. 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 unnaturally clean, with the corpse, blood, red wall circle, and the two teacups absent; Su-ri-yeong's backpack and doll remain slightly shifted in her room.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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 표정(잔뜩 구긴 미간, 벌린 입)을 정확히 연출했으며, 지정된 인물(김형사)만 등장시키는 제한 규칙을 완벽히 준수함." }, { "label": "A", "score": 5, "verdict_ko": "허용된 인물만 등장시켰으나, 텍스트가 요구한 핵심 연기 지시인 '비스듬히 벌린 입'과 '잔뜩 구긴 미간'이 누락됨." }, { "label": "B", "score": 3, "verdict_ko": "프로필 목록에 허용되지 않은 인물(소녀)을 화면에 임의로 추가하여 'never anyone else' 제약을 심각하게 위반함." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B07.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S14sh7.png" }, { "label": "CHARACTER REFERENCE — 김형사: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 카메라 쪽을 향한 채 왼팔을 뻗고 있으며(왼팔에 오른손이 달려 있는 해부학적 오류 포함), 이는 카메라를 등지고 오른팔을 뻗은 레이아웃 스케치의 자세와 일치하지 않습니다.", "fix_en": "Rotate the character so he faces away from the camera and points with his right arm, matching the pose in the layout sketch and resolving the anatomical error." }, { "issue_ko": "인물의 의상이 캐릭터 레퍼런스의 의상과 다릅니다.", "fix_en": "Change the character's clothing to a brown leather jacket over a blue shirt and dark tie to match the character reference." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the character so he faces away from the camera and points with his right arm, matching the pose in the layout sketch and resolving the anatomical error.\n- Change the character's clothing to a brown leather jacket over a blue shirt and dark tie to match 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." }, "S14sh7::ab_noconti": { "input_fingerprint": "181718cfd962aef6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 김형사가 미간을 잔뜩 구긴 채 손가락으로 수리영을 가리키며 입을 비스듬히 벌린 정지 동작.\n\nLOCATION (lock): A compact rooftop apartment with a metal-door entry, a living and dining area, and a small bedroom. The bedroom is now unnaturally clean and orderly, with a bed, curtain, backpack, and doll but no visible blood, body, or red wall mark. 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 unnaturally clean, with the corpse, blood, red wall circle, and the two teacups absent; Su-ri-yeong's backpack and doll remain slightly shifted in her room.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": 4, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 6, "verdict_ko": "지시된 표정과 가리키는 동작을 해부학적 오류 없이 잘 구현했으나, 침대 위치 등 공간 배치가 달라졌고 지정되지 않은 인물(수리영)의 뒷모습이 등장함." }, { "label": "B", "score": 4, "verdict_ko": "배경 재현도는 가장 높으나, 가리키는 팔이 가로지르는 각도와 손의 방향이 물리적으로 불가능한 해부학적 오류(하드 위반)가 있음." }, { "label": "C", "score": 3, "verdict_ko": "가리키는 팔이 가슴 중앙에서 뻗어 나오는 심각한 해부학적 오류(하드 위반)가 있으며, 배경의 욕실 구조도 잘못됨." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B07.png" }, { "label": "CHARACTER REFERENCE — 김형사: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "프롬프트의 인물 목록에 없는 다른 사람(화면 오른쪽 여성의 뒷모습)이 프레임에 포함되어 있습니다.", "fix_en": "Remove the woman on the right edge of the frame completely, replacing her with the apartment background." }, { "issue_ko": "김형사가 캐릭터 레퍼런스에 있는 회색 모자(플랫 캡)를 착용하고 있지 않습니다.", "fix_en": "Add the grey flat cap from the character reference onto Detective Kim's head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the woman on the right edge of the frame completely, replacing her with the apartment background.\n- Add the grey flat cap from the character reference onto Detective Kim's head.\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." }, "S14sh7::conti_ab_decision": { "fingerprint": "287470ffd9a8d325", "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 지시문에 묘사된 김형사의 찡그린 미간과 비스듬히 벌린 입을 정확한 프레이밍으로 보여주지만, 후보 1은 뒷모습으로 앵글을 잡아 핵심적인 표정 연기를 전혀 구현하지 못했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "1번 후보는 찡그린 미간과 벌어진 입 등 지문에서 요구한 김형사의 표정을 정확히 묘사한 반면, 2번 후보는 뒷모습으로 연출되어 표정을 전혀 확인할 수 없습니다." } ] } }, "S14sh7": { "input_fingerprint": "181718cfd962aef6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 김형사가 미간을 잔뜩 구긴 채 손가락으로 수리영을 가리키며 입을 비스듬히 벌린 정지 동작.\n\nLOCATION (lock): A compact rooftop apartment with a metal-door entry, a living and dining area, and a small bedroom. The bedroom is now unnaturally clean and orderly, with a bed, curtain, backpack, and doll but no visible blood, body, or red wall mark. 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 unnaturally clean, with the corpse, blood, red wall circle, and the two teacups absent; Su-ri-yeong's backpack and doll remain slightly shifted in her room.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": 4, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 6, "verdict_ko": "지시된 표정과 가리키는 동작을 해부학적 오류 없이 잘 구현했으나, 침대 위치 등 공간 배치가 달라졌고 지정되지 않은 인물(수리영)의 뒷모습이 등장함." }, { "label": "B", "score": 4, "verdict_ko": "배경 재현도는 가장 높으나, 가리키는 팔이 가로지르는 각도와 손의 방향이 물리적으로 불가능한 해부학적 오류(하드 위반)가 있음." }, { "label": "C", "score": 3, "verdict_ko": "가리키는 팔이 가슴 중앙에서 뻗어 나오는 심각한 해부학적 오류(하드 위반)가 있으며, 배경의 욕실 구조도 잘못됨." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B07.png" }, { "label": "CHARACTER REFERENCE — 김형사: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "프롬프트의 인물 목록에 없는 다른 사람(화면 오른쪽 여성의 뒷모습)이 프레임에 포함되어 있습니다.", "fix_en": "Remove the woman on the right edge of the frame completely, replacing her with the apartment background." }, { "issue_ko": "김형사가 캐릭터 레퍼런스에 있는 회색 모자(플랫 캡)를 착용하고 있지 않습니다.", "fix_en": "Add the grey flat cap from the character reference onto Detective Kim's head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the woman on the right edge of the frame completely, replacing her with the apartment background.\n- Add the grey flat cap from the character reference onto Detective Kim's head.\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.", "conti_ab": { "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 지시문에 묘사된 김형사의 찡그린 미간과 비스듬히 벌린 입을 정확한 프레이밍으로 보여주지만, 후보 1은 뒷모습으로 앵글을 잡아 핵심적인 표정 연기를 전혀 구현하지 못했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "1번 후보는 찡그린 미간과 벌어진 입 등 지문에서 요구한 김형사의 표정을 정확히 묘사한 반면, 2번 후보는 뒷모습으로 연출되어 표정을 전혀 확인할 수 없습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S14sh7__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S14sh7__ab_noconti_sel.png" }, "ref_mode": "플레이트+엔티티 (A/B: 무콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B07", "choice": "Candidate G", "confident": true, "reason_ko": "김형사가 수리영을 가리키는 장면으로, 혈흔이나 붉은 자국 없이 깨끗하게 치워진 침실 배경이 필요하므로 현재 지정된 G가 가장 적합합니다.", "kept": "L04B07" } }, "S15sh2": { "input_fingerprint": "acbb9d76c89692d0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수의 손에 들린 차가운 캔 음료가 수리영의 무릎 위 허공에 멈춰 있는 클로즈업.\n\nLOCATION (lock): A bus stop area along the coast, with seating positioned toward the sea and an adjacent roadside space. 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: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": 8, "A": 5, "C": 3 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 8, "verdict_ko": "프레이밍, 두 인물의 지정된 의상, 차가운 캔의 디테일 및 배경 요소를 스토리보드와 프롬프트에 맞춰 가장 충실하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "구도와 상황은 적절하나, 캔을 든 혜수의 팔이 지정된 검은색 재킷이 아닌 잘못된 색상의 옷으로 표현되었습니다." }, { "label": "C", "score": 3, "verdict_ko": "화면 배경에 프롬프트와 숏 텍스트에 명시되지 않은 정체불명의 인물이 등장하여 치명적인 위반(Hard Violation)입니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S15sh2.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캔 음료 표면에 영어와 정체불명의 문자가 생성되었습니다 ('No text' 지시 사항 위반).", "fix_en": "Remove all text and logos from the surface of the beverage can." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text and logos from the surface of the beverage 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": "6f151683862e645b", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, exterior light unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 한 손을 입가에 모으고 수리영의 뒷모습을 향해 입을 크게 벌린 정지 상태.\n\nLOCATION (lock): A bus stop area along the coast, with seating positioned toward the sea and an adjacent roadside space. 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: 수리영 (한국인 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": "구도와 동작은 훌륭하나, 혜수의 의상 색상이 레퍼런스와 다르고 스케치의 가방이 잘못 반영되었습니다." }, { "label": "B", "score": 7, "verdict_ko": "스케치의 가방을 배제하고 혜수의 검은색 의상을 유지하여 캐릭터 레퍼런스를 가장 충실히 따랐습니다." }, { "label": "C", "score": 4, "verdict_ko": "스토리보드 스케치의 가방 요소가 그대로 노출되었으며 전경 인물의 복장 디테일이 레퍼런스와 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S15sh5.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "전경에 있는 혜수가 캐릭터 레퍼런스에 있는 검은색 모자를 쓰고 있지 않습니다.", "fix_en": "Add the black baseball cap from the character reference to the foreground woman's head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the black baseball cap from the character reference to the foreground woman's head.\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" }, "S16sh3::ab_conti": { "input_fingerprint": "9c19a02268bdd2de", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 김형사의 얼굴 코앞에 스마트폰 액정 화면을 들이밀고 있는 손 클로즈업.\n\nLOCATION (lock): A police office with desks, computers, telephones, and workstations for investigators. Only one officer is on overnight duty when another investigator enters. 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): Hye-su continues to carry the phone containing Min-sook's saved number.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 머리); 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": "레이아웃 스케치의 좌우 인물 배치를 위반했으며, 스마트폰을 쥔 손가락의 해부학적 형태가 부자연스럽습니다." }, { "label": "B", "score": 7, "verdict_ko": "레이아웃 스케치의 구도와 인물 배치를 정확히 따랐으며, 김형사 얼굴 앞에 폰을 들이미는 샷 텍스트의 요구를 충실히 구현했습니다." }, { "label": "C", "score": 2, "verdict_ko": "김형사의 모습이 아예 배제되어 샷 텍스트의 핵심 연출을 위반했으며, 화면에 지시되지 않은 영문 텍스트가 나타났습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S16sh3.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 김형사: 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": "스마트폰 액정 화면이 김형사가 아닌 카메라 쪽을 향하고 있습니다.", "fix_en": "Rotate the smartphone and the hand holding it so the screen faces Detective Kim." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone and the hand holding it so the screen faces Detective Kim.\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." }, "S16sh3::ab_noconti": { "input_fingerprint": "27f6e10a721d7f61", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 김형사의 얼굴 코앞에 스마트폰 액정 화면을 들이밀고 있는 손 클로즈업.\n\nLOCATION (lock): A police office with desks, computers, telephones, and workstations for investigators. Only one officer is on overnight duty when another investigator enters. 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): Hye-su continues to carry the phone containing Min-sook's saved number.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 머리); 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": 2, "C": 4 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지시된 '손 클로즈업'과 소품 디테일을 잘 구현했으나, 액정 화면이 김형사가 아닌 카메라를 향하고 인물의 시선이 어긋난 점이 아쉽습니다." }, { "label": "B", "score": 2, "verdict_ko": "지시문에 없는 제3의 인물이 배경 좌측에 등장하여 '인물 추가 금지' 규칙을 위반한 치명적 오류(Hard Violation)입니다." }, { "label": "C", "score": 4, "verdict_ko": "혜수가 들고 있어야 할 스마트폰을 김형사의 가죽 재킷을 입은 팔이 쥐고 있어 행동 지시와 의상을 위반했으며, 화면 속 번호에도 오타가 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 김형사: 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": "스마트폰 화면이 김형사의 시선이 아닌 카메라 쪽을 향하고 있어, 그에게 화면을 보여준다는 설정과 어긋납니다.", "fix_en": "Rotate the smartphone so its screen faces directly toward Detective Kim's eyes." }, { "issue_ko": "혜수의 집게손가락이 김형사의 코를 찌르고 있으며, 손가락들이 화면 앞부분을 가리고 있어 부자연스럽습니다.", "fix_en": "Move the hand slightly away from his face so no fingers touch his nose, and reposition the fingers to hold only the edges and back of the phone." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone so its screen faces directly toward Detective Kim's eyes.\n- Move the hand slightly away from his face so no fingers touch his nose, and reposition the fingers to hold only the edges and 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." }, "S16sh3::conti_ab_decision": { "fingerprint": "7b0975f81f1e468a", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 스마트폰 화면이 김형사를 향해 올바르게 배치된 반면, 후보 2는 화면이 카메라를 향해 있어 김형사가 폰의 뒷면을 보게 되는 물리적 오류가 있습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 스마트폰 화면이 김형사의 시선을 향하도록 물리적 방향을 올바르게 연출한 반면, 후보 1은 화면을 카메라 쪽으로 부자연스럽게 노출했습니다." } ] } }, "S16sh3": { "input_fingerprint": "9c19a02268bdd2de", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 김형사의 얼굴 코앞에 스마트폰 액정 화면을 들이밀고 있는 손 클로즈업.\n\nLOCATION (lock): A police office with desks, computers, telephones, and workstations for investigators. Only one officer is on overnight duty when another investigator enters. 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): Hye-su continues to carry the phone containing Min-sook's saved number.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 머리); 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": "레이아웃 스케치의 좌우 인물 배치를 위반했으며, 스마트폰을 쥔 손가락의 해부학적 형태가 부자연스럽습니다." }, { "label": "B", "score": 7, "verdict_ko": "레이아웃 스케치의 구도와 인물 배치를 정확히 따랐으며, 김형사 얼굴 앞에 폰을 들이미는 샷 텍스트의 요구를 충실히 구현했습니다." }, { "label": "C", "score": 2, "verdict_ko": "김형사의 모습이 아예 배제되어 샷 텍스트의 핵심 연출을 위반했으며, 화면에 지시되지 않은 영문 텍스트가 나타났습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S16sh3.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 김형사: 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": "스마트폰 액정 화면이 김형사가 아닌 카메라 쪽을 향하고 있습니다.", "fix_en": "Rotate the smartphone and the hand holding it so the screen faces Detective Kim." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone and the hand holding it so the screen faces Detective Kim.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 스마트폰 화면이 김형사를 향해 올바르게 배치된 반면, 후보 2는 화면이 카메라를 향해 있어 김형사가 폰의 뒷면을 보게 되는 물리적 오류가 있습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 스마트폰 화면이 김형사의 시선을 향하도록 물리적 방향을 올바르게 연출한 반면, 후보 1은 화면을 카메라 쪽으로 부자연스럽게 노출했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S16sh3__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S16sh3__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L14B01", "B": "L14B03", "C": "L14B04", "D": "L14B05" }, "assigned": "L14B04", "choice": "Candidate C", "confident": true, "reason_ko": "경찰서 내부에서 김형사와 혜수가 대화하는 상황이며, 후보 A, B, C가 모두 동일한 사무실 내부를 보여주므로 현재 할당된 후보 C를 유지합니다.", "kept": "L14B04" } }, "S16sh5::ab_conti": { "input_fingerprint": "126fb381dd78dc1d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 두 손으로 책상 모서리를 짚은 채 고개를 숙이고 심각하게 인상을 쓴 상체 구도.\n\nLOCATION (lock): A police office with desks, computers, telephones, and workstations for investigators. Only one officer is on overnight duty when another investigator enters. 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): Hye-su retains the smartphone she has been using to call Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": 6 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지시된 상체 구도와 표정을 정확히 구현했으며, 불필요한 인물이 없어 가장 우수합니다." }, { "label": "B", "score": 3, "verdict_ko": "배경에 프롬프트에 명시되지 않은 인물이 추가되어 심각한 규정 위반입니다." }, { "label": "C", "score": 6, "verdict_ko": "전반적으로 양호하나 오른손을 주먹 쥐고 있어 '책상을 짚은 자세'의 구현이 A보다 다소 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S16sh5.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스마트폰을 짚고 있는 캐릭터의 왼손(화면 우측) 손가락이 기형적으로 묘사되어 해부학적으로 틀렸습니다.", "fix_en": "Correct the anatomy of the character's left hand (on the right side of the image) so it has five normal fingers resting naturally on the phone and desk." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Correct the anatomy of the character's left hand (on the right side of the image) so it has five normal fingers resting naturally on the phone and 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." }, "S16sh5::ab_noconti": { "input_fingerprint": "f94c508d99b3a234", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 두 손으로 책상 모서리를 짚은 채 고개를 숙이고 심각하게 인상을 쓴 상체 구도.\n\nLOCATION (lock): A police office with desks, computers, telephones, and workstations for investigators. Only one officer is on overnight duty when another investigator enters. 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): Hye-su retains the smartphone she has been using to call Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": "지정된 상체 구도와 표정을 잘 살렸고, 필수 소지품(스마트폰)과 참조 이미지의 모자를 유일하게 유지했습니다." }, { "label": "B", "score": 5, "verdict_ko": "책상을 짚은 자세는 적절하나, 지시된 스마트폰과 참조 이미지의 모자가 모두 누락되었습니다." }, { "label": "C", "score": 4, "verdict_ko": "모자가 누락되었고, 스마트폰이 좁은 파티션 위에 비현실적으로 아슬아슬하게 놓여 있어 물리적 연출 오류가 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 책상 모서리가 아닌 파티션(칸막이) 상단을 짚고 있습니다.", "fix_en": "Lower the character's hands and arms so they are resting directly on the horizontal edge of the desk." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Lower the character's hands and arms so they are resting directly on the horizontal edge of 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." }, "S16sh5::conti_ab_decision": { "fingerprint": "88fdccaa8a5f65c2", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "1번은 프롬프트에 명시된 대로 파티션이 아닌 책상 모서리를 두 손으로 짚고 있는 모습을 정확히 연출했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 인물이 파티션이 아닌 책상 모서리를 짚고 있는 모습을 정확하게 묘사하여 프롬프트의 지시를 더 충실히 이행했습니다." } ] } }, "S16sh5": { "input_fingerprint": "126fb381dd78dc1d", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): night, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 혜수가 두 손으로 책상 모서리를 짚은 채 고개를 숙이고 심각하게 인상을 쓴 상체 구도.\n\nLOCATION (lock): A police office with desks, computers, telephones, and workstations for investigators. Only one officer is on overnight duty when another investigator enters. 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): Hye-su retains the smartphone she has been using to call Min-sook.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": 6 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지시된 상체 구도와 표정을 정확히 구현했으며, 불필요한 인물이 없어 가장 우수합니다." }, { "label": "B", "score": 3, "verdict_ko": "배경에 프롬프트에 명시되지 않은 인물이 추가되어 심각한 규정 위반입니다." }, { "label": "C", "score": 6, "verdict_ko": "전반적으로 양호하나 오른손을 주먹 쥐고 있어 '책상을 짚은 자세'의 구현이 A보다 다소 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S16sh5.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스마트폰을 짚고 있는 캐릭터의 왼손(화면 우측) 손가락이 기형적으로 묘사되어 해부학적으로 틀렸습니다.", "fix_en": "Correct the anatomy of the character's left hand (on the right side of the image) so it has five normal fingers resting naturally on the phone and desk." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Correct the anatomy of the character's left hand (on the right side of the image) so it has five normal fingers resting naturally on the phone and 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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "1번은 프롬프트에 명시된 대로 파티션이 아닌 책상 모서리를 두 손으로 짚고 있는 모습을 정확히 연출했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 인물이 파티션이 아닌 책상 모서리를 짚고 있는 모습을 정확하게 묘사하여 프롬프트의 지시를 더 충실히 이행했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S16sh5__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S16sh5__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L14B01", "B": "L14B03", "C": "L14B04", "D": "L14B05" }, "assigned": "L14B04", "choice": "Candidate C", "confident": true, "reason_ko": "지문에서 '책상 모서리를 짚은 채'라고 언급되어 있으므로 책상이 있는 실내 사무실 공간이어야 합니다. 후보 A, B, C가 모두 동일한 실내를 보여주므로 현재 할당된 후보 C를 유지합니다.", "kept": "L14B04" } }, "S17sh1": { "input_fingerprint": "35d7e17c5abb31ad", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dawn, rooftop room lights on.\n\nSHOT TEXT (authoritative, Korean): 어스름한 새벽빛이 스며드는 낡은 옥탑방 마당 담벼락 위에 검은 고양이 한 마리가 웅크리고 앉아 불 켜진 창문을 향해 고개를 돌린 정지 구도.\n\nLOCATION (lock): The rooftop exterior of an old residential villa, enclosed by a low wall and attached to a small rooftop apartment. The apartment remains visibly lit as a cat crosses the wall. The shot takes place here — the attached LOCATION PHOTOGRAPH shows the exact spot.\n\nSTRUCTURE LOOK AUTHORITY: the attached STRUCTURE LOOK photograph is the identity of the fixed structure at this location — wherever that structure appears in the frame, its shape, proportions, openings, materials and colors are LOCKED to it. The LOCATION PHOTOGRAPH remains the authority for this shot's sub-space, surroundings, time of day and lighting. If the two conflict on the structure itself, the STRUCTURE LOOK photo wins; for everything else, the LOCATION PHOTOGRAPH wins.\n\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 lit 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": { "A": 6, "B": 4, "C": 2 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 6, "verdict_ko": "지정된 구조물 레퍼런스(회색 콘크리트 외관)를 정확히 반영하고 고양이가 창문을 향해 있으나, 금지된 레퍼런스의 구도와 주변 환경까지 그대로 모방한 점이 감점 요인입니다." }, { "label": "B", "score": 4, "verdict_ko": "고양이가 웅크리고 창문을 바라보는 모습은 잘 표현되었으나, 가장 중요한 건물 외관이 구조물 레퍼런스를 무시하고 벽돌로 렌더링되어 우선순위에서 밀립니다." }, { "label": "C", "score": 2, "verdict_ko": "구조물 레퍼런스를 완전히 무시한 채 위치 사진의 외관과 구도를 그대로 복사했으며, 고양이가 창문이 아닌 정면을 응시하여 프롬프트를 크게 위반했습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its spatial layout, surroundings, fixed features, time of day and lighting mood are spatial truth; stage the moment inside this place. If a STRUCTURE LOOK photograph is also attached, that photo wins for the fixed structure itself — this photograph wins for everything around it. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L11B01.png" }, { "label": "STRUCTURE LOOK — the confirmed photograph of the fixed structure at this location: wherever the structure appears in the frame, its shape, proportions, materials, colors and openings are LOCKED to this photo. Never copy its camera framing, time of day or lighting — the shot text and the LOCATION PHOTOGRAPH are the authorities for those.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/background_chain/seed_bg_rooftop_villa_sel.png" } ], "critique": { "issues": [ { "issue_ko": "고양이가 옥탑방 마당 담벼락(옥상 난간)이 아닌 1층 대문 지붕 위에 위치해 있습니다.", "fix_en": "Move the black cat from the ground-floor gate roof to crouch on the low parapet wall enclosing the rooftop." }, { "issue_ko": "고양이가 불 켜진 창문을 향해 고개를 돌리지 않고 건물 벽면을 바라보고 있습니다.", "fix_en": "Turn the cat's head so it is visibly looking up at the lit window of the rooftop room." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Move the black cat from the ground-floor gate roof to crouch on the low parapet wall enclosing the rooftop.\n- Turn the cat's head so it is visibly looking up at the lit window of the rooftop room.\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+seed만 (배경 전용)", "lane_policy": "ab_select_bypass:bg_only", "plate_select": { "candidates": { "A": "L11B01", "B": "L11B02" }, "assigned": "L11B01", "choice": "A", "confident": true, "reason_ko": "샷 텍스트에 묘사된 '낡은 옥탑방 마당 담벼락 위 검은 고양이', '불 켜진 창문', '어스름한 새벽빛'이 모두 후보 A에 정확히 일치합니다.", "chosen": "L11B01" } }, "S18sh1::ab_conti": { "input_fingerprint": "391b3082cbba5711", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 동이 튼 방 안, 수리영이 침대 매트리스에 이마를 기댄 채 눈을 감고 웅크린 상체 구도.\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-dining area. No specific interior light source is stated. 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 lit and Su-ri-yeong's room remains excessively clean, with no corpse or blood behind the curtain. Her doll-trimmed backpack is still in the room until she takes it out later in this 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "요구된 상체 구도를 정확히 따랐으며, 이마를 매트리스에 기댄 자세를 가장 사실적으로 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "상체 구도 지시를 무시하고 화각을 넓혔으며, 이마가 아닌 뺨을 기대고 있어 감점되었습니다." }, { "label": "C", "score": 5, "verdict_ko": "프레이밍은 적절하나, 이마가 매트리스에서 떨어져 있고 얼굴 하단부가 닿아 있어 포즈 정확도가 아쉽습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B05.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S18sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영이 매트리스에 이마가 아닌 코와 입 부근을 기대고 있습니다.", "fix_en": "Tilt the character's head down so her forehead is firmly resting on the mattress." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Tilt the character's head down so her forehead is firmly resting on the mattress.\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." }, "S18sh1::ab_noconti": { "input_fingerprint": "e9eb6e318a7d3637", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 동이 튼 방 안, 수리영이 침대 매트리스에 이마를 기댄 채 눈을 감고 웅크린 상체 구도.\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-dining area. No specific interior light source is stated. 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 lit and Su-ri-yeong's room remains excessively clean, with no corpse or blood behind the curtain. Her doll-trimmed backpack is still in the room until she takes it out later in this 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "요구된 상체 구도를 정확히 따랐으며, 이마를 침대 매트리스에 직접 기댄 핵심 동작을 가장 충실하게 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "상체 구도를 벗어나 전신이 침대 위에 올라가 있으며, 이마를 매트리스에 기댄 자세가 명확히 표현되지 않았습니다." }, { "label": "C", "score": 5, "verdict_ko": "구도와 배경은 좋으나 이마를 매트리스가 아닌 자신의 팔에 기대고 있어 세부 동작 묘사에서 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B05.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물이 캐릭터 레퍼런스의 올리브색 후드티가 아닌 위치 레퍼런스의 갈색 의상을 그대로 입고 있습니다.", "fix_en": "Change the character's clothing to the olive green hoodie 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- Change the character's clothing to the olive green hoodie 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." }, "S18sh1::conti_ab_decision": { "fingerprint": "f294936b7f811e87", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 프롬프트에 명시된 '인형이 달린 백팩'을 배경에 정확히 묘사했으며, 매트리스에 이마를 기댄 자연스러운 자세를 잘 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "2번이 이마를 매트리스에 완전히 기댄 자세와 샷 텍스트에 요구된 인형 장식이 달린 백팩을 더 정확하게 구현함." } ] } }, "S18sh1": { "input_fingerprint": "391b3082cbba5711", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 동이 튼 방 안, 수리영이 침대 매트리스에 이마를 기댄 채 눈을 감고 웅크린 상체 구도.\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-dining area. No specific interior light source is stated. 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 lit and Su-ri-yeong's room remains excessively clean, with no corpse or blood behind the curtain. Her doll-trimmed backpack is still in the room until she takes it out later in this 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 4, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "요구된 상체 구도를 정확히 따랐으며, 이마를 매트리스에 기댄 자세를 가장 사실적으로 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "상체 구도 지시를 무시하고 화각을 넓혔으며, 이마가 아닌 뺨을 기대고 있어 감점되었습니다." }, { "label": "C", "score": 5, "verdict_ko": "프레이밍은 적절하나, 이마가 매트리스에서 떨어져 있고 얼굴 하단부가 닿아 있어 포즈 정확도가 아쉽습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B05.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S18sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영이 매트리스에 이마가 아닌 코와 입 부근을 기대고 있습니다.", "fix_en": "Tilt the character's head down so her forehead is firmly resting on the mattress." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Tilt the character's head down so her forehead is firmly resting on the mattress.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 프롬프트에 명시된 '인형이 달린 백팩'을 배경에 정확히 묘사했으며, 매트리스에 이마를 기댄 자연스러운 자세를 잘 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "2번이 이마를 매트리스에 완전히 기댄 자세와 샷 텍스트에 요구된 인형 장식이 달린 백팩을 더 정확하게 구현함." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S18sh1__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S18sh1__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B05", "choice": "Candidate E", "confident": true, "reason_ko": "\"침대 매트리스에 이마를 기댄 채 눈을 감고 웅크린\"이라는 지문에 정확히 부합하는 침대 공간이므로 현재 할당된 후보를 유지합니다.", "kept": "L04B05" } }, "S18sh9": { "input_fingerprint": "fbdd52fd411f343f", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 거울에 비친 욕실 타일 벽면에 붉은색 원 형태의 뚜렷한 표식이 새겨져 있는 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-dining area. No specific interior light source is stated. 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): Su-ri-yeong now has her doll-trimmed backpack with her. The red circle exists only as a momentary apparition on the reflected bathroom wall.\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": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "실제 벽면은 비워두고 거울 속 반사된 타일 벽면에만 붉은 원이 나타나도록 한 핵심 제약 조건을 가장 정확하게 연출했습니다." }, { "label": "B", "score": 3, "verdict_ko": "거울 속뿐만 아니라 물리적인 실제 벽면에도 붉은 원이 존재하여 프롬프트의 지시사항을 정면으로 위반했습니다." }, { "label": "C", "score": 5, "verdict_ko": "거울 속에 나타난 붉은 원을 보여주지만, 실제 벽면과의 대조가 프레임에 담기지 않아 지시사항의 의도가 명확히 드러나지 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B04.png" } ], "critique": { "issues": [ { "issue_ko": "거울 오른쪽 벽면에 레퍼런스 사진에는 없는 창문이 임의로 생성되었습니다.", "fix_en": "Remove the window on the right wall and replace it with solid white bathroom tiles matching the rest of the room." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the window on the right wall and replace it with solid white bathroom tiles matching the rest of the room.\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)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B04", "choice": "Candidate D", "confident": true, "reason_ko": "지문에서 명시한 '거울', '욕실 타일 벽면', '붉은색 원 형태의 표식'이 모두 등장하는 유일한 컷입니다.", "kept": "L04B04" } }, "S18sh11": { "input_fingerprint": "adb6d5deb0fefbe8", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 거실 식탁에 놓인 액자 속 강민숙의 활짝 웃는 얼굴 형상이 마치 진흙처럼 아래로 기괴하게 흘러내리는 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-dining area. No specific interior light source is stated. 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): Su-ri-yeong continues to carry her doll-trimmed backpack; the framed photograph itself remains normal after the fleeting distortion.\n\nNO PEOPLE IN THIS SHOT: the shot text shows only the place and its state — no living person or any body part appears in frame, unless the shot text itself explicitly says so.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 8, "C": 5, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 8, "verdict_ko": "레퍼런스의 카메라 구도를 피하라는 지침을 잘 따랐으며, 피사체에 집중하여 웃는 여성이 진흙처럼 기괴하게 흘러내리는 묘사를 가장 정확하게 구현했습니다." }, { "label": "C", "score": 5, "verdict_ko": "지문대로 한 여성의 웃는 얼굴이 흘러내리도록 연출했으나, 위치 레퍼런스의 카메라 구도를 그대로 복사하지 말라는 지침을 어겼습니다." }, { "label": "A", "score": 3, "verdict_ko": "레퍼런스의 카메라 구도를 그대로 복사했을 뿐만 아니라, 지문이 요구한 단일 인물이 아닌 세 명의 인물을 묘사하여 텍스트에 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B04.png" } ], "critique": { "issues": [ { "issue_ko": "사진 속 세 인물이 모두 흘러내리고 있으나, 프롬프트는 '강민숙의 얼굴 형상' 하나만 흘러내려야 한다고 명시하고 있습니다.", "fix_en": "Apply the melting effect to only one face in the photograph, keeping the other two figures completely normal." }, { "issue_ko": "진흙이 탁자, 의자, 바닥까지 방대하게 쏟아져 내리고 있습니다. 얼굴 형상이 아래로 흘러내리는 묘사로 제한해야 합니다.", "fix_en": "Confine the melting mud effect strictly to the photograph and the immediate area under the frame, removing all mud from the chairs, table edge, and floor." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Apply the melting effect to only one face in the photograph, keeping the other two figures completely normal.\n- Confine the melting mud effect strictly to the photograph and the immediate area under the frame, removing all mud from the chairs, table edge, and 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)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B04", "choice": "Candidate D", "confident": true, "reason_ko": "\"거실 식탁에 놓인 액자\"와 그 안의 형상이 \"진흙처럼 아래로 기괴하게 흘러내리는 찰나\"라는 묘사가 Candidate D의 식탁 위 녹아내리는 액자 모습과 정확히 일치합니다.", "kept": "L04B04" } }, "S19sh6::ab_conti": { "input_fingerprint": "faf27728a6bf367b", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 입술을 굳게 다물고 마트 천장 모서리를 가리키는 손끝 클로즈업.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong still carries her doll-trimmed backpack and has arrived by bicycle.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 5, "B": 7, "C": 4 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "A", "score": 5, "verdict_ko": "의상과 소품은 레퍼런스와 일치하나, 굳게 다문 입술(지시문) 대신 입술을 내민 표정을 연출해 감점되었습니다." }, { "label": "B", "score": 7, "verdict_ko": "입술을 굳게 다문 표정을 가장 정확히 묘사했으며, 지정된 의상과 가방 색상도 충실히 반영했습니다." }, { "label": "C", "score": 4, "verdict_ko": "표정 지시를 어기고 입술을 내밀었으며, 가방의 색상(흰색)이 지정된 레퍼런스와 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B02.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S19sh6.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지시문에 어떤 텍스트도 포함하지 말 것(No text)을 명시했으나, 배경의 간판과 계산대에 '알정선잡', 'Utility Zone', '서비스 조몬 4' 등의 글자가 생성되었습니다.", "fix_en": "Remove all text from the overhead signs and the counter in the background, leaving the surfaces completely blank." } ] } }, "S19sh8": { "input_fingerprint": "501b5050ea843b03", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 마트 외부, 천장 구석에 매달린 둥근 CCTV 카메라 렌즈 클로즈업.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong's doll-trimmed backpack remains in her possession while she and the employee inspect the exterior CCTV position.\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": 6, "B": 4, "C": 7 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "A", "score": 6, "verdict_ko": "클로즈업 프레이밍이 다소 넓으며, 지시문과 달리 천장이 아닌 벽면에 설치되어 있습니다." }, { "label": "B", "score": 4, "verdict_ko": "카메라 렌즈의 구조가 물리적으로 부자연스럽고 기형적으로 표현되었습니다." }, { "label": "C", "score": 7, "verdict_ko": "천장에 매달린 둥근 CCTV 렌즈의 클로즈업을 정확히 구현했고, 배경의 마트 외벽과 거리도 잘 일치합니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B03.png" } ], "critique": { "issues": [ { "issue_ko": "CCTV 카메라의 돔 유리에 카메라를 들고 있는 사람의 모습이 반사되어 있어 인물이 등장해서는 안 된다는 지시사항을 위반했습니다.", "fix_en": "Remove the reflection of the person holding a camera from the glass dome of the CCTV camera." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the reflection of the person holding a camera from the glass dome of the CCTV camera.\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)", "plate_select": { "candidates": { "A": "L09B02", "B": "L09B03" }, "assigned": "L09B03", "choice": "Candidate B", "confident": true, "reason_ko": "지문에서 '마트 외부'라고 명시하고 있으므로, 마트 안쪽을 보여주는 A 대신 외경을 보여주는 B가 올바른 공간입니다.", "kept": "L09B03" } }, "S19sh10::ab_conti": { "input_fingerprint": "0356b6bae3eb9126", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 스마트폰을 오른쪽 귀에 바짝 댄 채 입을 벌린 상체 구도.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong retains her backpack and smartphone, using the phone to explain the CCTV lead to Hye-su.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "오른쪽 귀에 올바른 방향으로 폰을 대고 있으며, 추가 인물 없이 상체 구도와 배경을 정확하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "구도와 배경은 우수하나, 스마트폰 화면이 바깥을 향하게 쥐고 있어 소품 방향 규칙을 위반했습니다." }, { "label": "B", "score": 3, "verdict_ko": "지시문에 없는 인물이 등장했고 왼쪽 귀에 폰을 대고 있으며, 배경에 콜라주 형태의 공간적 오류가 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B03.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S19sh10.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": "배경의 간판, 상점 창문, 파란색 플라스틱 상자 등에 텍스트가 생성되어 '어느 곳에도 텍스트 금지' 규칙을 위반했습니다.", "fix_en": "Remove all text and letters from the street signs, store windows, and blue crates in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text and letters from the street signs, store windows, and blue crates 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." }, "S20sh3::ab_conti": { "input_fingerprint": "419b5c5d006814d6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 혜수와 한국인 남자 보안직원 뒤에 서서 두 손을 모아 꽉 쥔 채 굳은 표정으로 모니터를 응시하는 상체.\n\nLOCATION (lock): A supermarket administrative and security office centered on a CCTV monitoring station, with screens and playback equipment for reviewing recorded 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): Su-ri-yeong continues to have her doll-trimmed backpack and smartphone with her while viewing the footage.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 머리); 한국인 남자 보안직원 (한국인 성인 남성, 자연스러운 얼굴, 단정한 머리, 짙은색 짧은 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": 9, "B": 3, "C": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 9, "verdict_ko": "수리영이 뒤에 서서 두 손을 꽉 쥐고 모니터를 응시하는 상체 샷의 구도, 굳은 표정, 그리고 백팩과 스마트폰 소품까지 프롬프트를 매우 충실하게 구현했습니다." }, { "label": "B", "score": 3, "verdict_ko": "수리영의 기준 의상(올리브색 후드티)을 입은 인물이 앉아 있고, 서 있는 인물은 프롬프트에 없는 교복을 입고 있어 캐릭터의 신원과 배치가 완전히 어긋납니다." }, { "label": "C", "score": 5, "verdict_ko": "수리영이 뒤에 서 있는 배치는 맞으나, '두 손을 모아 꽉 쥔' 핵심 포즈가 누락되었으며 얼굴이 잘 보이지 않아 굳은 표정을 살리지 못했습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L15B03.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S20sh3.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "레이아웃 스케치에 따르면 혜수와 보안직원은 의자에 앉아 있어야 하지만, 생성된 이미지에서는 두 사람 모두 서서 책상에 기대어 있습니다.", "fix_en": "Seat Hye-su and the security guard in the office chairs, exactly matching their sitting poses from the layout sketch." }, { "issue_ko": "텍스트 생성이 금지되어 있으나, 모니터 화면들에 'CAM 3'를 비롯한 영어와 숫자 텍스트가 나타나 있습니다.", "fix_en": "Remove all text and numbers from the monitor screens." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Seat Hye-su and the security guard in the office chairs, exactly matching their sitting poses from the layout sketch.\n- Remove all text and numbers from the monitor screens.\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." }, "S20sh5": { "input_fingerprint": "0a303ad9e06a3ca2", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 어깨를 움츠린 채 아랫입술을 질끈 깨문 불안한 얼굴 클로즈업.\n\nLOCATION (lock): A supermarket administrative and security office centered on a CCTV monitoring station, with screens and playback equipment for reviewing recorded footage. 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 security-room lighting, monitor glow, muted office materials, and the young woman's clothing from the reference. Exclude the older woman, the security employee, their bodies, and the wider monitor arrangement from the facial 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\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong retains her backpack and smartphone as the CCTV segment remains too dark to identify Min-sook's male contact.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "지시된 얼굴 클로즈업과 입술을 깨문 표정을 정확히 담았으며, 이전 샷의 비니 복장 연출도 잘 유지했습니다." }, { "label": "A", "score": 5, "verdict_ko": "클로즈업과 표정 연출은 적절하나, 이전 샷에서 요구된 비니 복장이 누락되어 연속성이 떨어집니다." }, { "label": "B", "score": 3, "verdict_ko": "얼굴 클로즈업 프레이밍 지시를 위반하여 모니터 배경이 넓게 드러났으며, 입술을 깨문 표정도 반영되지 않았습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S20sh3_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "카메라 프레임에 어깨가 명확히 보이지만, 유지해야 할 배낭의 어깨끈이 양쪽 모두 누락되었습니다.", "fix_en": "Add the thick green backpack straps over both of her visible shoulders." }, { "issue_ko": "지시문에는 입술을 깨물었다고만 되어 있으나, 프롬프트에 없는 피(상처)가 입술에 추가되었습니다.", "fix_en": "Remove the blood from her bitten lower lip." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add the thick green backpack straps over both of her visible shoulders.\n- Remove the blood from her bitten lower lip.\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": "prev+엔티티" }, "S21sh2::ab_conti": { "input_fingerprint": "0085a7dfdeca46a6", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 김형사가 한 손에 서류철을 든 채 삐딱한 자세로 혜수를 노려보는 상체.\n\nLOCATION (lock): A police station with a general investigators' office, clustered desks, computer monitors, paperwork, and a nearby meeting area. Enlarged crime-scene photographs are spread across one 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): Su-ri-yeong has brought her doll-trimmed backpack and smartphone into the police station.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "상체 구도, 삐딱한 자세, 지정된 소품(서류철, 인형 가방, 스마트폰) 및 캐릭터 외형을 충실하게 재현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "캐릭터 의상이 기준과 다르고, 지시된 자세와 어긋나며 지정된 가방 소품이 누락되었습니다." }, { "label": "C", "score": 3, "verdict_ko": "지문에 없는 다수의 인물이 배경에 추가된 치명적 오류가 있으며, 프레이밍이 과도하게 넓습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S21sh2.png" }, { "label": "CHARACTER REFERENCE — 김형사: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "여학생이 들고 있는 스마트폰의 화면(기능면)이 본인이 아닌 카메라 방향을 향하고 있습니다.", "fix_en": "Rotate the smartphone so its back faces the camera and the screen faces the girl." }, { "issue_ko": "서류철을 쥐고 있는 김형사의 오른손 손가락들이 여러 개로 뭉개져 해부학적으로 일그러져 있습니다.", "fix_en": "Correct the anatomy of Detective Kim's right hand to show normal human fingers holding the file folder." }, { "issue_ko": "화면 좌측 파티션 벽에 붙은 종이에 텍스트(문자)가 명확하게 생성되어 있어 '텍스트 없음' 규칙을 위반했습니다.", "fix_en": "Remove all text and writing from the paper pinned to the partition on the left." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone so its back faces the camera and the screen faces the girl.\n- Correct the anatomy of Detective Kim's right hand to show normal human fingers holding the file folder.\n- Remove all text and writing from the paper pinned to the partition on the left.\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." }, "S21sh9": { "input_fingerprint": "18b68f5979b711b5", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 눈을 동그랗게 뜨고 입을 떡 벌린 채 충격에 굳은 얼굴 클로즈업.\n\nLOCATION (lock): A police station with a general investigators' office, clustered desks, computer monitors, paperwork, and a nearby meeting area. Enlarged crime-scene photographs are spread across one 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 daylight police-office look, desk-filled background, institutional materials, and the young woman's clothing from the reference. Exclude the detective holding the file and the earlier confrontation; keep other officers outside the tight shocked-face frame.\n\nREALIZE FIGURATIVE LANGUAGE AS A LIVE-ACTION SHOT: the Korean shot\ntext may describe characters metaphorically, figuratively or with\nexaggeration. Photograph what a real movie camera would actually\nrecord on a physical set — exaggerated or figurative impressions\nbecome realistic staging choices (lighting, distance, angle,\nwardrobe), not literal fantasy imagery.\nEVERY CHARACTER IS A HUMAN BEING: unless the story explicitly\nfeatures non-human or virtual beings (as in science fiction or\nfantasy), every character — however indirectly, vaguely or\nfiguratively the text describes them — IS a real human. When the\ntext gives no direct visual description of a person, IMAGINE one\nand still show them as a concrete, fully-formed human being:\nalways render the human form to the maximum extent — build,\nposture, face, hands, clothing — never reduce a person to a shape,\nblob, solid silhouette or abstract mass.\n\nEXPRESSIONS ARE ACTED, NEVER ANATOMICAL: when the text describes a\nperson's eyes, face or presence emotionally or figuratively (vacant,\nhollow, dazed, burning, lifeless gaze and the like), it describes an\nACTOR'S PERFORMANCE captured by a real camera — realize it ONLY\nthrough gaze direction, focus, eyelids, facial muscles, stillness\nand posture. Every living person's eyes remain anatomically normal\nhuman eyes with a natural iris and pupil, natural sclera and normal\nproportions; NEVER whiten, blank out, cloud over, glow, enlarge or\notherwise alter eyeballs, skin or anatomy — unless the story\nexplicitly declares that being non-human or supernatural in form.\n\nPROPS FACE THE RIGHT WAY: every handheld or used object must be\noriented exactly as its real-world use requires. A person reading,\nwatching or operating something (a phone, a photograph, a paper,\nany device) has its functional side — screen, front, page — facing\nTHEIR OWN eyes; the camera then sees whatever side the staging\ngeometry implies (often its back). Show the functional side to the\ncamera ONLY when the shot text itself stages it toward the viewer.\nNever flip, mirror or reverse an object's front and back.\n\nCARRIED STATE (persist exactly — must match the neighbouring shots of this scene): Su-ri-yeong still has her doll-trimmed backpack and smartphone while she examines the location result.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 7, "B": 5, "C": 3 }, "selected": "A", "ranking": [ "A", "B", "C" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "충격받은 얼굴 클로즈업을 정확히 구현했으며, 기준 배경, 묶은 머리, 스마트폰 소품까지 완벽히 일치합니다." }, { "label": "B", "score": 5, "verdict_ko": "표정과 샷의 크기는 적합하나, 배경 구조가 기준 이미지와 다소 다르고 지정된 스마트폰이 보이지 않습니다." }, { "label": "C", "score": 3, "verdict_ko": "배경과 표정 연출은 우수하지만, 지정된 '묶은 머리'를 위반하고 머리를 풀고 있어 감점되었습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S21sh2_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터의 양손 해부학이 좌우가 뒤바뀌어 있습니다. 스마트폰을 쥔 오른손과 주먹 쥔 왼손 모두 엄지손가락이 신체 안쪽이 아닌 바깥쪽을 향해 잘못 그려졌습니다.", "fix_en": "Redraw both hands with correct anatomy: place the thumb of the right hand holding the phone on the inside (viewer's right), and the thumb of the left fist on the inside (viewer's left)." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Redraw both hands with correct anatomy: place the thumb of the right hand holding the phone on the inside (viewer's right), and the thumb of the left fist on the inside (viewer's left).\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": "prev+엔티티" }, "S21sh17": { "input_fingerprint": "53ae617508353760", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 접힌 메모지를 바지 주머니 틈새로 반쯤 밀어 넣은 정지 동작.\n\nLOCATION (lock): A police station with a general investigators' office, clustered desks, computer monitors, paperwork, and a nearby meeting area. Enlarged crime-scene photographs are spread across one 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 daylight office lighting, desk-area materials, and the young woman's clothing from the reference. Exclude the detective, the file confrontation, and other officers because she acts covertly while the office is empty; frame only the hand placing the folded note into the 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): Su-ri-yeong puts the folded note containing Min-sook's last-location coordinates into her pocket. She also retains her doll-trimmed 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 5, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "지시된 클로즈업보다 넓게 프레이밍되었으나, 인형이 달린 배낭과 정확한 경찰서 배경을 유일하게 모두 충족했습니다." }, { "label": "B", "score": 5, "verdict_ko": "레퍼런스(캐비닛과 책상)와 다른 창문 배경이 나타나 장소 고정 지시를 위반했으며 배낭의 인형 장식도 누락되었습니다." }, { "label": "A", "score": 3, "verdict_ko": "배경과 메모지 텍스트는 완벽하나, 왼팔에 오른손이 달려 있는 심각한 해부학적 오류(하드 위반)가 있습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S21sh2_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 위치 좌표를 적은 노란색 메모지: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "샷 텍스트에 '접힌 메모지'로 명시되었으나, 손에 든 메모지가 전혀 접히지 않은 평평한 상태입니다.", "fix_en": "Transform the flat yellow note in the right hand into a folded piece of yellow paper." }, { "issue_ko": "유지해야 할 소지품으로 지시된 스마트폰이 화면 내에 보이지 않습니다.", "fix_en": "Add a smartphone being held in the character's visible left hand hanging down." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Transform the flat yellow note in the right hand into a folded piece of yellow paper.\n- Add a smartphone being held in the character's visible left hand hanging down.\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": "prev+엔티티" }, "S22sh1::ab_conti": { "input_fingerprint": "1ee7ed8a27cb9ca9", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 맑은 낮 경찰서 건물 앞, 수리영이 자전거 안장에 엉덩이를 뗀 채 한쪽 페달을 강하게 밟아 누른 역동적인 정지 포즈.\n\nLOCATION (lock): The exterior entrance and parking area of a police station, with a bicycle parked nearby and direct access to the street. 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): Su-ri-yeong rides 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 3, "B": 8, "C": 6 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "자전거 앞바퀴 포크의 물리적 구조 오류와 오른팔 누락이 발생했으며, 명시된 소지품이 제대로 묘사되지 않았습니다." }, { "label": "B", "score": 8, "verdict_ko": "안장에서 일어선 역동적인 포즈와 지정된 장소를 훌륭하게 재현했으며, 메모지, 스마트폰, 배낭 등 모든 소지품의 상태를 정확하게 구현했습니다." }, { "label": "C", "score": 6, "verdict_ko": "요구된 포즈와 장소 묘사는 양호하나 스마트폰이 누락되었고, 불필요하게 뒤를 돌아보는 자세가 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B05.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S22sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "레이아웃 스케치에 지정된 포즈와 달리 오른팔이 핸들을 잡고 있으며, 스마트폰이 손이 아닌 백팩 옆면에 비현실적으로 붙어 있습니다.", "fix_en": "Reposition the right arm to reach backward and hold the smartphone in the right hand, exactly matching the pose in the layout sketch." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Reposition the right arm to reach backward and hold the smartphone in the right hand, exactly matching the pose in the layout sketch.\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." }, "S23sh1": { "input_fingerprint": "8765e53f888588cf", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 어두운 PC방 내부, 모니터 불빛에 비친 수리영이 마우스를 쥔 채 화면을 뚫어지게 응시하는 상체.\n\nLOCATION (lock): A public computer café with individual computer stations, monitors, internet access, and a printer used to produce several map sheets. 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 Su-ri-yeong's pocket, and several printed maps are folded into her doll-trimmed backpack alongside her other carried belongings.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 6, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "지시된 구도와 조명, 기준 의상을 정확히 재현했으며, 인형 달린 가방과 배경의 프린터까지 세부 설정을 가장 충실히 반영했습니다." }, { "label": "C", "score": 6, "verdict_ko": "인물, 의상, 구도는 지시사항에 잘 부합하나, 장소 설정에 명시된 프린터 등의 디테일이 화면에 포함되지 않았습니다." }, { "label": "A", "score": 3, "verdict_ko": "프롬프트에 언급되지 않은 인물들이 배경에 등장하여 심각한 위반에 해당하며, 기준 이미지의 의상도 반영되지 않았습니다." } ], "refs": [ { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "모니터 화면(좌측 상단 좌표 및 지도 위)에 텍스트와 숫자가 포함되어 있습니다.", "fix_en": "Remove all text, numbers, and labels from the computer screen." }, { "issue_ko": "마우스를 쥐고 있는 왼쪽 팔에 오른손이 달려 있습니다(엄지손가락의 위치가 반대임).", "fix_en": "Replace the hand on the mouse with an anatomically correct left hand." }, { "issue_ko": "프롬프트에 명시된 '가방에 접어 넣은 인쇄된 지도들'이 보이지 않습니다.", "fix_en": "Add several folded printed maps visibly sticking out of the backpack." } ] }, "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 labels from the computer screen.\n- Replace the hand on the mouse with an anatomically correct left hand.\n- Add several folded printed maps visibly sticking out of the backpack.\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" }, "S24sh3": { "input_fingerprint": "3e91b5b6d64b100a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, heavy rain and storm clouds.\n\nSHOT TEXT (authoritative, Korean): 수리영의 머리카락과 얼굴이 빗물에 흠뻑 젖은 채 앞을 매섭게 노려보는 측면 얼굴 클로즈업.\n\nLOCATION (lock): An exposed coastal route beside the sea, traversed by bicycle during a heavy shower. Thick storm clouds, thunder, lightning, and driving rain dominate the setting. 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): Su-ri-yeong rides through the rain carrying the printed maps in her backpack, the coordinate note in her pocket, and 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 6, "A": 5 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "스토리보드의 시선 방향(우측)과 측면 클로즈업 구도를 정확히 따랐으며, 비에 젖은 디테일과 매서운 표정 지시를 충실히 구현했습니다." }, { "label": "B", "score": 6, "verdict_ko": "환경 묘사와 표정은 우수하나, 스토리보드에 제시된 인물의 위치와 시선 방향이 좌우 반전되어 구도 일치도가 떨어집니다." }, { "label": "A", "score": 5, "verdict_ko": "스토리보드와 좌우 구도가 반전되었으며, 텍스트가 요구한 '매섭게 노려보는' 표정에 비해 시선이 다소 밋밋합니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S24sh3.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": "lane(map_marker): 스케치+엔티티" }, "S25sh5": { "input_fingerprint": "4353b5b6f5290341", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 붉은 원형 표식 정중앙에 핏자국이 묻은 낡은 사진 한 장이 핀으로 꽂혀 있는 클로즈업.\n\nLOCATION (lock): A ransacked rooftop apartment with muddy or dusty footprints, an open living-room window, overturned dining chairs, and household items pulled apart. In the adjoining bedroom, a red circle marks the wall around a bloodstained 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 apartment has been ransacked, with the living-room window open and dining chairs overturned. A bloodstained old photograph is pinned at the center of a red circle on Su-ri-yeong's wall, while her maps remain in her 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": { "A": 8, "B": 6, "C": 7 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 8, "verdict_ko": "클로즈업 샷의 의도를 가장 잘 살렸으며, 질감과 해질녘의 자연스러운 빛 표현이 우수합니다." }, { "label": "B", "score": 6, "verdict_ko": "프레이밍이 약간 넓은 편이며, 조명 톤이 지나치게 푸른빛을 띠어 해질녘의 느낌이 덜합니다." }, { "label": "C", "score": 7, "verdict_ko": "클로즈업 구도와 핏자국 묘사는 좋으나, 색감이 다소 따뜻하게 치우쳐 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B02.png" } ], "critique": { "issues": [ { "issue_ko": "사진이 붉은 원형 표식의 정중앙에 있지 않고 왼쪽으로 눈에 띄게 치우쳐 있습니다.", "fix_en": "Center the pinned photograph exactly within the drawn red circle." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Center the pinned photograph exactly within the drawn red circle.\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)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B02", "choice": "Candidate B", "confident": true, "reason_ko": "샷 설명에서 '붉은 원형 표식 정중앙에 핏자국이 묻은 낡은 사진'을 명확히 요구하고 있으며, 침실 벽면에 붉은 원과 그 안에 핀으로 꽂힌 사진이 있는 후보 B가 정확히 일치합니다.", "kept": "L04B02" } }, "S25sh8::ab_conti": { "input_fingerprint": "266fe90c2b16a0f5", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 동공이 커진 채 사진을 눈앞에 바짝 들고 굳어버린 얼굴.\n\nLOCATION (lock): A ransacked rooftop apartment with muddy or dusty footprints, an open living-room window, overturned dining chairs, and household items pulled apart. In the adjoining bedroom, a red circle marks the wall around a bloodstained 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): Su-ri-yeong has removed the bloodstained old photograph from the red wall circle and holds it tightly. She wears her backpack containing the printed maps; the apartment remains ransacked behind 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 5, "A": 4 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "사진을 캐릭터가 보는 방향(뒷면이 카메라를 향함)으로 올바르게 들고 있으며, 표정과 난장판이 된 공간의 묘사가 지시문과 가장 잘 일치합니다." }, { "label": "C", "score": 5, "verdict_ko": "침실의 붉은 원 위치 등 배경 묘사는 정확하나, 사진의 앞면이 카메라를 향하게 들어 소품 방향 규칙을 위반했습니다." }, { "label": "A", "score": 4, "verdict_ko": "사진이 카메라를 향해 있고 지도를 배낭이 아닌 손에 함께 들고 있어 소품 및 휴대 상태 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S25sh8.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 두 어린 여자아이가 찍힌 피 묻은 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물의 눈알과 눈동자가 해부학적으로 비정상적으로 크게 부풀려져 있어, 표정을 위해 신체 구조를 변형하지 말라는 지침을 위반했습니다.", "fix_en": "Restore the character's eyes to normal, anatomically correct human size and proportions while keeping her shocked expression." }, { "issue_ko": "프롬프트상 인접한 침실에 있어야 할 붉은색 원이 거실의 TV 위 벽면에 잘못 배치되어 있습니다.", "fix_en": "Remove the red circle from the wall above the television in the background." }, { "issue_ko": "공간 레퍼런스에서 막힌 벽이었던 화면 우측 식탁 뒤 벽면에 창문이 임의로 추가되었습니다.", "fix_en": "Remove the window on the right wall and replace it with a solid, plain wallpapered 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- Restore the character's eyes to normal, anatomically correct human size and proportions while keeping her shocked expression.\n- Remove the red circle from the wall above the television in the background.\n- Remove the window on the right wall and replace it with a solid, plain wallpapered 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." }, "S26sh3": { "input_fingerprint": "e9ee08399cd5ec24", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, dense fog after rain.\n\nSHOT TEXT (authoritative, Korean): 갯바위에 놓인 수리영의 스마트폰 액정 화면에 '혜수이모'라는 발신자 이름이 떠 있는 클로즈업.\n\nLOCATION (lock): An empty harbor wrapped in dense fog after a shower, with vessels obscured and rocky shoreline seating near the water. A lone small boat approaches through the fog and ties up at the dock. 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): Su-ri-yeong retains the old photograph, her map-filled backpack, and her smartphone while waiting at the foggy harbor.\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": 5 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "스토리보드와 일치하는 타이트한 클로즈업 앵글을 완벽히 구현했으며 지시된 화면 텍스트도 정확합니다." }, { "label": "B", "score": 3, "verdict_ko": "스토리보드의 스케치용 진동 효과선이 실사 화면에 그대로 노출되는 치명적인 규정 위반이 있습니다." }, { "label": "C", "score": 5, "verdict_ko": "카메라 앵글이 스토리보드보다 낮고 넓어 프레이밍이 일치하지 않으며, 스마트폰 인터페이스에 영문이 포함되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S26sh3.png" } ], "critique": { "issues": [ { "issue_ko": "스마트폰 화면 상단에 표시된 시간 '13:40'은 프롬프트에 지정된 시간대인 '해질녘(dusk)'과 모순됩니다.", "fix_en": "Change the time displayed on the top left of the phone screen to a time appropriate for dusk (e.g., 18:40) or remove it." }, { "issue_ko": "프롬프트에 지시되지 않은 의미 불명의 텍스트가 통화 수락(녹색) 및 거절(빨간색) 버튼 아래에 생성되어 있습니다.", "fix_en": "Remove the garbled Korean text located directly below the green and red call button icons on the screen." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the time displayed on the top left of the phone screen to a time appropriate for dusk (e.g., 18:40) or remove it.\n- Remove the garbled Korean text located directly below the green and red call button icons on 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": "lane(map_marker): 스케치" }, "S26sh8": { "input_fingerprint": "a2ab916c90aa085f", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, dense fog after rain.\n\nSHOT TEXT (authoritative, Korean): 수리영이 인우 앞에 마주 선 채 두 손을 모으고 절박한 눈빛으로 올려다보는 상체 구도.\n\nLOCATION (lock): An empty harbor wrapped in dense fog after a shower, with vessels obscured and rocky shoreline seating near the water. A lone small boat approaches through the fog and ties up at the dock. 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): Su-ri-yeong still has the old photograph and wears the backpack containing her printed maps as she pleads with In-woo.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 인우 (한국인 젊은 남성, 24세, 젊은 얼굴, 짙은색 머리, 짧은 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": 8, "A": 6, "C": 3 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 8, "verdict_ko": "구도와 절박한 감정 표현이 훌륭하며, 인우의 의상 레퍼런스와 정박된 보트의 디테일을 가장 잘 살렸습니다." }, { "label": "A", "score": 6, "verdict_ko": "전반적인 구도와 지시사항은 따랐으나, 인우의 의상이 레퍼런스와 일치하지 않고 정박된 보트의 묘사가 부족합니다." }, { "label": "C", "score": 3, "verdict_ko": "수리영의 손가락이 기형적으로 렌더링된 치명적인 해부학적 오류가 있어 감점되었습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S26sh8.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": "지문에서 카메라를 향해 보여주라는 지시가 없었음에도, 수리영이 들고 있는 오래된 사진의 앞면(이미지)이 카메라를 향하고 있습니다.", "fix_en": "Rotate the photograph in Su-ri-yeong's hands so that only its blank back side faces the camera." }, { "issue_ko": "수리영이 모으고 있는 두 손의 손가락 개수가 너무 많고 관절이 부자연스럽게 얽혀 해부학적 오류가 있습니다.", "fix_en": "Correct the anatomy of Su-ri-yeong's clasped hands to show a natural grip with the correct number of normal fingers." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the photograph in Su-ri-yeong's hands so that only its blank back side faces the camera.\n- Correct the anatomy of Su-ri-yeong's clasped hands to show a natural grip with the correct number of normal fingers.\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" }, "S26sh12": { "input_fingerprint": "450695c167dc7e96", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, dense fog after rain.\n\nSHOT TEXT (authoritative, Korean): 수리영이 눈을 지그시 감은 채 옅은 미소를 띤 얼굴 클로즈업.\n\nLOCATION (lock): An empty harbor wrapped in dense fog after a shower, with vessels obscured and rocky shoreline seating near the water. A lone small boat approaches through the fog and ties up at the dock. 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 dense harbor fog, muted dusk palette, concrete pier setting, and the young woman's clothing from the reference. Exclude the boatman, the pleading hand pose, and the wider boat interaction from this close facial composition.\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-yeong retains the old photograph and map-filled backpack and uses her smartphone to leave Hye-su a voice message before departing.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 8, "C": 7, "A": 5 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 8, "verdict_ko": "지시된 완벽한 얼굴 클로즈업 앵글을 구현했으며, 눈을 감고 미소 짓는 표정과 안개 낀 항구 배경이 매우 정확합니다." }, { "label": "C", "score": 7, "verdict_ko": "표정과 배경 묘사는 훌륭하나, B에 비해 프레이밍이 미세하게 넓어 타이트한 클로즈업의 느낌이 다소 줄었습니다." }, { "label": "A", "score": 5, "verdict_ko": "소품을 포함시키기 위해 지시된 '얼굴 클로즈업' 샷의 프레이밍을 임의로 넓힌 점이 감점 요인입니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S26sh12.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S26sh8_sel.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": "lane(map_marker): 스케치+prev+엔티티" }, "S27sh2::ab_conti": { "input_fingerprint": "058e1ff8f3043896", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 고개를 치켜든 채 옆에 선 김형사를 향해 두 눈을 크게 뜬 혜수의 얼굴 클로즈업.\n\nLOCATION (lock): The same rooftop apartment after it has been disturbed, now empty and desolate. The entrance door rattles in the wind while the interior remains without a stated light source. 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 rooftop apartment remains ransacked and empty while Hye-su checks Su-ri-yeong's recorded message.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": 6, "C": 8 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "A", "score": 5, "verdict_ko": "기준 의상을 가장 정확하게 재현했으나, 배경 복도에 원본에 없는 외부 출입문을 임의로 생성하여 공간 구조(우선순위 3)를 위반했습니다." }, { "label": "B", "score": 6, "verdict_ko": "공간 구조와 스마트폰 파지 방향은 적절하나, 고개를 치켜든 포즈가 다소 밋밋하며 기준 의상(목티)을 반영하지 못했습니다." }, { "label": "C", "score": 8, "verdict_ko": "고개를 치켜든 포즈, 정확한 공간 구조(화장실 복도 및 어질러진 바닥), 일몰 조명, 올바른 스마트폰 방향을 모두 잘 구현했으나 의상(목티)이 누락되었습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S27sh2.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "혜수가 쥔 휴대폰의 뒷면(카메라 렌즈 부분)이 그녀의 얼굴을 향하고 있어 기기 방향이 잘못되었습니다.", "fix_en": "Flip the smartphone in the woman's hand so the screen faces her face and the camera lenses face outward." }, { "issue_ko": "카메라를 향해 있는 혜수가 캐릭터 레퍼런스에 지정된 의상(모자, 회색 터틀넥, 검은색 전술 재킷)을 입고 있지 않습니다.", "fix_en": "Dress the front-facing woman exactly as shown in the character reference, including the cap, grey turtleneck, and black field jacket." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Flip the smartphone in the woman's hand so the screen faces her face and the camera lenses face outward.\n- Dress the front-facing woman exactly as shown in the character reference, including the cap, grey turtleneck, and black field 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." }, "S27sh6": { "input_fingerprint": "35f60d02e58383ef", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 엉망이 된 빈 거실에 짙은 그림자만이 길게 드리워진 쓸쓸한 실내 전경.\n\nLOCATION (lock): The same rooftop apartment after it has been disturbed, now empty and desolate. The entrance door rattles in the wind while the interior remains without a stated light source. 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 empty apartment remains in its ransacked state, with the disturbed furnishings left unrestored.\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": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "해질녘의 어스름한 조명과 길게 드리워진 그림자로 엉망이 된 빈 공간의 쓸쓸한 분위기를 가장 정확하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "실내 기물과 난장판이 된 상태는 원본과 흡사하나, 창문 방향과 맞지 않는 우측 광원이 공간의 사실성을 떨어뜨립니다." }, { "label": "B", "score": 3, "verdict_ko": "아무도 없어야 할 공간임에도 복도 화장실 문틀에 정체를 알 수 없는 신체 일부(팔)가 생성되어 지침을 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" } ], "critique": { "issues": [ { "issue_ko": "왼쪽 열린 문의 손잡이가 문이 열리는 가장자리(오른쪽)가 아닌 경첩이 있는 쪽(왼쪽)에 달려 있습니다.", "fix_en": "Move the silver door handle from the left edge of the left door panel to its right edge." }, { "issue_ko": "바닥에 쓰러진 TV에 스탠드가 그대로 붙어 있는데, TV 장식장 위에도 빈 TV 스탠드가 중복해서 남아 있습니다.", "fix_en": "Remove the empty TV stand base from the top of the black TV console." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Move the silver door handle from the left edge of the left door panel to its right edge.\n- Remove the empty TV stand base from the top of the black TV console.\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)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B03", "choice": "Candidate C", "confident": true, "reason_ko": "\"빈 거실\"과 \"실내 전경\"이라는 묘사에 가장 잘 부합하는 거실 전체 샷이므로 현재 할당된 샷을 유지합니다.", "kept": "L04B03" } }, "S28sh5": { "input_fingerprint": "2c24a9107e959e5e", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 수리영의 손가락 끝이 낡은 사진 표면을 짚은 채 인우 눈앞 허공에 멈춰 있는 클로즈업.\n\nLOCATION (lock): A small fishing boat well out at sea, with an open working deck and an enclosed wheelhouse. The deck provides an unobstructed view of the surrounding 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): Su-ri-yeong has brought the printed maps in her backpack and the bloodstained old photograph aboard In-woo'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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 8, "B": 5, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 8, "verdict_ko": "지정된 1인칭 클로즈업 프레이밍을 정확히 따랐으며, 스토리보드의 배경 배치(좌측 해치)와 소품 디테일을 훌륭하게 구현했습니다." }, { "label": "B", "score": 5, "verdict_ko": "클로즈업 구도는 준수하나, 스토리보드에 명시된 배경 구조(좌측 해치)를 무시하고 조타실을 중앙에 배치했습니다." }, { "label": "A", "score": 3, "verdict_ko": "손가락 클로즈업 지시를 어기고 인물의 뒷모습을 과도하게 노출했으며, 사진을 쥐고 있는 손의 해부학적 구조가 불가능합니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S28sh5.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 두 어린 여자아이가 찍힌 피 묻은 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "사진을 붙잡고 있는 손이나 지지대가 없어 사진이 허공에 떠 있음.", "fix_en": "Add fingers or a hand holding the edges of the photograph so it does not float in mid-air." }, { "issue_ko": "사진 하단에 레퍼런스와 다른 알아볼 수 없는 텍스트가 임의로 생성됨.", "fix_en": "Replace the gibberish text on the photograph with the exact text from the reference ('1979 - 식당 앞 in the restaurant')." }, { "issue_ko": "허용된 인물 목록(수리영)에 없는 남성 캐릭터가 배경에 등장함.", "fix_en": "Remove the male character from the background entirely." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add fingers or a hand holding the edges of the photograph so it does not float in mid-air.\n- Replace the gibberish text on the photograph with the exact text from the reference ('1979 - 식당 앞 in the restaurant').\n- Remove the male character from the background entirely.\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" }, "S28sh6": { "input_fingerprint": "19a56ab40a954941", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk.\n\nSHOT TEXT (authoritative, Korean): 인우가 두 눈을 크게 뜨고 입을 살짝 벌린 채 당황한 표정의 얼굴 클로즈업.\n\nLOCATION (lock): A small fishing boat well out at sea, with an open working deck and an enclosed wheelhouse. The deck provides an unobstructed view of the surrounding 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): The old photograph remains held out between Su-ri-yeong and In-woo, while her remaining printed maps stay in her backpack.\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 젊은 남성, 24세, 젊은 얼굴, 짙은색 머리, 짧은 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": "당황한 표정과 구도는 적절하나 레퍼런스의 의상(비니, 재킷)이 누락되었고, 사진의 앞면이 카메라를 향해 소품 방향 지침을 위반했습니다." }, { "label": "B", "score": 7, "verdict_ko": "스토리보드 구도, 인물의 의상 및 표정을 정확히 구현했으며, 사진의 뒷면 노출과 배낭 속 지도 등 세부 지침을 완벽하게 준수했습니다." }, { "label": "C", "score": 4, "verdict_ko": "스토리보드와 달리 인물이 중앙에 배치되었고, 재킷 누락 및 사진 앞면 노출 등 주요 지침을 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S28sh6.png" }, { "label": "CHARACTER REFERENCE — 인우: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "사진의 앞면(이미지)이 카메라를 향하고 있어 인우가 뒷면을 보고 있으므로, 소품의 방향 규칙에 어긋납니다.", "fix_en": "Flip the photograph so its image side faces In-woo and its blank back faces the camera." }, { "issue_ko": "지도는 수리영의 배낭 안에 있어야 하는데, 배경에 있는 배낭 밖으로 지도가 노출되어 있습니다.", "fix_en": "Remove the exposed map from the backpack in the background so it is completely hidden." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Flip the photograph so its image side faces In-woo and its blank back faces the camera.\n- Remove the exposed map from the backpack in the background so it is completely hidden.\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" }, "S29sh1::ab_conti": { "input_fingerprint": "298975fdff5ef1f0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 좁은 조타실 내부, 조타기를 향해 한 팔을 뻗은 인우와 두 팔을 양옆으로 뻗어 그의 앞을 가로막은 수리영의 정지 자세.\n\nLOCATION (lock): The compact wheelhouse of a small fishing boat, containing the steering controls and navigation area. Numerous circular forms are visible around the cabin, including a newly perceived red circle. 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): Su-ri-yeong retains the old photograph and her map-filled backpack as she blocks In-woo inside the wheelhouse.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 인우 (한국인 젊은 남성, 24세, 젊은 얼굴, 짙은색 머리, 짧은 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 지도 배낭과 옛 사진 유지, 인우가 조타기를 향해 팔을 뻗는 핵심 동작을 모두 정확히 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "수리영의 옛 사진과 지도가 누락되었으며, 인우가 조타기를 향해 팔을 뻗는 동작이 보이지 않습니다." }, { "label": "C", "score": 4, "verdict_ko": "옛 사진과 지도가 없고, 인우가 조타기가 아닌 다른 곳을 향해 팔을 뻗고 있어 요구사항을 충족하지 못했습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L21B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S29sh1.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": "수리영이 들고 있는 사진의 앞면이 카메라를 향하고 있지만, 샷 텍스트에 사진을 보여준다는 지시가 없습니다.", "fix_en": "Flip the photograph in Su-ri-yeong's left hand so its blank back side faces the camera." }, { "issue_ko": "대시보드의 모니터 화면에 텍스트가 생성되어 있습니다 ('텍스트 금지' 규칙 위반).", "fix_en": "Remove all text from the monitor screen on the dashboard." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Flip the photograph in Su-ri-yeong's left hand so its blank back side faces the camera.\n- Remove all text from the monitor screen on the dashboard.\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." }, "S29sh4": { "input_fingerprint": "74f49d05331e721a", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영의 집게손가락 끝이 낡은 사진 속 어린 여자아이의 흑백 얼굴 위를 강하게 짚고 있는 손 클로즈업.\n\nLOCATION (lock): The compact wheelhouse of a small fishing boat, containing the steering controls and navigation area. Numerous circular forms are visible around the cabin, including a newly perceived red circle. 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 cramped wheelhouse lighting, worn interior materials, the old photograph, and the young woman's clothing from the reference. Exclude the full-body blocking pose, the man reaching for the controls, and the steering equipment from the hand close-up unless only softly visible at the edge.\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-yeong still holds the old photograph, pressing her finger onto the pictured child identified as Min-sook; her printed maps remain in her backpack.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "A": 5, "C": 4 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "지정된 손 클로즈업 구도를 잘 보여주며, 사진을 들고 여자아이의 얼굴을 짚는 동작과 기준 이미지의 회색 바람막이 소매를 가장 정확하게 재현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "클로즈업 구도는 훌륭하나, 사진을 들고 있는 손이 생략되었고 기준과 다른 어둡고 두꺼운 재질의 소매가 노출되어 아쉽습니다." }, { "label": "C", "score": 4, "verdict_ko": "명시된 '손 클로즈업' 앵글을 벗어나 인물의 얼굴 측면까지 프레임에 포함시키는 넓은 구도를 잡아 프레이밍 원칙을 위반했습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S29sh1_sel.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 두 어린 여자아이가 찍힌 피 묻은 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "집게손가락 끝이 사진 속 여자아이의 얼굴이 아닌, 그 위쪽의 건물 벽면을 짚고 있습니다.", "fix_en": "Move the tip of the index finger down so that it presses directly onto the face of the young girl sitting on the bench in the photograph." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Move the tip of the index finger down so that it presses directly onto the face of the young girl sitting on the bench in 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": "prev+엔티티" }, "S29sh10": { "input_fingerprint": "9221fc0cf67aa563", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영의 창백한 이마가 인우의 단단한 어깨 위에 닿은 채 멈춘 찰나.\n\nLOCATION (lock): The compact wheelhouse of a small fishing boat, containing the steering controls and navigation area. Numerous circular forms are visible around the cabin, including a newly perceived red circle. 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 cramped wheelhouse look, dusk light, worn controls, and both characters' clothing from the reference. Preserve their positions near the controls but replace the earlier blocking stance with the young woman collapsed against the man's shoulder. Exclude the prominently displayed photograph and the earlier outstretched arms.\n\nIMMOBILE CHARACTER POSE — CANONICAL (identical wherever this character appears in ANY panel; on any conflict THIS POSE WINS): She is slumped forward against Inwoo with her pale forehead resting on his shoulder and her head supported there. Her body hangs downward in the collapse, with her limbs no longer holding her upright.\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-yeong is slumped forward against In-woo with her pale forehead resting on his shoulder and her head supported there, unconscious. With no stated transfer or disposal, the old photograph remains with her and the printed maps remain in her backpack.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 인우 (한국인 젊은 남성, 24세, 젊은 얼굴, 짙은색 머리, 짧은 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": 4, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "지시된 카메라 앵글 변경을 성공적으로 수행했으며, 캐릭터의 자세와 의상, 소품(배낭 속 지도)을 이전 장면과 일관되게 잘 묘사했습니다." }, { "label": "C", "score": 4, "verdict_ko": "자세와 앵글은 양호하지만, 이전 장면에 없던 파란색 앞치마(캐릭터 레퍼런스 의상)가 혼합되어 의상 일관성을 위반했습니다." }, { "label": "A", "score": 3, "verdict_ko": "이전 장면의 카메라 구도를 그대로 모방하는 금지 사항을 어겼으며, 창문에 인위적인 빨간색 원이 렌더링되었습니다." } ], "refs": [ { "label": "PREVIOUS SHOT STILL — a visually related earlier shot of this same place: the location's look, materials, fixed features, lighting mood and each person's clothing are LOCKED to this photo; never copy its camera framing. If this photo shows a character who cannot move (dead or unconscious), that character's exact pose, position and orientation are ALSO LOCKED — treat their body as a fixed prop of the set that only the camera moves around.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S29sh1_sel.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": "수리영의 머리가 지정된 '긴 묶은 머리'가 아닌 짧은 단발머리로 묘사되었습니다.", "fix_en": "Change Su-ri-yeong's hair to long dark hair tied back, matching the character reference." }, { "issue_ko": "수리영의 '창백한 이마'가 인우의 어깨에 닿아야 한다는 샷 텍스트의 지시와 달리, 뺨과 얼굴 측면이 어깨에 닿아 있습니다.", "fix_en": "Adjust Su-ri-yeong's head posture so that her forehead rests directly against Inwoo's shoulder." }, { "issue_ko": "텍스트 생성이 금지되었음에도 불구하고 조타실 벽면과 계기판 주변('g함어음' 등)에 문자가 무작위로 생성되었습니다.", "fix_en": "Remove all text, numbers, and gibberish lettering from the control panels and background walls." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change Su-ri-yeong's hair to long dark hair tied back, matching the character reference.\n- Adjust Su-ri-yeong's head posture so that her forehead rests directly against Inwoo's shoulder.\n- Remove all text, numbers, and gibberish lettering from the control panels and background walls.\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": "prev+엔티티" }, "S30sh3": { "input_fingerprint": "cab5522f69c8c3da", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): sunset, golden light and dense fog.\n\nSHOT TEXT (authoritative, Korean): 낡은 소형 어선의 앞머리가 짙고 탁한 안개 덩어리 속으로 절반쯤 밀려 들어간 찰나의 멈춘 구도.\n\nLOCATION (lock): Open sea crossed by a small fishing boat, with a dense bank of fog forming a boundary across the water. The boat gradually disappears into the fog. 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): Su-ri-yeong remains unconscious aboard In-woo's boat as it enters the fog, still accompanied by her backpack of printed maps and the old photograph unless moved by In-woo.\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": 8, "A": 6, "B": 6 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 8, "verdict_ko": "스토리보드의 구도와 석양빛, 짙은 안개를 훌륭하게 구현했으며, 프롬프트에 명시된 배낭, 인쇄된 지도, 오래된 사진을 프레임 내에 정확히 배치하여 가장 높은 일치도를 보입니다." }, { "label": "A", "score": 6, "verdict_ko": "지정된 멈춘 구도와 탁한 안개의 분위기를 잘 살렸으나, 지시된 특정 소품(배낭, 지도, 사진)이 묘사되지 않았습니다." }, { "label": "B", "score": 6, "verdict_ko": "요구된 앵글과 시간대를 적절히 반영했지만 안개의 무게감이 다소 가벼우며, 명시된 소품들이 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/lane_sketch_S30sh3.png" } ], "critique": { "issues": [ { "issue_ko": "어선의 앞머리가 안개 덩어리 속으로 절반쯤 밀려 들어간 상태여야 하나, 뱃머리 전체가 안개 밖으로 선명하게 노출되어 있습니다.", "fix_en": "Envelop the front half of the boat in dense fog so that the bow is fully obscured and pushed inside the fog mass." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Envelop the front half of the boat in dense fog so that the bow is fully obscured and pushed inside the fog mass.\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" }, "S19sh6::ab_noconti": { "input_fingerprint": "23bebafe9d380c9c", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 입술을 굳게 다물고 마트 천장 모서리를 가리키는 손끝 클로즈업.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong still carries her doll-trimmed backpack and has arrived by bicycle.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 5, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "클로즈업 샷의 타이트한 구도를 가장 잘 살렸으며, 굳게 다문 입술과 천장을 가리키는 손끝의 디테일이 지문을 정확히 반영했습니다." }, { "label": "C", "score": 5, "verdict_ko": "표정과 자세는 적절하나 프레임이 넓어 클로즈업 요건에서 밀리며, 원본에 없는 '냉동식품' 텍스트를 배경에 생성했습니다." }, { "label": "A", "score": 3, "verdict_ko": "굳게 다문 입술이라는 지문을 어기고 입을 내밀고 있으며, 팔이 얼굴을 가려 인물의 식별이 어렵고 프레임도 다소 넓습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B02.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지시문에는 '입술을 굳게 다물고' 있다고 명시되어 있으나, 사진의 인물은 입술을 앞으로 둥글게 삐죽 내밀고 있습니다.", "fix_en": "Change the character's mouth so the lips are pressed firmly and flatly together in a tight line." }, { "issue_ko": "화면 좌측 하단에 인물의 오른쪽 어깨가 분명히 보이지만, 지시된 백팩을 메고 있음을 보여주는 어깨 끈이 전혀 없습니다.", "fix_en": "Add the thick green backpack strap resting over the character's right shoulder, matching the character reference." }, { "issue_ko": "카메라를 향해 뻗은 손의 해부학적 구조가 크게 왜곡되어 있습니다(관절 형태가 없는 비정상적으로 굵은 집게손가락과 기형적인 모양의 엄지손가락).", "fix_en": "Redraw the foreground hand to have correct normal human anatomical proportions, distinct finger joints, and a properly formed thumb." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the character's mouth so the lips are pressed firmly and flatly together in a tight line.\n- Add the thick green backpack strap resting over the character's right shoulder, matching the character reference.\n- Redraw the foreground hand to have correct normal human anatomical proportions, distinct finger joints, and a properly formed thumb.\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." }, "S19sh6::conti_ab_decision": { "fingerprint": "9ddcad7697c717e5", "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "2번 후보는 지시문이 요구한 손끝 클로즈업 샷 크기와 굳게 다문 입술의 표정을 훨씬 더 정확하게 구현했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "요청된 '손끝 클로즈업' 샷 크기를 더 충실하게 반영하여 손끝과 굳게 다문 입술을 강조했기 때문입니다." } ] } }, "S19sh6": { "input_fingerprint": "23bebafe9d380c9c", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 입술을 굳게 다물고 마트 천장 모서리를 가리키는 손끝 클로즈업.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong still carries her doll-trimmed backpack and has arrived by bicycle.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 5, "A": 3 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "클로즈업 샷의 타이트한 구도를 가장 잘 살렸으며, 굳게 다문 입술과 천장을 가리키는 손끝의 디테일이 지문을 정확히 반영했습니다." }, { "label": "C", "score": 5, "verdict_ko": "표정과 자세는 적절하나 프레임이 넓어 클로즈업 요건에서 밀리며, 원본에 없는 '냉동식품' 텍스트를 배경에 생성했습니다." }, { "label": "A", "score": 3, "verdict_ko": "굳게 다문 입술이라는 지문을 어기고 입을 내밀고 있으며, 팔이 얼굴을 가려 인물의 식별이 어렵고 프레임도 다소 넓습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B02.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지시문에는 '입술을 굳게 다물고' 있다고 명시되어 있으나, 사진의 인물은 입술을 앞으로 둥글게 삐죽 내밀고 있습니다.", "fix_en": "Change the character's mouth so the lips are pressed firmly and flatly together in a tight line." }, { "issue_ko": "화면 좌측 하단에 인물의 오른쪽 어깨가 분명히 보이지만, 지시된 백팩을 메고 있음을 보여주는 어깨 끈이 전혀 없습니다.", "fix_en": "Add the thick green backpack strap resting over the character's right shoulder, matching the character reference." }, { "issue_ko": "카메라를 향해 뻗은 손의 해부학적 구조가 크게 왜곡되어 있습니다(관절 형태가 없는 비정상적으로 굵은 집게손가락과 기형적인 모양의 엄지손가락).", "fix_en": "Redraw the foreground hand to have correct normal human anatomical proportions, distinct finger joints, and a properly formed thumb." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Change the character's mouth so the lips are pressed firmly and flatly together in a tight line.\n- Add the thick green backpack strap resting over the character's right shoulder, matching the character reference.\n- Redraw the foreground hand to have correct normal human anatomical proportions, distinct finger joints, and a properly formed thumb.\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.", "conti_ab": { "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "2번 후보는 지시문이 요구한 손끝 클로즈업 샷 크기와 굳게 다문 입술의 표정을 훨씬 더 정확하게 구현했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "요청된 '손끝 클로즈업' 샷 크기를 더 충실하게 반영하여 손끝과 굳게 다문 입술을 강조했기 때문입니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S19sh6__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S19sh6__ab_noconti_sel.png" }, "ref_mode": "플레이트+엔티티 (A/B: 무콘티 승)", "plate_select": { "candidates": { "A": "L09B02", "B": "L09B03" }, "assigned": "L09B02", "choice": "A", "confident": true, "reason_ko": "지문에서 '마트 천장 모서리'를 가리킨다고 명시하고 있으므로, 마트 내부의 천장이 보이는 공간인 후보 A가 적합합니다.", "chosen": "L09B02" } }, "S19sh10::ab_noconti": { "input_fingerprint": "ab412d887ecefc0e", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 스마트폰을 오른쪽 귀에 바짝 댄 채 입을 벌린 상체 구도.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong retains her backpack and smartphone, using the phone to explain the CCTV lead to Hye-su.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 3, "B": 7, "C": 3 }, "selected": "B", "ranking": [ "B", "A", "C" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "스마트폰 화면이 바깥을 향한 채 귀에 대고 있어 물리적 사용 방향 지침을 위반했습니다." }, { "label": "B", "score": 7, "verdict_ko": "스마트폰 방향이 올바르고 의상과 배낭이 정확히 일치하며, 입을 벌린 상체 구도를 잘 구현했습니다." }, { "label": "C", "score": 3, "verdict_ko": "스마트폰 화면 방향 오류라는 심각한 위반이 있으며, 캐릭터 의상 색상도 레퍼런스와 다릅니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B03.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": "스마트폰의 후면 카메라 모듈이 아래쪽을 향하고 있어, 기기를 거꾸로 들고 있는 상태입니다.", "fix_en": "Rotate the smartphone 180 degrees so the rear camera module is positioned at the top." }, { "issue_ko": "프롬프트에서 텍스트 생성을 금지했으나, 창문, 우측 녹색 간판, 파란색 플라스틱 상자 등에 문자가 포함되어 있습니다.", "fix_en": "Remove all text and characters from the glass window, the green sign on the brick wall, and the blue crates." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone 180 degrees so the rear camera module is positioned at the top.\n- Remove all text and characters from the glass window, the green sign on the brick wall, and the blue crates.\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." }, "S19sh10::conti_ab_decision": { "fingerprint": "6e8fa3c8605eb4c0", "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "2번은 프롬프트에 명시된 대형 마트 내부의 상품 진열대와 외부 자전거 주차 구역을 정확히 반영했으며, 스마트폰을 오른쪽 귀에 대고 입을 벌린 상체 구도 역시 충실히 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 1은 스마트폰을 쥔 오른손의 관절 방향과 손바닥 위치가 해부학적으로 불가능한 구조인 반면, 후보 2는 자연스러운 파지법을 보여주며 지문에 명시된 외부 보안 카메라까지 배경에 정확히 구현했습니다." } ] } }, "S19sh10": { "input_fingerprint": "0356b6bae3eb9126", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 스마트폰을 오른쪽 귀에 바짝 댄 채 입을 벌린 상체 구도.\n\nLOCATION (lock): A large supermarket in the quiet early hours, with merchandise aisles, staff areas, and bicycle parking outside. Exterior security cameras cover part of the area where employees enter or meet. 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): Su-ri-yeong retains her backpack and smartphone, using the phone to explain the CCTV lead to Hye-su.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 5, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "오른쪽 귀에 올바른 방향으로 폰을 대고 있으며, 추가 인물 없이 상체 구도와 배경을 정확하게 구현했습니다." }, { "label": "A", "score": 5, "verdict_ko": "구도와 배경은 우수하나, 스마트폰 화면이 바깥을 향하게 쥐고 있어 소품 방향 규칙을 위반했습니다." }, { "label": "B", "score": 3, "verdict_ko": "지시문에 없는 인물이 등장했고 왼쪽 귀에 폰을 대고 있으며, 배경에 콜라주 형태의 공간적 오류가 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L09B03.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S19sh10.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": "배경의 간판, 상점 창문, 파란색 플라스틱 상자 등에 텍스트가 생성되어 '어느 곳에도 텍스트 금지' 규칙을 위반했습니다.", "fix_en": "Remove all text and letters from the street signs, store windows, and blue crates in the background." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove all text and letters from the street signs, store windows, and blue crates 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.", "conti_ab": { "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "2번은 프롬프트에 명시된 대형 마트 내부의 상품 진열대와 외부 자전거 주차 구역을 정확히 반영했으며, 스마트폰을 오른쪽 귀에 대고 입을 벌린 상체 구도 역시 충실히 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 1은 스마트폰을 쥔 오른손의 관절 방향과 손바닥 위치가 해부학적으로 불가능한 구조인 반면, 후보 2는 자연스러운 파지법을 보여주며 지문에 명시된 외부 보안 카메라까지 배경에 정확히 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S19sh10__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S19sh10__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L09B02", "B": "L09B03" }, "assigned": "L09B03", "choice": "Candidate B", "confident": false, "reason_ko": "스마트폰 통화 중이라는 인물 묘사 외에 실내 매장인지 실외인지 하위 공간을 특정할 단서가 샷 텍스트에 없으므로 기존 할당된 후보를 유지함.", "kept": "L09B03" } }, "S20sh3::ab_noconti": { "input_fingerprint": "d7e49f2155f27a40", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 혜수와 한국인 남자 보안직원 뒤에 서서 두 손을 모아 꽉 쥔 채 굳은 표정으로 모니터를 응시하는 상체.\n\nLOCATION (lock): A supermarket administrative and security office centered on a CCTV monitoring station, with screens and playback equipment for reviewing recorded 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): Su-ri-yeong continues to have her doll-trimmed backpack and smartphone with her while viewing the footage.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 머리); 한국인 남자 보안직원 (한국인 성인 남성, 자연스러운 얼굴, 단정한 머리, 짙은색 짧은 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": 9, "C": 4 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "혜수가 누락되었고, 인물들의 배치와 시선이 모니터를 향해 같은 방향을 바라보는 것이 아니라 서로 마주보는 형태로 잘못 연출되었습니다." }, { "label": "B", "score": 9, "verdict_ko": "수리영이 다른 두 인물 뒤에 서서 두 손을 꽉 쥐고 모니터를 응시하는 상체 샷의 요구사항을 위치, 표정, 소품까지 가장 완벽하게 구현했습니다." }, { "label": "C", "score": 4, "verdict_ko": "뒷모습으로 촬영되어 수리영의 굳은 표정을 확인할 수 없으며, 수리영이 다른 인물들 뒤에 서 있다는 위치 설정도 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L15B03.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영이 '두 손을 모아' 꽉 쥐고 있어야 하나, 양손이 떨어져 따로 주먹을 쥐고 있습니다.", "fix_en": "Clasp Su-ri-yeong's two hands tightly together in front of her chest." }, { "issue_ko": "혜수가 캐릭터 레퍼런스 이미지에 있는 짙은 회색 모자를 쓰고 있지 않습니다.", "fix_en": "Add the dark grey cap from the reference image to Hye-su's head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Clasp Su-ri-yeong's two hands tightly together in front of her chest.\n- Add the dark grey cap from the reference image to Hye-su's head.\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." }, "S20sh3::conti_ab_decision": { "fingerprint": "b5a56c2de2ccb602", "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 수리영이 두 사람 뒤에 서서 두 손을 꽉 쥐고 모니터를 응시하는 모습과 배낭, 스마트폰 등 지시된 모든 디테일을 정확히 구현했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "수리영이 두 인물 뒤에 서서 두 손을 모아 꽉 쥔 채 모니터를 응시하는 자세를 정확히 연출했으며, 지정된 가방과 스마트폰 소품도 잘 구현했습니다." } ] } }, "S20sh3": { "input_fingerprint": "d7e49f2155f27a40", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): early morning, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 혜수와 한국인 남자 보안직원 뒤에 서서 두 손을 모아 꽉 쥔 채 굳은 표정으로 모니터를 응시하는 상체.\n\nLOCATION (lock): A supermarket administrative and security office centered on a CCTV monitoring station, with screens and playback equipment for reviewing recorded 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): Su-ri-yeong continues to have her doll-trimmed backpack and smartphone with her while viewing the footage.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 혜수 (한국인 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 머리); 한국인 남자 보안직원 (한국인 성인 남성, 자연스러운 얼굴, 단정한 머리, 짙은색 짧은 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": 9, "C": 4 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "혜수가 누락되었고, 인물들의 배치와 시선이 모니터를 향해 같은 방향을 바라보는 것이 아니라 서로 마주보는 형태로 잘못 연출되었습니다." }, { "label": "B", "score": 9, "verdict_ko": "수리영이 다른 두 인물 뒤에 서서 두 손을 꽉 쥐고 모니터를 응시하는 상체 샷의 요구사항을 위치, 표정, 소품까지 가장 완벽하게 구현했습니다." }, { "label": "C", "score": 4, "verdict_ko": "뒷모습으로 촬영되어 수리영의 굳은 표정을 확인할 수 없으며, 수리영이 다른 인물들 뒤에 서 있다는 위치 설정도 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L15B03.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "수리영이 '두 손을 모아' 꽉 쥐고 있어야 하나, 양손이 떨어져 따로 주먹을 쥐고 있습니다.", "fix_en": "Clasp Su-ri-yeong's two hands tightly together in front of her chest." }, { "issue_ko": "혜수가 캐릭터 레퍼런스 이미지에 있는 짙은 회색 모자를 쓰고 있지 않습니다.", "fix_en": "Add the dark grey cap from the reference image to Hye-su's head." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Clasp Su-ri-yeong's two hands tightly together in front of her chest.\n- Add the dark grey cap from the reference image to Hye-su's head.\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.", "conti_ab": { "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 수리영이 두 사람 뒤에 서서 두 손을 꽉 쥐고 모니터를 응시하는 모습과 배낭, 스마트폰 등 지시된 모든 디테일을 정확히 구현했습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "수리영이 두 인물 뒤에 서서 두 손을 모아 꽉 쥔 채 모니터를 응시하는 자세를 정확히 연출했으며, 지정된 가방과 스마트폰 소품도 잘 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S20sh3__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S20sh3__ab_noconti_sel.png" }, "ref_mode": "플레이트+엔티티 (A/B: 무콘티 승)" }, "S21sh2::ab_noconti": { "input_fingerprint": "a52b6d461f1be8f9", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 김형사가 한 손에 서류철을 든 채 삐딱한 자세로 혜수를 노려보는 상체.\n\nLOCATION (lock): A police station with a general investigators' office, clustered desks, computer monitors, paperwork, and a nearby meeting area. Enlarged crime-scene photographs are spread across one 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): Su-ri-yeong has brought her doll-trimmed backpack and smartphone into the police station.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": 3 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지문에 없는 인물을 완벽히 배제하여 1순위 규정을 지켰으나, 가죽 재킷 누락(4순위)과 두 손으로 서류를 쥔 점(2순위)이 감점 요인입니다." }, { "label": "B", "score": 3, "verdict_ko": "지문에 없는 배경 인물들을 다수 생성했고, 서류철에 임의의 텍스트가 누출되어 심각한 1순위 규정 위반입니다." }, { "label": "C", "score": 3, "verdict_ko": "지문에 언급되지 않은 다수의 인물을 배경에 임의로 추가하여 1순위(발명된 사람) 규정을 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.png" }, { "label": "CHARACTER REFERENCE — 김형사: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지문은 '한 손에 서류철을 든 채'라고 명시했으나, 인물이 두 손으로 서류철을 잡고 있습니다.", "fix_en": "Remove the hand holding the bottom of the file folder so the character holds it with only one hand." }, { "issue_ko": "프롬프트에서 모든 텍스트를 금지했으나, 화면 왼쪽 파티션에 붙어있는 문서에 텍스트가 쓰여 있습니다.", "fix_en": "Erase all text and text-like squiggles from the paper document attached to the partition on the left." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the hand holding the bottom of the file folder so the character holds it with only one hand.\n- Erase all text and text-like squiggles from the paper document attached to the partition on the left.\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." }, "S21sh2::conti_ab_decision": { "fingerprint": "8a8859a46f559b1b", "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 허용된 유일한 인물(김형사)만 등장시키고 지정된 상체 프레이밍을 준수했으나, 후보 1은 금지된 인물이 프레임에 포함되었고 스마트폰을 거꾸로 들고 있는 치명적인 오류가 있습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "후보 2는 전경에 배치된 여성의 왼손에 오른손이 달려 있는 심각한 해부학적 오류가 있으며 프롬프트에 허용되지 않은 인물을 등장시킨 반면, 후보 1은 단독 인물 제한 규정을 정확히 준수하고 해부학적 오류 없이 삐딱한 자세와 책상 위 사진들을 잘 구현함." } ] } }, "S21sh2": { "input_fingerprint": "a52b6d461f1be8f9", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day, office lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 김형사가 한 손에 서류철을 든 채 삐딱한 자세로 혜수를 노려보는 상체.\n\nLOCATION (lock): A police station with a general investigators' office, clustered desks, computer monitors, paperwork, and a nearby meeting area. Enlarged crime-scene photographs are spread across one 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): Su-ri-yeong has brought her doll-trimmed backpack and smartphone into the police station.\n\nPEOPLE: the SHOT TEXT alone decides whether any person is visible in this shot. IF a person appears, they must be one of: 김형사 (한국인 성인 남성, 30대, 자연스러운 얼굴, 짙은색 머리, 짧은 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": 3 }, "selected": "A", "ranking": [ "A", "C", "B" ], "verdicts": [ { "label": "A", "score": 7, "verdict_ko": "지문에 없는 인물을 완벽히 배제하여 1순위 규정을 지켰으나, 가죽 재킷 누락(4순위)과 두 손으로 서류를 쥔 점(2순위)이 감점 요인입니다." }, { "label": "B", "score": 3, "verdict_ko": "지문에 없는 배경 인물들을 다수 생성했고, 서류철에 임의의 텍스트가 누출되어 심각한 1순위 규정 위반입니다." }, { "label": "C", "score": 3, "verdict_ko": "지문에 언급되지 않은 다수의 인물을 배경에 임의로 추가하여 1순위(발명된 사람) 규정을 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B04.png" }, { "label": "CHARACTER REFERENCE — 김형사: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "지문은 '한 손에 서류철을 든 채'라고 명시했으나, 인물이 두 손으로 서류철을 잡고 있습니다.", "fix_en": "Remove the hand holding the bottom of the file folder so the character holds it with only one hand." }, { "issue_ko": "프롬프트에서 모든 텍스트를 금지했으나, 화면 왼쪽 파티션에 붙어있는 문서에 텍스트가 쓰여 있습니다.", "fix_en": "Erase all text and text-like squiggles from the paper document attached to the partition on the left." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Remove the hand holding the bottom of the file folder so the character holds it with only one hand.\n- Erase all text and text-like squiggles from the paper document attached to the partition on the left.\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.", "conti_ab": { "winner": "B", "outer": { "winner": "B", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 2는 허용된 유일한 인물(김형사)만 등장시키고 지정된 상체 프레이밍을 준수했으나, 후보 1은 금지된 인물이 프레임에 포함되었고 스마트폰을 거꾸로 들고 있는 치명적인 오류가 있습니다." }, { "order": "BA", "winner_label": "1", "winner_branch": "B", "reason_ko": "후보 2는 전경에 배치된 여성의 왼손에 오른손이 달려 있는 심각한 해부학적 오류가 있으며 프롬프트에 허용되지 않은 인물을 등장시킨 반면, 후보 1은 단독 인물 제한 규정을 정확히 준수하고 해부학적 오류 없이 삐딱한 자세와 책상 위 사진들을 잘 구현함." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S21sh2__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S21sh2__ab_noconti_sel.png" }, "ref_mode": "플레이트+엔티티 (A/B: 무콘티 승)", "plate_select": { "candidates": { "A": "L14B01", "B": "L14B03", "C": "L14B04", "D": "L14B05" }, "assigned": "L14B01", "choice": "Candidate A", "confident": true, "reason_ko": "김형사가 서류철을 들고 있는 실내 사무실 장면이므로, 경찰서 내부 책상과 캐비닛이 있는 현재 할당된 이미지가 적합합니다.", "kept": "L14B01" } }, "S22sh1::ab_noconti": { "input_fingerprint": "42c1ca06b90700f5", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 맑은 낮 경찰서 건물 앞, 수리영이 자전거 안장에 엉덩이를 뗀 채 한쪽 페달을 강하게 밟아 누른 역동적인 정지 포즈.\n\nLOCATION (lock): The exterior entrance and parking area of a police station, with a bicycle parked nearby and direct access to the street. 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): Su-ri-yeong rides 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "A": 4, "B": 3 }, "selected": "C", "ranking": [ "C", "A", "B" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "안장에서 엉덩이를 떼고 한쪽 페달을 강하게 밟는 역동적인 자세를 지침에 맞게 가장 훌륭히 구현했습니다." }, { "label": "A", "score": 4, "verdict_ko": "자전거 안장에 엉덩이를 붙이고 앉아 있어 '엉덩이를 뗀' 핵심 포즈 지침을 위반했습니다." }, { "label": "B", "score": 3, "verdict_ko": "자전거 핸들바가 허공에 떠 있고 구조가 끊어져 있는 물리적 형태 오류(하드 위반)가 발생했습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B05.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "캐릭터의 두 발이 모두 공중에 떠 있어, '한쪽 페달을 강하게 밟아 누른'이라는 포즈 지시를 명백히 위반했습니다.", "fix_en": "Reposition the character's right foot to rest firmly on the right pedal, pressing it down." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Reposition the character's right foot to rest firmly on the right pedal, pressing it down.\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." }, "S22sh1::conti_ab_decision": { "fingerprint": "844f859a7bf5405a", "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 1은 오른팔이 뒤로 꺾인 물리적으로 불가능한 신체 구조를 보여주는 반면, 후보 2는 안장에서 엉덩이를 떼고 페달을 밟는 역동적인 포즈를 자연스럽고 정확하게 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "첫 번째 후보는 자전거 페달 부근에 다리와 발이 하나 더 생성되는 신체 구조적 오류가 있으므로 탈락입니다." } ] } }, "S22sh1": { "input_fingerprint": "1ee7ed8a27cb9ca9", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): day.\n\nSHOT TEXT (authoritative, Korean): 맑은 낮 경찰서 건물 앞, 수리영이 자전거 안장에 엉덩이를 뗀 채 한쪽 페달을 강하게 밟아 누른 역동적인 정지 포즈.\n\nLOCATION (lock): The exterior entrance and parking area of a police station, with a bicycle parked nearby and direct access to the street. 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): Su-ri-yeong rides 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "A": 3, "B": 8, "C": 6 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "A", "score": 3, "verdict_ko": "자전거 앞바퀴 포크의 물리적 구조 오류와 오른팔 누락이 발생했으며, 명시된 소지품이 제대로 묘사되지 않았습니다." }, { "label": "B", "score": 8, "verdict_ko": "안장에서 일어선 역동적인 포즈와 지정된 장소를 훌륭하게 재현했으며, 메모지, 스마트폰, 배낭 등 모든 소지품의 상태를 정확하게 구현했습니다." }, { "label": "C", "score": 6, "verdict_ko": "요구된 포즈와 장소 묘사는 양호하나 스마트폰이 누락되었고, 불필요하게 뒤를 돌아보는 자세가 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L14B05.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S22sh1.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "레이아웃 스케치에 지정된 포즈와 달리 오른팔이 핸들을 잡고 있으며, 스마트폰이 손이 아닌 백팩 옆면에 비현실적으로 붙어 있습니다.", "fix_en": "Reposition the right arm to reach backward and hold the smartphone in the right hand, exactly matching the pose in the layout sketch." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Reposition the right arm to reach backward and hold the smartphone in the right hand, exactly matching the pose in the layout sketch.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "tie", "verdicts": [ { "order": "AB", "winner_label": "2", "winner_branch": "B", "reason_ko": "후보 1은 오른팔이 뒤로 꺾인 물리적으로 불가능한 신체 구조를 보여주는 반면, 후보 2는 안장에서 엉덩이를 떼고 페달을 밟는 역동적인 포즈를 자연스럽고 정확하게 구현했습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "첫 번째 후보는 자전거 페달 부근에 다리와 발이 하나 더 생성되는 신체 구조적 오류가 있으므로 탈락입니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S22sh1__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S22sh1__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L14B01", "B": "L14B03", "C": "L14B04", "D": "L14B05" }, "assigned": "L14B05", "choice": "Candidate D", "confident": true, "reason_ko": "\"경찰서 건물 앞\"이라는 묘사와 야외 자전거가 있는 풍경이 일치합니다.", "kept": "L14B05" } }, "S25sh8::ab_noconti": { "input_fingerprint": "a6a6f05a483d2316", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 동공이 커진 채 사진을 눈앞에 바짝 들고 굳어버린 얼굴.\n\nLOCATION (lock): A ransacked rooftop apartment with muddy or dusty footprints, an open living-room window, overturned dining chairs, and household items pulled apart. In the adjoining bedroom, a red circle marks the wall around a bloodstained 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): Su-ri-yeong has removed the bloodstained old photograph from the red wall circle and holds it tightly. She wears her backpack containing the printed maps; the apartment remains ransacked behind 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "C": 7, "B": 4, "A": 3 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "C", "score": 7, "verdict_ko": "커진 동공, 눈앞에 바짝 든 사진, 배낭 속 지도 등 지문을 훌륭히 구현했으나, 사진 앞면이 카메라를 향한 방향 오류가 있습니다." }, { "label": "B", "score": 4, "verdict_ko": "레퍼런스 사진의 카메라 구도를 그대로 모방하는 위반이 있으며, 지문과 다른 눈물 흘리는 표정과 지도 누락이 감점 요인입니다." }, { "label": "A", "score": 3, "verdict_ko": "배경 구조가 위치 레퍼런스와 완전히 다르며, 배낭 속 지도가 없고 사진을 가슴 높이로 들고 있어 지문 일치도가 낮습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 두 어린 여자아이가 찍힌 피 묻은 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "사진의 앞면(이미지)이 카메라를 향해 있어 인물이 사진의 뒷면을 보고 있습니다.", "fix_en": "Turn the photograph around so its front image faces the character's eyes, showing only its blank back to the camera." }, { "issue_ko": "인물의 안구가 해부학적으로 부자연스럽게 확대되고 기괴하게 변형되었습니다.", "fix_en": "Restore the character's eyes to anatomically normal proportions and a natural appearance without enlarging the eyeballs." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Turn the photograph around so its front image faces the character's eyes, showing only its blank back to the camera.\n- Restore the character's eyes to anatomically normal proportions and a natural appearance without enlarging the eyeballs.\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." }, "S25sh8::conti_ab_decision": { "fingerprint": "551d59fae1278626", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 사진을 눈앞에 바짝 든 자세와 확장된 동공을 묘사하라는 지문을 정확히 구현한 반면, 후보 2는 사진을 옆으로 들고 있습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 요구된 '피 묻은 사진'의 디테일을 명확히 포함하고 있으며, 사진을 눈앞에 바짝 들고 굳어버린 얼굴이라는 지문 묘사를 더 정확하게 구현했습니다." } ] } }, "S25sh8": { "input_fingerprint": "266fe90c2b16a0f5", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 수리영이 동공이 커진 채 사진을 눈앞에 바짝 들고 굳어버린 얼굴.\n\nLOCATION (lock): A ransacked rooftop apartment with muddy or dusty footprints, an open living-room window, overturned dining chairs, and household items pulled apart. In the adjoining bedroom, a red circle marks the wall around a bloodstained 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): Su-ri-yeong has removed the bloodstained old photograph from the red wall circle and holds it tightly. She wears her backpack containing the printed maps; the apartment remains ransacked behind 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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리) — never anyone else, and never add a person the shot text does not show. IF only part of a person is in frame (a hand, arm, foot, back, silhouette), that body part belongs to the specific person the shot text names — its sex, age, build, skin and grooming must unmistakably match that person's profile above.\n\nNo text, captions, watermarks or annotations anywhere.", "totals": { "B": 7, "C": 5, "A": 4 }, "selected": "B", "ranking": [ "B", "C", "A" ], "verdicts": [ { "label": "B", "score": 7, "verdict_ko": "사진을 캐릭터가 보는 방향(뒷면이 카메라를 향함)으로 올바르게 들고 있으며, 표정과 난장판이 된 공간의 묘사가 지시문과 가장 잘 일치합니다." }, { "label": "C", "score": 5, "verdict_ko": "침실의 붉은 원 위치 등 배경 묘사는 정확하나, 사진의 앞면이 카메라를 향하게 들어 소품 방향 규칙을 위반했습니다." }, { "label": "A", "score": 4, "verdict_ko": "사진이 카메라를 향해 있고 지도를 배낭이 아닌 손에 함께 들고 있어 소품 및 휴대 상태 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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S25sh8.png" }, { "label": "CHARACTER REFERENCE — 수리영: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" }, { "label": "PROP REFERENCE — 두 어린 여자아이가 찍힌 피 묻은 낡은 사진: the exact object appearing in this shot; match its look, material and wear exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "인물의 눈알과 눈동자가 해부학적으로 비정상적으로 크게 부풀려져 있어, 표정을 위해 신체 구조를 변형하지 말라는 지침을 위반했습니다.", "fix_en": "Restore the character's eyes to normal, anatomically correct human size and proportions while keeping her shocked expression." }, { "issue_ko": "프롬프트상 인접한 침실에 있어야 할 붉은색 원이 거실의 TV 위 벽면에 잘못 배치되어 있습니다.", "fix_en": "Remove the red circle from the wall above the television in the background." }, { "issue_ko": "공간 레퍼런스에서 막힌 벽이었던 화면 우측 식탁 뒤 벽면에 창문이 임의로 추가되었습니다.", "fix_en": "Remove the window on the right wall and replace it with a solid, plain wallpapered 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- Restore the character's eyes to normal, anatomically correct human size and proportions while keeping her shocked expression.\n- Remove the red circle from the wall above the television in the background.\n- Remove the window on the right wall and replace it with a solid, plain wallpapered 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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 1은 사진을 눈앞에 바짝 든 자세와 확장된 동공을 묘사하라는 지문을 정확히 구현한 반면, 후보 2는 사진을 옆으로 들고 있습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 2는 요구된 '피 묻은 사진'의 디테일을 명확히 포함하고 있으며, 사진을 눈앞에 바짝 들고 굳어버린 얼굴이라는 지문 묘사를 더 정확하게 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S25sh8__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S25sh8__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B03", "choice": "Candidate B", "confident": true, "reason_ko": "\"사진을 눈앞에 바짝 들고\"라는 지문을 통해, 장소 설명에 명시된 '붉은 동그라미 안의 피 묻은 사진'이 있는 침실 벽면이 이 숏의 배경임을 알 수 있습니다.", "kept": "L04B03" } }, "S27sh2::ab_noconti": { "input_fingerprint": "ab00477745ba9f52", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 고개를 치켜든 채 옆에 선 김형사를 향해 두 눈을 크게 뜬 혜수의 얼굴 클로즈업.\n\nLOCATION (lock): The same rooftop apartment after it has been disturbed, now empty and desolate. The entrance door rattles in the wind while the interior remains without a stated light source. 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 rooftop apartment remains ransacked and empty while Hye-su checks Su-ri-yeong's recorded message.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": "지문이 요구한 인물 배치와 표정, 기준 이미지의 복장과 장소를 가장 잘 구현했으나, 휴대폰 화면이 카메라를 향해 있어 소품 방향 규칙을 위반함." }, { "label": "B", "score": 3, "verdict_ko": "지문에 명시된 김형사가 완전히 누락되었으며, 기준 이미지의 핵심 복장(모자와 외투)도 반영되지 않아 우선순위 위반이 큼." }, { "label": "C", "score": 5, "verdict_ko": "인물의 구도와 표정은 적절하나 모자를 누락했고, 복도 끝 화장실을 외부 출입문으로 왜곡하여 장소 일관성을 훼손함." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "스마트폰 화면이 사용자인 혜수가 아닌 카메라를 향하고 있습니다.", "fix_en": "Rotate the smartphone so its screen faces Hye-su's eyes, showing its back to the camera." }, { "issue_ko": "스마트폰 화면에 텍스트가 표시되어 있습니다 (텍스트 금지 규칙 위반).", "fix_en": "Remove all text from the smartphone screen." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Rotate the smartphone so its screen faces Hye-su's eyes, showing its back to the camera.\n- Remove all text from the smartphone 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." }, "S27sh2::conti_ab_decision": { "fingerprint": "926d7b0842e1ee58", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "1번 후보는 고개를 치켜든 자세를 텍스트에 맞게 정확히 연출했으며, 전화기를 든 손의 해부학적 구조도 자연스럽습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "2번 후보는 창밖으로 지정된 시간대(해질녘)를 정확히 묘사했으며, 고개를 치켜들고 김형사를 바라보는 혜수의 자세와 시선을 텍스트에 맞게 충실히 구현했습니다." } ] } }, "S27sh2": { "input_fingerprint": "058e1ff8f3043896", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, interior lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 고개를 치켜든 채 옆에 선 김형사를 향해 두 눈을 크게 뜬 혜수의 얼굴 클로즈업.\n\nLOCATION (lock): The same rooftop apartment after it has been disturbed, now empty and desolate. The entrance door rattles in the wind while the interior remains without a stated light source. 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 rooftop apartment remains ransacked and empty while Hye-su checks Su-ri-yeong's recorded message.\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 중년 여성, 40대 중반, 성숙한 얼굴, 짙은색 머리, 중간 길이 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": 6, "C": 8 }, "selected": "C", "ranking": [ "C", "B", "A" ], "verdicts": [ { "label": "A", "score": 5, "verdict_ko": "기준 의상을 가장 정확하게 재현했으나, 배경 복도에 원본에 없는 외부 출입문을 임의로 생성하여 공간 구조(우선순위 3)를 위반했습니다." }, { "label": "B", "score": 6, "verdict_ko": "공간 구조와 스마트폰 파지 방향은 적절하나, 고개를 치켜든 포즈가 다소 밋밋하며 기준 의상(목티)을 반영하지 못했습니다." }, { "label": "C", "score": 8, "verdict_ko": "고개를 치켜든 포즈, 정확한 공간 구조(화장실 복도 및 어질러진 바닥), 일몰 조명, 올바른 스마트폰 방향을 모두 잘 구현했으나 의상(목티)이 누락되었습니다." } ], "refs": [ { "label": "LOCATION PHOTOGRAPH — the exact place of this shot: its architecture, materials, fixed features and lighting mood are spatial truth; stage the moment inside this place. Never copy its camera framing.", "path": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L04B01.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S27sh2.png" }, { "label": "CHARACTER REFERENCE — 혜수: the exact person appearing in this shot; match face, hair and build exactly.", "path": "" } ], "critique": { "issues": [ { "issue_ko": "혜수가 쥔 휴대폰의 뒷면(카메라 렌즈 부분)이 그녀의 얼굴을 향하고 있어 기기 방향이 잘못되었습니다.", "fix_en": "Flip the smartphone in the woman's hand so the screen faces her face and the camera lenses face outward." }, { "issue_ko": "카메라를 향해 있는 혜수가 캐릭터 레퍼런스에 지정된 의상(모자, 회색 터틀넥, 검은색 전술 재킷)을 입고 있지 않습니다.", "fix_en": "Dress the front-facing woman exactly as shown in the character reference, including the cap, grey turtleneck, and black field jacket." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Flip the smartphone in the woman's hand so the screen faces her face and the camera lenses face outward.\n- Dress the front-facing woman exactly as shown in the character reference, including the cap, grey turtleneck, and black field 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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "1번 후보는 고개를 치켜든 자세를 텍스트에 맞게 정확히 연출했으며, 전화기를 든 손의 해부학적 구조도 자연스럽습니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "2번 후보는 창밖으로 지정된 시간대(해질녘)를 정확히 묘사했으며, 고개를 치켜들고 김형사를 바라보는 혜수의 자세와 시선을 텍스트에 맞게 충실히 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S27sh2__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S27sh2__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)", "plate_select": { "candidates": { "A": "L04B01", "B": "L04B02", "C": "L04B03", "D": "L04B04", "E": "L04B05", "F": "L04B06", "G": "L04B07" }, "assigned": "L04B03", "choice": "Candidate C", "confident": true, "reason_ko": "클로즈업 샷으로 특정 서브공간을 요구하는 묘사가 없으므로, 현재 배정된 기본 실내 배경을 유지함.", "kept": "L04B03" } }, "S29sh1::ab_noconti": { "input_fingerprint": "9277bd2f52579263", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 좁은 조타실 내부, 조타기를 향해 한 팔을 뻗은 인우와 두 팔을 양옆으로 뻗어 그의 앞을 가로막은 수리영의 정지 자세.\n\nLOCATION (lock): The compact wheelhouse of a small fishing boat, containing the steering controls and navigation area. Numerous circular forms are visible around the cabin, including a newly perceived red circle. 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): Su-ri-yeong retains the old photograph and her map-filled backpack as she blocks In-woo inside the wheelhouse.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 인우 (한국인 젊은 남성, 24세, 젊은 얼굴, 짙은색 머리, 짧은 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": "수리영이 배낭을 메고 두 팔을 벌려 인우를 막아서는 자세와 인우가 조타기를 향해 팔을 뻗는 동작, 소지품이 모두 지시문대로 정확히 연출되었습니다." }, { "label": "B", "score": 4, "verdict_ko": "인물이 마주 서 있을 뿐 수리영이 양팔을 뻗어 가로막는 동작이 없으며, 수리영의 머리 모양이 묶은 긴 머리가 아닌 단발로 묘사되었습니다." }, { "label": "A", "score": 3, "verdict_ko": "화면 우측에 몸과 연결되지 않은 허공에 뜬 손이 나타나는 등 신체 구조에 치명적인 오류(Hard Violation)가 있습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L21B04.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": "조타실 내부의 원형 형태들 중 명시된 '새로 인지된 붉은색 원'이 묘사되지 않았습니다.", "fix_en": "Add a distinct red circle among the circular forms in the wheelhouse." }, { "issue_ko": "조타실 제어판에 금지된 텍스트('지안하' 등)가 생성되었습니다.", "fix_en": "Remove all text and characters from the control panel." }, { "issue_ko": "사진을 들고 인우를 가로막는 수리영의 오른팔이 공간에 비해 비정상적으로 길게 묘사되어 해부학적으로 어긋납니다.", "fix_en": "Adjust Su-ri-yeong's right arm to a natural human length and proportion." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Add a distinct red circle among the circular forms in the wheelhouse.\n- Remove all text and characters from the control panel.\n- Adjust Su-ri-yeong's right arm to a natural human length and proportion.\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." }, "S29sh1::conti_ab_decision": { "fingerprint": "47f61ce8b4412738", "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 2는 수리영의 오른팔 끝에 인우의 재킷 소매를 입은 정체불명의 세 번째 손이 사진을 들고 있는 심각한 신체 구조적 오류(Hard Violation)가 있으므로 후보 1이 우승작입니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 1은 조타기를 잡은 인우의 왼손이 형체 없이 뭉개지는 해부학적 오류(하드 위반)가 발생했으며 사진의 앞면을 카메라를 향해 노출해 소품 방향 규칙을 어긴 반면, 후보 2는 치명적 신체 오류 없이 수리영이 인우의 앞을 가로막는 핵심 자세와 소품의 뒷면 노출을 규칙에 맞게 구현했습니다." } ] } }, "S29sh1": { "input_fingerprint": "298975fdff5ef1f0", "prompt": "Create ONE FINAL photorealistic live-action film still of the moment below — coastal Incheon, South Korea, 2026; all people are Korean unless stated. TIME OF DAY (lock): dusk, wheelhouse lighting unspecified.\n\nSHOT TEXT (authoritative, Korean): 좁은 조타실 내부, 조타기를 향해 한 팔을 뻗은 인우와 두 팔을 양옆으로 뻗어 그의 앞을 가로막은 수리영의 정지 자세.\n\nLOCATION (lock): The compact wheelhouse of a small fishing boat, containing the steering controls and navigation area. Numerous circular forms are visible around the cabin, including a newly perceived red circle. 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): Su-ri-yeong retains the old photograph and her map-filled backpack as she blocks In-woo inside the wheelhouse.\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 젊은 여성, 21세, 타원형 얼굴, 짙은색 긴 머리, 묶은 머리); 인우 (한국인 젊은 남성, 24세, 젊은 얼굴, 짙은색 머리, 짧은 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 지도 배낭과 옛 사진 유지, 인우가 조타기를 향해 팔을 뻗는 핵심 동작을 모두 정확히 구현했습니다." }, { "label": "B", "score": 4, "verdict_ko": "수리영의 옛 사진과 지도가 누락되었으며, 인우가 조타기를 향해 팔을 뻗는 동작이 보이지 않습니다." }, { "label": "C", "score": 4, "verdict_ko": "옛 사진과 지도가 없고, 인우가 조타기가 아닌 다른 곳을 향해 팔을 뻗고 있어 요구사항을 충족하지 못했습니다." } ], "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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/episodes/69af710d-a81f-4102-b000-cf2ef259c91b/images/background_chain/L21B04.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/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/conti/conti_S29sh1.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": "수리영이 들고 있는 사진의 앞면이 카메라를 향하고 있지만, 샷 텍스트에 사진을 보여준다는 지시가 없습니다.", "fix_en": "Flip the photograph in Su-ri-yeong's left hand so its blank back side faces the camera." }, { "issue_ko": "대시보드의 모니터 화면에 텍스트가 생성되어 있습니다 ('텍스트 금지' 규칙 위반).", "fix_en": "Remove all text from the monitor screen on the dashboard." } ] }, "fix_prompt": "Edit this photograph. Apply ONLY the corrections listed below,\neach in place, changing nothing else:\n\nCORRECTIONS:\n- Flip the photograph in Su-ri-yeong's left hand so its blank back side faces the camera.\n- Remove all text from the monitor screen on the dashboard.\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.", "conti_ab": { "winner": "A", "outer": { "winner": "A", "verdicts": [ { "order": "AB", "winner_label": "1", "winner_branch": "A", "reason_ko": "후보 2는 수리영의 오른팔 끝에 인우의 재킷 소매를 입은 정체불명의 세 번째 손이 사진을 들고 있는 심각한 신체 구조적 오류(Hard Violation)가 있으므로 후보 1이 우승작입니다." }, { "order": "BA", "winner_label": "2", "winner_branch": "A", "reason_ko": "후보 1은 조타기를 잡은 인우의 왼손이 형체 없이 뭉개지는 해부학적 오류(하드 위반)가 발생했으며 사진의 앞면을 카메라를 향해 노출해 소품 방향 규칙을 어긴 반면, 후보 2는 치명적 신체 오류 없이 수리영이 인우의 앞을 가로막는 핵심 자세와 소품의 뒷면 노출을 규칙에 맞게 구현했습니다." } ] }, "outer_judged_this_run": true, "sel_conti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S29sh1__ab_conti_sel.png", "sel_noconti": "/Users/manta/Documents/Projects/TheRoad-I1/projects/3d854b4e-d628-4d4e-a68c-c81a9c8fbf00/images/69af710d-a81f-4102-b000-cf2ef259c91b/scene/recipe/S29sh1__ab_noconti_sel.png" }, "ref_mode": "플레이트+콘티+엔티티 (A/B: 콘티 승)" } }