outdoor_canon_photo
outdoor_canon_map
프로젝트 7872eda9 · 64스텝 완주 · 스틸 71/71 · 직행 31(prev 연속성 13) · RCV 0 · 정정 스코핑·Luna 폐지·연속성 fix·팩 v2 반영
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: state-specific character appearance reference (severely_injured): 한국인 성인 남성. 짙은 머리와 평범한 성인 남성의 얼굴형, 자연스러운 이목구비. At dusk in a forest, the character shown in image 3 flees through dense undergrowth with one shoe missing. A deep, large bite mark scars his thigh, with a trail of blood running down from the wound. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as state-variant reference (severely_injured) — match the character's body language for this specific state - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: BACKGROUND from a previous shot (SAME ROOM) — use as-is Reference image 4: state-specific character appearance reference (severely_injured): 한국인 성인 남성. 짙은 머리와 평범한 성인 남성의 얼굴형, 자연스러운 이목구비. At dusk in a forest, amid dense undergrowth along a pursuit path, a black humanoid silhouette overtakes the character shown in image 4 and crashes into his back, folding his upper body forward at the moment of impact. The character shown in image 4 has lost one shoe, and a large bite wound on his thigh is bleeding as the black humanoid silhouette attacks him. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use the background, furniture layout, walls, and lighting from image 3 as-is
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: state-specific character appearance reference (severely_injured): 한국인 성인 여성. 짙은 머리와 평범한 성인 여성의 얼굴형, 자연스러운 이목구비. Reference image 4: character appearance reference: 동아시아계 성인 남성. 짙은 머리와 평범한 얼굴선, 자연스럽고 균형 잡힌 이목구비. At night, beside the utility pole at the encounter point in an urban back alley under a dim streetlight, the character shown in image 3 and the character shown in image 4 are entangled against the pole. The character shown in image 4 has his teeth embedded in the flesh of the character shown in image 3’s neck, tearing it away as labored breathing gives way to violent sucking. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as state-variant reference (severely_injured) — match the character's body language for this specific state -
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: state-specific character appearance reference (dead): 한국인 성인 여성. 짙은 머리와 평범한 성인 여성의 얼굴형, 자연스러운 이목구비. At night in an urban back-alley passage, the character shown in image 4 lies collapsed and completely motionless on the ground beneath a utility pole, her full body visible after a violent attack to her neck. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the attached LOCATION PHOTOGRAPH/SITE PLAN stay th
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Local news footage shows the daytime investigation scene along the 금월산 hiking trail. Police lines cordon off the forest incident site, enclosing forensic teams examining earlier discovery locations and officers collecting two unrecognizable bodies at a designated recovery point. A lead detective directs the assembled investigators from a nearby command position, while an expanded police presence and a temporary investigation headquarters convey a Level A emergency response. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same sta
Photorealistic cinematic close-up. Refine the FIRST reference image into a sharp, detailed close-up. It is a zoomed-in crop of a film frame: keep EVERY element exactly where it is — same poses, same object positions and angles, same surfaces, same lighting. Do NOT move, add, remove or restage anything. Only enhance sharpness, texture and detail lost in the enlargement. Locked visual facts (must remain exactly as in the crop): 혜수 stands among the detectives as the officer directing the investigation.; 혜수 is nearly frontal with a resolute face and a braced command posture.; 혜수’s eyes remain fixed just beyond the lens rather than looking at the audience.; The detective group is clustered behind 혜수, establishing her as its central figure.; The police line remains installed as a lateral boundary at the investigation scene.; Both views are presented as local-news footage of the investigation.
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only In daylight, an isolated rooftop room stands on the open roof of an old multi-household villa, above the narrow alley approach to the building. No people are present. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the scene description explicitly describes. If a characte
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 40대 초반. 짙은 머리와 성숙한 타원형 얼굴, 자연스럽고 부드러운 이목구비. At eye level, the character from Image 1 in apron over homewear, a Korean woman in her early 40s with dark hair, a mature oval face, and soft features, occupies the right two-thirds of the frame at three-quarter angle from the right, her head fixed in a turn away from the switched-off television and her eyes fixed off-screen toward her daughter beyond the right edge. Her guarded, rigid expression is the sharp focal point. The television remains smaller and softly defined in the left background, its unlit front display angled partly toward the camera and showing no image. A narrow trace of subdued natural daylight from a small window sliver at the upper frame edge rests on visible facial surfaces and the apron. Restrained low-to-mid-key illumination, muted natural color, one camera position, no other person visible. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Reference image 2: character appearance reference: 한국인 여성, 40대 초반. 짙은 머리와 성숙한 타원형 얼굴, 자연스럽고 부드러운 이목구비. the character from Image 1 in a gray hoodie, Korean woman in her 20s, occupies the middle-left midground in right profile, her face inclined toward the character from Image 2 and her eyes fixed on the character from Image 2; the character from Image 2 in homewear and apron, Korean woman in her 40s, appears partially in the middle-right midground in left profile, chin tilted toward the character from Image 1 and eyes fixed on the character from Image 1. From a shoulder-height observer position beside the entrance, the character from Image 1's head and shoulders dominate the left half in clear focus, with long black hair, an oval face, and distinct features; the character from Image 2's near-side profile and apron-clad shoulder enter from the right edge in slightly softer focus, with dark hair, a mature oval face, and soft features. Their faces share the middle band with a narrow gap, linked by a single reciprocal line of sight within the entrance space. Subdued natural daylight from the off-frame small window rests softly across their visible facial surfaces, with muted color and restrained contrast. A fixed vertical entrance edge beside t
Photorealistic cinematic still. Reference image 1: BACKGROUND from a previous shot (SAME ROOM) — use as-is Reference image 2: character appearance reference: 한국인 여성, 40대 초반. 짙은 머리와 성숙한 타원형 얼굴, 자연스럽고 부드러운 이목구비. the character from Image 2's marked wrist occupies the middle-center foreground, her hand and forearm crossing the central frame at wrist height, held close to her torso just beyond the frame; the small circular mark is in precise focus with open frame space above it. the character from Image 2, an East Asian woman in her 40s, is represented only by her naturally poised hand and forearm, her body angled away and her off-screen gaze remaining outside the frame. A narrow, softly rendered side edge of the entrance door from the established room background remains behind the forearm. Subdued natural daylight from the established small window rests across the skin and circular mark, with restrained contrast and muted natural color. Generate one image: - use the background, furniture layout, walls, and lighting from image 1 as-is - do NOT copy any standing/moving people from image 1 - from image 1: keep fixed vertical entrance edge under subdued natural daylight - from image 1: ignore Ignore the young woman in the gray hoodie and the apron-clad woman beside her. - use image 2 as character appearance reference — match the person's identity where visible in the scene - do not
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. In daylight on the coastal road beside the sea at the property shown in image 1, the character shown in image 3 wears earphones and breathes in the scent of the sea with a bright, lively demeanor, turning her gaze toward the sound of a barking dog. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. the character from Image 1 in a gray hoodie, a Korean 21-year-old woman with long black hair, an oval face, and clearly defined facial features, faces the camera at eye level with shoulders locked and chin lowered, fully conscious, her centered face filling nearly the entire frame. Her rigid eyes are fixed directly on the off-screen Rottweiler aligned almost exactly with the lens; her firmly set jaw and sustained facial tension are the sharp focal plane. Restrained daytime ambient light gives her visible facial surfaces firm contrast in muted natural tones without a dramatic color cast. Soft oblique portions of kennel enclosure fronts remain at the extreme frame edges, with no other figure or animal visible. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks f
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Reference image 3: POSE + PLACEMENT GUIDE for this shot — line-art mannequins registered over a faint underlay of this room (pose/placement reference only, NOT identity, appearance, style or environment) Photorealistic cinematic still: the character from Image 2 in a gray hoodie, a young Korean woman with long black hair, an oval face, and distinct features, occupies the middle-left midground in right profile beside the occupied dog enclosures in the middle-right background. An eye-level observer camera perpendicular to her sightline preserves her clean profile in the left third, with her upper body inclined toward the existing kennel enclosures across the right half. She holds a smartphone below her chin and looks down at it with a fixed, hardened expression; the phone’s front display faces the camera obliquely and visibly shows an unanswered-call screen, even though its details are small at this wide scale. Her lips are slightly parted and her shoulders are subtly lowered. Focus holds on her profile, the phone, and several confined dogs already present within the established background reference. The enclosure fronts face partly toward th
Photorealistic cinematic still. Reference image 1: PREVIOUS SHOT (SAME MOMENT, zoomed-in reframing) — continuity base Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. From an observer position just below eye level, the character from Image 2 in a gray hoodie, a young Korean woman, occupies the upper half at three-quarter angle from the right, her long black hair framing an oval face with clearly defined features, her upright posture and visibly rigid, fixed and hardened expression held as she gazes downward at the smartphone she holds below her chin. The smartphone sits in the lower foreground, proportionally smaller than her face; its front display faces the camera obliquely, showing a legible unanswered-call screen. Balanced shallow focus favors her facial tension without obscuring the display. Neutral, restrained daytime ambient light reveals both surfaces in muted natural color without a dramatic cast. Only indistinct portions of the shelter enclosure fronts remain softly visible behind her. Generate one image: - use image 1 as the continuity base for a zoomed-in reframing — same moment, same environment, the camera moved closer - keep the environment, lighting, and spatial layout from image 1; render the subjects and their poses as described in the prompt text (for a true zoom they match image 1) - do NOT duplica
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At eye level within a single supermarket aisle, the character from Image 1 in a gray hoodie, a 21-year-old Korean woman, occupies center-left with her face and upper shoulders filling the frame at a three-quarter angle from the left, leaning slightly toward the speaker and looking off-screen; an East Asian female supermarket employee’s blurred shoulder forms a soft edge in the right foreground, both aligned within the same aisle. Long black hair frames the character from Image 1’s oval face and distinct features, her mouth held tight and her facial muscles visibly rigid. Focus rests on her fixed off-screen eyes and tightened mouth. Supermarket display edges remain compressed and recede behind her. Restrained neutral ambient illumination covers her visible face in muted natural tones with controlled contrast, without a prominent visible fixture. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the
Photorealistic cinematic still. No reference images. At eye level, a Korean adult male supermarket employee in a supermarket uniform fills the central third from the chest upward, nearly frontal and facing the camera, with short black hair, ordinary facial contours, and neat features; his eyes are directed toward an off-screen listener beside the camera. Both hands support a stocking box at the shelf edge below shoulder level, its top and one side visible. The aisle-facing edge of the shelving frames one side of him and recedes laterally toward the off-screen listener. Focus is locked on his eyes and mouth, with the box and immediate shelf edge slightly softer. Restrained neutral supermarket ambient illumination gives his visible features controlled contrast without a prominent identifiable light fixture. Generate one image: - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the scene description explicitly describes. If a cha
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. the character from Image 1 in a gray hoodie, a 21-year-old Korean woman, occupies the middle-center foreground from a side observer position, her face and upper shoulders filling the frame. She is at three-quarter angle from the right, alive, shoulders tense, looking off-screen rather than at the camera. Her long black hair frames an oval face with distinct features and visible facial tension. A smartphone rests low at the bottom edge near her lower eyeline, its front display angled toward the camera and showing an unanswered call without invented content. Dusk ambient light rests softly across her face; restrained neon accents create low-to-mid contrast. Defocused exterior sign faces remain behind her as unreadable color accents, with no added text. Focus rests on her tightened expression and the unanswered-call display within the same shallow plane. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night on a coastal road, the character shown in image 3 rides a bicycle hurriedly home, her gaze fixed on the road ahead with an urgent expression. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording -
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only At night, on the open rooftop surrounding the rooftop room above the villa, a rooftop water tank, a clothesline, and a wooden platform holding dried sweet-flag leaves stand at ordinary real-world scale. The clothesline knocks intermittently in the wind. No people are present. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT a
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only At night outside the rooftop room, an unidentified man’s finger presses an advertisement leaflet firmly back into the gap of the iron entrance door. The latch has just disengaged from the lock with a click, leaving the door slightly ajar before he enters. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the attached LOCATION PHOTOGRAPH/SITE PLAN stay the primary source for the property's overall identity and layout - do NOT
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only At night, beside the reflective window of a rooftop room and the surrounding open rooftop, the window reflection shows an East Asian male stranger attacking a woman, his teeth sunk into her shoulder. The iron entrance door through which he entered the rooftop room is nearby. The figures read clearly as reflections, not additional physical bodies. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the sce
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, outside the rooftop room, the character shown in image 3 pauses on the retrofit open-tread steel exterior staircase while hurrying upward, wrinkling her nose with a disgusted expression after suddenly detecting a foul smell. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conf
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, on the rooftop beside the retrofit open-tread steel exterior staircase, the character shown in image 4 has just arrived from the courtyard and pauses in revulsion at a disgusting smell, staring at the rooftop room’s half-open iron entrance door. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the attached LO
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity The existing opened curtain edges of the object shown in image 1 occupy the middle-left foreground, their parted vertical boundaries framing the revealed body. A middle-aged Korean woman is fully visible, collapsed on the floor, seated motionless and lifeless, with her head and gaze lowered near the upper center and her legs resting on the floor below. Her exposed shoulder and collarbone show torn flesh, rough bite-like wounds, dangling tissue, and blood running from the damaged areas. The existing deep-red blood pool beneath and immediately around her body and the existing bloody footprint trail ending at the curtain gap remain consistent with the object shown in image 1. Restrained, dim illumination from the faint interior stand light preserves muted natural colors and deep shadows; the blood remains the principal deep-red accent and emits no light. Generate one image: - use image 1 for this location's wall/floor/ceiling materials, furniture identity, and lighting mood - follow the camera position, framing, and action requested in the prompt text — do NOT copy the camera position or composition of image 1 unless the prompt explicitly asks for it - do NOT copy any people from image 1 - architectural openings (doors, windows, thresholds) must m
Photorealistic cinematic still. No reference images. From the observer's exact eye position beside the body, the camera looks sharply downward at an East Asian adult woman's slack exposed wrist in the center, with part of her limp hand, forearm, and the immediately surrounding floor visible; the wrist occupies less than half the frame. The woman lies motionless and lifeless beyond the focused area, her head outside the frame, oriented away with its sightline off-screen. The wrist tattoo is transformed into a vivid saturated red mark, sharply focused and visibly part of the skin, without emitting light. Extensive deep-red bloodstains and pooled blood cover the floor beneath and around the wrist, subdued enough not to compete with the tattoo. Restrained dim low-key interior illumination, muted natural colors, deep shadows, and shallow depth of field suppress the rest of the body. Generate one image: - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITI
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity The existing bed beneath the circle spans the lower-center midground, with its upper edge crossing the lower frame. The wall bearing the completed red circle fills the upper-center background. The background shown in image 1 retains its established geometry: an old, cramped rooftop dwelling containing a small living area, bedroom, bathroom, and kitchenette, illuminated by a small window and a living-room floor lamp. No person is visible. The completed red circle above the bed is sharply defined against the subdued wall. Restrained, dim interior illumination, muted natural colors, and deep shadows surround the clearly visible red mark. The red circle is non-luminous and casts no light. Generate one image: - use image 1 for this location's wall/floor/ceiling materials, furniture identity, and lighting mood - follow the camera position, framing, and action requested in the prompt text — do NOT copy the camera position or composition of image 1 unless the prompt explicitly asks for it - do NOT copy any people from image 1 - architectural openings (doors, windows, thresholds) must match image 1 — same position, type, and material; do NOT invent additional doors or windows unless the prompt text explicitly asks for them - if the prompt text places th
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, outside the rooftop room, the character shown in image 3 leans out through the open iron entrance door, screaming as she flees toward the retrofit open-tread steel exterior staircase while a siren sounds. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the refer
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, the character shown in image 4 stands dazed and unfocused in the villa courtyard in front of the rooftop-room building, beside the retrofit open-tread steel exterior staircase and beneath the rooftop room. A siren is sounding around the old multi-household villa exterior. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consisten
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. At night in the Villa Courtyard, beneath the rooftop room and the retrofitted open-tread steel exterior staircase, the character shown in image 4 freezes in shock upon hearing that her mother is dead, her face stunned in the moment before she runs toward the rooftop room. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the
Photorealistic cinematic still. Reference image 1: BACKGROUND from a previous shot (SAME ROOM) — use as-is Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At eye level inside the established room background, the character from Image 2 in a gray hoodie, young East Asian woman, occupies a slightly off-center position with her upper body visible. Her long black hair frames an oval face with clearly defined features; her face is pale, her shoulders drawn inward, her upper body contracted, and her living body held in a mid-turn at three-quarter left, torso angled toward the established curtain area, eyes looking off-screen toward the bed side of the room. Neutral, natural-looking nighttime ambient illumination supports a restrained muted palette and preserves realistic facial and room detail without a conspicuous new light source. Behind her, the established curtain area reveals empty space with no body; the room floor is excessively clean and orderly with no bloodstains, the clean wall above the established bed has no red mark, and the bed remains neatly arranged. The established backpack and doll remain visible in their existing background positions. Camera focus rests on her pale face and contracted shoulders, with the empty curtain area and unmarked bed wall legible in the same frame. Generate one image: - use the bac
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 남성, 30대. 짧은 검은 머리와 평범한 성인 남성의 얼굴선, 단정한 이목구비. the character from Image 1 in a worn leather jacket, a Korean man in his 30s with short black hair, ordinary facial contours, and neat features, fills most of the center-right frame at a three-quarter angle from the left, his upper body inclined forward in a reprimanding posture, his sharp expression and eyes fixed on the young woman at lower left. A soft partial over-shoulder sliver of a young Korean woman occupies the lower-left edge, her shoulders drawn inward into a contracted upper-body posture, her visible gaze directed down. Both figures occupy the space beside a clean dining table; only its near edge and a small portion of its top remain softly visible behind them. Shallow depth of field isolates the character from Image 1's tense facial detail and the compressed line of sight between the two figures. Restrained natural-looking ambient illumination rests across the visible facial surfaces, with low-to-mid-key contrast and a muted dark palette appropriate to the nighttime interior. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in
Photorealistic cinematic still. Reference image 1: BACKGROUND from a previous shot (SAME ROOM) — use as-is The exact established room remains unchanged. An everyday fabric backpack with a front compartment, shoulder straps, and a small hanging plush-doll ornament is positioned in the lower-center midground, occupying less than one-third of the frame. The doll is in its established position, and the meticulous surrounding order reveals its existing slight positional mismatch. The backpack remains visible immediately behind or beside the doll, with its outward-facing side visible. The orderly bed anchors the background, with its side and part of its top surface visible. Sharp focus emphasizes the doll’s slight positional mismatch, while the backpack and bed provide scale and alignment cues. The uncluttered room appears excessively clean and fully restored, with no visible bloodstains anywhere and no red markings on the bed. Restrained, natural-looking nighttime ambient illumination creates a muted palette, low-to-mid-key contrast, and realistic object detail without any conspicuous new light source or motivated color. Generate one image: - use the background, furniture layout, walls, and lighting from image 1 as-is - do NOT copy any standing/moving people from image 1 - from image 1: keep excessively clean and orderly room floor - from image 1: keep neatly arranged bed in its
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, the character shown in image 3 sits facing the sea in the coastal bus stop waiting area, dazed and motionless, holding a phone whose screen shows an unanswered call to “강민숙(민숙).” Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the ref
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, the character shown in image 3 sits in the coastal bus stop waiting area facing the sea. Holding her unanswered phone, she stares blankly toward the water, struggling to maintain a calm expression as she denies the possibility that her mother has been in an accident and questions whether what she saw was real. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that c
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. Reference image 5: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At night, the character shown in image 5 runs away along the coastal road from the bus stop, seen from behind in a running stride. Near the sea-facing bus stop waiting area, the character shown in image 4 watches her leave and calls after her, telling her to have her mother phone if she sees her. The sea remains visible from the bus stop. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into t
Photorealistic cinematic still. No reference images. At hand height just inside the office entrance, a modern touchscreen mobile phone with a glossy black screen and a thin rectangular body occupies the lower-center foreground, clearly filling less than half the frame. Its hand-sized, slim body rests in one hand, with the active display facing forward. The contact name 강민숙 and a contemporary unanswered outgoing-call interface are sharply legible on the screen, with no portrait, person, location, or surveillance image displayed. Focus remains on the display and unanswered-call status; the supporting fingertips are crisp, while only the interior edge of the entrance and indistinct forms of the police office are softly visible around the phone. Restrained low-to-mid-key nighttime office lighting keeps the hand, phone, and immediate interior legible without any conspicuous new light source or added color cast. Generate one image: - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. At eye level from across the duty desk, the character from Image 1 in a dark navy blouson, a Korean woman in her mid-40s, fills the frame from the upper torso upward, her face slightly left of center at a three-quarter angle from the left. She is leaning across the duty desk, dark hair framing a mature oval face with neat features, her eyes fixed on the off-screen detective and her lips set around a firm syllable. One hand holds a smartphone near the desk line; its screen is angled toward the off-screen detective rather than directly toward the camera, with an oblique visible display bearing a saved contact number. The near desk edge crosses the bottom of the frame, and its receding top plane forms a physical boundary between her and the unseen detective. Shallow depth of field holds her resolute gaze and the supporting phone legible, with the immediate surroundings softened. Restrained natural office illumination rests across visible surfaces of her face and blouson, preserving firm low-to-mid-key contrast and muted neutral color without a conspicuous source or added color cast. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or composit
Photorealistic cinematic still. Reference image 1: BACKGROUND from a previous shot (SAME ROOM) — use as-is Reference image 2: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. Reference image 3: character appearance reference: 한국인 남성, 30대. 짧은 검은 머리와 평범한 성인 남성의 얼굴선, 단정한 이목구비. The background shown in image 1 occupies the middle-center midground, with a duty desk, computers, documents, partitioned desks, and case files filling the police workspace beneath fluorescent lights. The character shown in image 2 occupies the middle-left midground, in right profile with both palms braced on the desk and her gaze fixed on the character shown in image 3. The character shown in image 3 occupies the middle-right midground, in left profile, seated back from the desk with his gaze fixed on the character shown in image 2. The desk’s near end and long receding edge form a central perspective line between their opposing profiles. Restrained nighttime office illumination keeps both characters and the desk surface clearly visible without a conspicuous new light source or color cast. A muted dark palette and low-to-mid-key contrast preserve depth throughout the office. Generate one image: - use the background, furniture layout, walls, and lighting from image 1 as-is - do NOT copy any standing/moving people from image 1 - from image 1: keep duty desk with its near ed
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only At daybreak, the light remains on inside the rooftop room above the villa. On the open rooftop, a black cat crosses the rooftop boundary wall while glancing back toward the illuminated room, crying out at dawn. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the attached LOCATION PHOTOGRAPH/SITE PLAN stay the primary source for the property's overall identity and layout - do NOT copy lighting or time of day from image 3 —
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only At daybreak on an open villa rooftop, a black cat pauses beside the rooftop boundary wall, shown in profile as it looks up at the illuminated rooftop room before crossing the wall. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the attached LOCATION PHOTOGRAPH/SITE PLAN stay the primary source for the property's overall identity and layout - do NOT copy lighting or time of day from image 3 — the shot text is the sole time
Photorealistic cinematic still. Reference image 1: BACKGROUND from a previous shot (SAME ROOM) — use as-is Reference image 2: state-specific character appearance reference (dead): 한국인 여성, 40대 초반. 짙은 머리와 성숙한 타원형 얼굴, 자연스럽고 부드러운 이목구비. The character shown in image 2 sits motionless and lifeless in the lower center of the frame, turned three-quarters toward the right, with her head dropped forward and her eyes closed. Her seated body, the existing pool of blood, and the blood spatter below form a single descending visual path. The background shown in image 1 remains unchanged, with the existing curtain hanging on both sides of the body as a constricted foreground border. Subdued soft ambient illumination, restrained contrast, desaturated colors, and dark tactile realism fill the space without a visible light source. Generate one image: - use the background, furniture layout, walls, and lighting from image 1 as-is - do NOT copy any standing/moving people from image 1 - from image 1: keep curtain edges framing the recessed floor area - from image 1: keep deep-red blood pool on the floor inside the curtain gap - from image 1: keep blood spatter below the curtained recess - use image 2 as state-variant reference (dead) — match the character's body language for this specific state - do not copy poses or compositions from reference images - do not alter character identities where their
Photorealistic cinematic still. No reference images. The camera occupies the unseen observer's exact eye position, square to the bathroom mirror. A bathroom mirror frame borders the image, enclosing a smooth glossy reflective plane with a crisp, undistorted, left-right-reversed view of the bathroom wall. The reflected bathroom wall fills the center, with a single anomalous red circle fixed on that wall inside the mirror image. No person, face, or additional reflection is visible. Optical focus is locked on the red circle and the reflected wall plane, with the mirror frame legible at the margins. Subdued existing bathroom ambience keeps the visible surfaces dim and restrained in color, leaving the red circle as the sole saturated deep-red accent. Generate one image: - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the scene description explicitly describes. If a character's face is visible in the scene, match it to the refere
Photorealistic cinematic still. Reference image 1: object appearance reference: 동아시아계 40대 여성과 20대 초반 딸이 함께 밝게 웃는 모습을 담은 가정용 컬러사진. The object shown in image 1 lies flat on a tabletop, centered across the lower third with enough surrounding surface visible to preserve its scale. Within the printed portrait, only the adult woman’s face is sharply distorted, vertically crushed into a downward-running smear across the photographic surface, while the daughter’s face remains intact and clearly readable beside it. Emphasis rests on the tactile photograph and the contrast between the malformed maternal face and the intact daughter’s face. Subdued ambient illumination covers the photograph and tabletop, with restrained saturation and no wider room visible. Generate one image: - include the object shown in image 1 - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the scene description explicitly describes. If a character's face is visib
Photorealistic cinematic still. Reference image 1: PREVIOUS SHOT (SAME MOMENT, zoomed-in reframing) — continuity base Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At her eye level from the off-screen speaker's side, the character from Image 2 in a gray hoodie, a young Korean woman, holds a three-quarter-left angle with her face occupying the left-center of the frame. Her upper torso is inclined forward in a questioning posture, lips parted, eyes widened and looking off-screen just past the lens. Long black hair frames an oval face with clearly defined features; the camera focus rests sharply on her widened eyes and immediate facial reaction. A soft, minimally visible store passage borders the frame behind her, without any other person or distinct object. Subdued early-morning store illumination, muted natural tones, restrained facial contrast, shallow depth of field, naturalistic tension. Generate one image: - use image 1 as the continuity base for a zoomed-in reframing — same moment, same environment, the camera moved closer - keep the environment, lighting, and spatial layout from image 1; render the subjects and their poses as described in the prompt text (for a true zoom they match image 1) - do NOT duplicate body parts - render the focused region (hand, wrist, face area, object, etc.) enlarged, consistent wit
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity Photorealistic cinematic still: the store entrance coverage area occupies the middle-center background beneath a wall-mounted CCTV camera in the upper-right background. From outside at chest height, the observer camera is offset from the mounted unit’s axis and tilted upward; the entrance door remains visible below and slightly offset, both fixed within the same exterior wall plane. The CCTV camera’s lens-facing end angles toward the entrance area rather than toward the observer. Its orientation makes the likely field of coverage spatially legible without a surveillance-feed viewpoint or graphic overlay. Preserve the attached background reference exactly for the existing exterior layout and fixtures. Focus sharply on the CCTV camera and the geometric relationship between its aim and the entrance coverage area. Soft natural early-morning light rests on visible exterior surfaces, with muted tones and restrained contrast. Generate one image: - use image 1 for this location's wall/floor/ceiling materials, furniture identity, and lighting mood - follow the camera position, framing, and action requested in the prompt text — do NOT copy the camera position or composition of image 1 unless the prompt explicitly asks for it - do NOT copy any people from
Photorealistic cinematic still. Reference image 1: PREVIOUS SHOT (SAME MOMENT, zoomed-in reframing) — continuity base Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Outside the store, the character shown in image 2 faces forward, looking off-screen with a tense brow and lips parted mid-sentence. A hand-sized, slim modern touchscreen smartphone with a black glass screen and thin rectangular body is pressed to one ear against her cheek, its screen facing inward. A softly blurred store entrance is visible behind her. Subdued natural early-morning light and muted tones create restrained low-to-mid-key contrast. Generate one image: - use image 1 as the continuity base for a zoomed-in reframing — same moment, same environment, the camera moved closer - keep the environment, lighting, and spatial layout from image 1; render the subjects and their poses as described in the prompt text (for a true zoom they match image 1) - do NOT duplicate body parts - render the focused region (hand, wrist, face area, object, etc.) enlarged, consistent with the frame context of image 1 - from image 1: keep store entrance edge and adjoining exterior wall - from image 1: keep soft natural early-morning illumination on the storefront - from image 1: ignore Ignore the wall-mounted CCTV camera. - use image 2 as character appearance reference —
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Reference image 3: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. The character shown in image 2 stands at the left foreground edge with her back facing outward, positioned behind the seated pair and looking down toward the CCTV display. At the same workstation, an adult security employee in a security uniform sits in left profile, with dark hair, ordinary facial contours, and neat features, their ethnicity and nationality visually indiscernible. One hand rests on the controls, and their gaze is directed downward. The character shown in image 3 remains seated in the existing chair to the right at a three-quarter angle, leaning forward and looking down toward the monitor. The existing desk supports the shared workstation, arranging the two seated viewers and the monitor in a triangular formation beneath the shoulder line of the character shown in image 2. The CCTV monitor is angled toward the seated pair, its active playback screen clearly visible. It displays an exceptionally dark, crushed, low-detail section of footage with heavily muted tonal separation, where two apparent human figures remain visible on
Photorealistic cinematic still. Reference image 1: PREVIOUS SHOT (SAME MOMENT, zoomed-in reframing) — continuity base The existing CCTV monitor from the attached background reference faces the camera squarely at seated eye level, its displayed image filling nearly the entire frame, with only a narrow trace of the monitor's front rim at the edge. The monitor is playing the same specific section of footage, exceptionally dark, crushed, and too low-detail for identification. Within this indistinct CCTV space, two apparent adult human figures remain present as vague forms oriented toward each other, with neither figure identifiable and no discernible facial features, bodily detail, clothing, or identity cues. Heavily muted tonal separation reduces the image to an abstract low-detail field with slight digital texture and reduced resolution. No viewers or surrounding room furnishings are visible. Generate one image: - use image 1 as the continuity base for a zoomed-in reframing — same moment, same environment, the camera moved closer - keep the environment, lighting, and spatial layout from image 1; render the subjects and their poses as described in the prompt text (for a true zoom they match image 1) - do NOT duplicate body parts - render the focused region (hand, wrist, face area, object, etc.) enlarged, consistent with the frame context of image 1 - from image 1: keep CCTV mon
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. the character from Image 1 in a dark navy blouson, Korean, mid-40s, sits at eye level in a three-quarter angle from the left, leaning forward in her chair; her forward-set face occupies the center-right, with dark hair, a mature oval face, neat features, slightly parted lips, and level brows. Her eyes look off-screen past the camera toward the security employee. The camera focus rests on her set mouth and steady brows. [L##: A restrained strip of the workstation crosses the lower edge at an oblique angle immediately in front of her.] Beside her, the CCTV monitor is only partly visible, its playback face turned away from the camera, with the unusually dark section remaining indistinct. Subdued neutral interior ambient illumination rests across the visible surfaces, with restrained contrast and controlled definition on her face. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the
Photorealistic cinematic still. Reference image 1: BACKGROUND/ENVIRONMENT from a previous shot at the same location Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At face height and slightly to the side of her seated position, the character from Image 2 in a gray hoodie, a 21-year-old Korean woman with long black hair, an oval face, and defined features, fills the frame at a three-quarter angle from the right, perched forward in the chair and looking down. Sharp focus rests on her pale face, widened eyes, slightly parted lips, and fixed downward attention. The upper edge of the desk enters the lower foreground beneath scattered victim photographs; their image-bearing fronts face upward and partly toward the camera, with cropped views of severe injuries across the neck, shoulder, and facial areas of unidentified Korean adult victims, softly out of focus. Subdued neutral ambient illumination rests across visible surfaces, with muted natural tones, low-to-mid-key brightness, restrained contrast, and no visible light source. Generate one image: - use ONLY the lighting, color palette, and environment mood from image 1 - do NOT copy characters, people, or their appearances from image 1 - do NOT copy the composition or character poses from image 1 - use image 2 as character appearance reference — match the person's identit
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity The monitor already present in the background reference fills the frame from display height, its active front face square to the camera, with a narrow slice of bezel retained around the image. Focus is locked on the central coordinate readout “37.484400 N / 126.072486 E” over a sea map; compressed peripheral interface information reads “Missing Kang Min-sook (41),” “LAST LOCATION,” “Timestamp 2026-02-18 11:21:08,” “Source: GPS / Network,” and “Status: offline.” Subdued neutral office ambience and restrained low-to-mid-key contrast keep the information legible, with no additional visible light source. Slight digital texture and modest screen-resolution character remain visible. Generate one image: - use image 1 for this location's wall/floor/ceiling materials, furniture identity, and lighting mood - follow the camera position, framing, and action requested in the prompt text — do NOT copy the camera position or composition of image 1 unless the prompt explicitly asks for it - do NOT copy any people from image 1 - architectural openings (doors, windows, thresholds) must match image 1 — same position, type, and material; do NOT invent additional doors or windows unless the prompt text explicitly asks for them - if the prompt text places the camera
Photorealistic cinematic still. Reference image 1: BACKGROUND from a previous shot (SAME ROOM) — use as-is Reference image 2: object appearance reference: 흰 종이에 북위·동경 수치가 손글씨로 적힌 작은 메모지. 접어서 주머니에 넣을 수 있는 크기다. One hand presses the folded object shown in image 2 partway into an open backpack pocket beside the desk. Only its folded outer face is visible, with the written coordinates concealed inside the folds. A partial view of a 21-year-old Korean woman in a gray hoodie appears at the right side, leaning over the backpack and looking off-screen. Part of her torso and long black hair provide scale, while her face and lower body remain outside the frame. The pocket opening angles toward her hand and is clearly visible. Only the near edge of the existing desk crosses the upper portion beside the backpack. Preserve the same police-office background shown in image 1 under subdued neutral ambient light, low-to-mid-key illumination, muted natural color, and restrained contrast, with no additional visible light source. The fingertips, the object shown in image 2, and the pocket opening are sharply focused, while the surrounding torso and office details remain secondary. Generate one image: - use the background, furniture layout, walls, and lighting from image 1 as-is - do NOT copy any standing/moving people from image 1 - from image 1: keep the same desk edge beside the seated work ar
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Outside the police station in daylight, the character shown in image 3 rides away from the parked-bicycle starting position, urgently staring in the direction of travel while seated on the bicycle. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follo
Photorealistic cinematic still. No reference images. The marked target point on the map occupies the middle-center midground. A square-on digital map fills the frame without perspective skew, its coastline, surrounding sea, entered numerical coordinates, route markings, and plotted destination clearly legible. Static focus rests sharply on the destination marker and the adjacent coastal contours. Slight digital pixel texture and restrained neutral tonal rendering preserve map clarity under even, subdued illumination with no conspicuous directional light source; no device frame or screen border visible. Generate one image: - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the scene description explicitly describes. If a character's face is visible in the scene, match it to the reference. If only a body part is shown, do NOT add the face. - CRITICAL: Output exactly ONE single continuous photograph of ONE moment — never a panel
Photorealistic cinematic still. Reference image 1: object appearance reference: 인천 연안과 서해 해역, 접근 경로가 표시된 여러 장의 인쇄 지도. 좌표와 해안선, 방향 정보가 보인다. The object shown in image 1 fills the center and upper frame as several gathered, overlapping printed map sheets, all unfolded with their printed faces turned upward. Distinct coastlines, route lines, map markings, and multiple coastal approach routes remain readable across the different sheets. An open backpack occupies a smaller area along the lower edge, its open compartment facing upward beneath the pages. Only the immediate workstation surface appears around the frame edges, unobtrusive and free of additional objects. Sharp focus emphasizes the alternate coastal approach routes spread across the object shown in image 1 rather than any single sheet. Even, subdued neutral interior illumination maintains restrained contrast and clear printed detail without a conspicuous directional source, within a restrained dark palette. Generate one image: - include the object shown in image 1 - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only A coastal edge beside a riding route and seaside bicycle corridor is engulfed by a fierce downpour beneath dense storm clouds. Thunder and lightning accompany the heavy rain as a young woman rides a bicycle through the storm with urgent determination, focused solely on finding her mother. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is descri
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Along the property’s coastal riding route, the character shown in image 3 rides a bicycle through heavy rain, her face set with unwavering determination to find her mother. Dark storm clouds loom overhead as thunder rolls and lightning flashes across the coastal landscape. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; i
Photorealistic cinematic still. Reference image 1: PREVIOUS SHOT (SAME MOMENT, zoomed-in reframing) — continuity base Reference image 2: object appearance reference: 1970~80년대 색감의 낡은 사진. 시골 식당과 지붕의 둥근 구조물, 앞에 앉은 동아시아계 12세 여자아이 둘이 담겼다. — Keep the prop at the consistent real-world scale of a small handheld photographic print relative to hands, walls, furniture, and the surrounding environment. A handheld, rectangular, worn glossy color photographic print with faded yellow-green tones, softened corners, light scratches, small surface stains, crumpling creases, and visible bloodstaining. This is the SAME single physical object in every shot it appears in — match this reference's printed content, crease pattern, stains and wear EXACTLY. The object shown in image 2 occupies the center midground, fixed precisely at the center of a complete deep-red circle on the bedroom wall. The entire circle remains visible around the smaller object, with a narrow perimeter of wall surrounding it. Its image-bearing face is oriented directly toward the viewer and remains clearly legible, with the printed scene and blood marks in sharp focus. Fading natural dusk light casts subdued, restrained contrast across the dim neutral wall, making the deep-red circle and blood accents selectively distinct. Generate one image: - use image 1 as the continuity base for a zoomed-in reframing — same moment, same
Photorealistic cinematic close-up. Refine the FIRST reference image into a sharp, detailed close-up. It is a zoomed-in crop of a film frame: keep EVERY element exactly where it is — same poses, same object positions and angles, same surfaces, same lighting. Do NOT move, add, remove or restage anything. Only enhance sharpness, texture and detail lost in the enlargement. Locked visual facts (must remain exactly as in the crop): The same old, blood-stained photograph is shown front-facing in both shots.; The photograph depicts two twelve-year-old girls seated together in front of a rural restaurant.; Young Kang Min-sook sits at a three-quarter-left angle beside the other girl and looks off-screen.; The second twelve-year-old girl is in right profile with her knees drawn close and her gaze lowered.; The depicted restaurant has a round installation mounted on its roof and visible above the roofline.; The restaurant is depicted as a rural restaurant from approximately the 1970s–1980s, with its front facade facing the photographed viewpoint. Object references: the additional labeled reference images show the true appearance of specific objects present in the crop. For each such object, keep its position, orientation, scale and physical condition exactly as in the crop, but its printed or displayed content (picture, text, screen) must match the object reference image — the enlargement
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity Reference image 2: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Reference image 3: object appearance reference: 1970~80년대 색감의 낡은 사진. 시골 식당과 지붕의 둥근 구조물, 앞에 앉은 동아시아계 12세 여자아이 둘이 담겼다. — Keep the prop at the consistent real-world scale of a small handheld photographic print relative to hands, walls, furniture, and the surrounding environment. A handheld, rectangular, worn glossy color photographic print with faded yellow-green tones, softened corners, light scratches, small surface stains, crumpling creases, and visible bloodstaining. This is the SAME single physical object in every shot it appears in — match this reference's printed content, crease pattern, stains and wear EXACTLY. The character shown in image 2 stands at the center in a natural upright posture, mid-stride toward the entrance door, with her gaze directed off-screen beyond the doorway. She clenches the object shown in image 3 close to her torso with one hand, its small rectangular image-bearing face partly turned toward the viewer. The rooftop-apartment interior remains unchanged, with the sides and undersides of toppled dining chairs at the periphery and an open window visible deeper in the room. Fading natural dusk light fills the room wi
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Near sunset, after a rain shower, dense fog fills an otherwise empty harbor. A single boat cuts through the harbor water toward the boat-accessible landing edge, its bow pointed toward the harbor. No people or other boats are visible. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only r
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 남성, 24세. 짙은 머리와 매끈한 얼굴선, 균형 잡힌 수려한 이목구비를 지닌 청년. At dusk, at the harbor landing edge beside the shore-to-boat walking approach and mooring connection point, the character shown in image 4 stands after securing the single boat’s mooring line and gestures permissively toward the boat, signaling someone off-frame to come aboard. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with i
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. At dusk, the character shown in image 4 sits on the rocky shore facing the open sea, recording a voice message for her aunt saying that she has somewhere to go and asking her not to worry. She struggles to keep her expression composed as she looks out over the water. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - where this shot overlaps image 3's view, keep the same stable visible details (openings, fixed fittings, signage, local object placement) consistent with it - the
Photorealistic cinematic still. Reference image 1: BACKGROUND/ENVIRONMENT from a previous shot at the same location Reference image 2: character appearance reference: 한국인 여성, 40대 중반. 짙은 머리와 성숙한 타원형 얼굴, 선명하고 단정한 이목구비. Photorealistic cinematic still featuring the character from Image 2 in a dark navy blouson, a Korean woman in her 40s, in the middle-center foreground with her face weighted toward center-right. Directly observed at eye level, she is turned at a three-quarter angle from the right, alive and alert, her shoulders hunched over the mobile phone, chin slightly raised, and eyes fixed off-screen. Her dark hair, mature oval face, neat facial features, parted lips, and tightened brow fill most of the frame; sharp focus rests on the anxious expression and lifted gaze. Only a narrow portion of the mobile phone appears along the lower edge, its illuminated screen facing her, with the voice-message playback interface visible to the camera at an oblique angle. Minimal out-of-focus rooftop-room interior remains behind her under restrained low-to-mid-key dusk ambience. Generate one image: - use ONLY the lighting, color palette, and environment mood from image 1 - do NOT copy characters, people, or their appearances from image 1 - do NOT copy the composition or character poses from image 1 - use image 2 as character appearance reference — match the person's identity and outfit w
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: object appearance reference: 1970~80년대 색감의 낡은 사진. 시골 식당과 지붕의 둥근 구조물, 앞에 앉은 동아시아계 12세 여자아이 둘이 담겼다. — Keep the prop at the consistent real-world scale of a small handheld photographic print relative to hands, walls, furniture, and the surrounding environment. A handheld, rectangular, worn glossy color photographic print with faded yellow-green tones, softened corners, light scratches, small surface stains, crumpling creases, and visible bloodstaining. This is the SAME single physical object in every shot it appears in — match this reference's printed content, crease pattern, stains and wear EXACTLY. At dusk on the exposed fishing-boat deck, the object shown in image 3 is held out between two people. One hand points to the small word “Geumwol” visible within the photograph. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: character appearance reference: 한국인 남성, 24세. 짙은 머리와 매끈한 얼굴선, 균형 잡힌 수려한 이목구비를 지닌 청년. At dusk on the exposed fishing-boat deck by the map-exchange position, the character shown in image 3 looks visibly flustered, his wavering gaze betraying recognition of the place depicted in a photograph bearing the small word “Geumwol,” even as he denies knowing it. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - use image 3 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clo
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only Reference image 3: PREVIOUS STILL at the same property — continuity source for stable visible details only Reference image 4: character appearance reference: 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Reference image 5: character appearance reference: 한국인 남성, 24세. 짙은 머리와 매끈한 얼굴선, 균형 잡힌 수려한 이목구비를 지닌 청년. At dusk inside the fishing-boat wheelhouse, the character shown in image 5 stands before the steering controls, urgently attempting to turn the bow. The character shown in image 4 has followed him inside and blocks him from operating the controls. The adjoining deck, wheelhouse access, and fishing-boat bow remain visible in the fading evening light. Generate one image: - image 1 is the PRIMARY source of how this place looks (building, materials, aging, colours, surroundings); the previous-still reference refines stable visible details in overlapping areas - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or
Photorealistic cinematic still. Reference image 1: character appearance reference: 한국인 남성, 24세. 짙은 머리와 매끈한 얼굴선, 균형 잡힌 수려한 이목구비를 지닌 청년. Photorealistic cinematic still: the character from Image 1 in fisher workwear, a young East Asian man, occupies the middle-center foreground. Seen directly across from his eye height at a three-quarter-left angle, his face fills most of the frame, his head held abruptly and unnaturally still, alive and upright. His rigid, scrutinizing eyes sit just above center, fixed on an off-screen Korean young woman. His pallid complexion, dark hair, smooth facial lines, and refined features remain sharply defined, with the camera focus concentrated on his frozen gaze. Subdued dusk illumination rests across his visible facial surfaces in restrained low-to-mid-key contrast and muted colors. The compressed wheelhouse interior falls softly out of focus, with faint circular forms visible behind one shoulder. Generate one image: - use image 1 as character appearance reference — match the person's identity and outfit where visible in the scene - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the refere
Photorealistic cinematic still. No reference images. The red circle occupies the middle-right midground, its visible face presenting one deep-red circular mark directly toward the unseen observer’s viewpoint. Focus locked on the red circle; multiple surrounding circular forms retain clean round shapes in a subtly rotational peripheral pattern, their spacing optically skewed across the unstable visual field. Interior frame edges of a small fishing-boat wheelhouse at dusk, muted surroundings under subdued illumination, peripheral geometry subtly distorted around recurring circular forms. Restrained low-to-mid-key dusk contrast, muted surrounding colors, and the deep-red circle as the only forceful color accent; no person visible. Generate one image: - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the scene description explicitly describes. If a character's face is visible in the scene, match it to the reference. If only a bod
Photorealistic cinematic still. Reference image 1: pre-rendered BACKGROUND environment reference — layout/material/lighting identity Reference image 2: state-specific character appearance reference (unconscious): 한국인 여성, 21세. 긴 검은 머리와 자연스러운 타원형 얼굴, 또렷한 이목구비를 지닌 젊은 성인. Reference image 3: character appearance reference: 한국인 남성, 24세. 짙은 머리와 매끈한 얼굴선, 균형 잡힌 수려한 이목구비를 지닌 청년. the character from Image 2 in a gray hoodie, a young Korean woman, occupies the middle-left foreground in right profile, unconscious with eyes closed, slack and limp, her upper body lying diagonally across the shoulder of the character from Image 3; the character from Image 3 in fisher workwear, a young Korean man, occupies the middle-right midground at a three-quarter angle from the left, torso rigid under her weight, eyes fixed on the character from Image 2 with a rigid, scrutinizing gaze. Her long black hair falls beside an oval face with clearly defined features; his dark hair, smooth facial contours, and handsome defined features remain visible. A shoulder-height viewpoint beside the pair angles slightly downward, framing their overlapping upper bodies and excluding their lower bodies. Both figures occupy the same referenced wheelhouse interior through the shared shoulder contact. The existing helm console and steering wheel remain peripheral and unattended, with their operator-facing sides angled away fro
Photorealistic cinematic still. Reference image 1: LOCATION PHOTOGRAPH — the real filming property this shot happens at Reference image 2: SITE-PLAN DRAWING of the same property — layout source only At sea at sunset, a small fishing boat slowly enters a dense fog bank, with part of its bow crossing the sharp boundary of the fog. The setting sun casts a golden reflection across the open water. No people are visible. Generate one image: - image 1 is the SOLE source of how this place looks: building, materials, aging, colours, surroundings - do NOT copy image 1's camera angle, composition, time of day or lighting — the prompt text defines this shot's moment - do NOT copy any people from image 1 - use image 2 ONLY to understand where things are on the property (positions, adjacency, routes) - NEVER draw image 2 itself or its colours, circles, markers or labels into the output - do not copy poses or compositions from reference images - do not alter character identities where their face is visible in the scene - each character's clothing/outfit must match that character's appearance reference image EXACTLY; if any clothing wording in the text conflicts with the reference image, follow the reference image and ignore that wording - only render what the scene description asks for — if only a hand or wrist is described, do NOT add the character's face - CRITICAL: Only render what the
직행 스틸 합계: 31 / 71