Создаем сцену из человека паука с вашими героями.

Нужен аккаунт в Syntx.ai (https://syntx.ai/welcome/mEQ2WKuT)
*Подписка PRO и выше.
Нейросети: Gpt Image 2 для фото, Seedance 2.5 для видео.
Нужно: 2 фотки героев + видео-исходник (1 часть и 2 часть)

Take the person's appearance from the uploaded photos and create a character reference sheet, divided into three vertical panels on a neutral white or light-gray background.

RIGHT PANEL (close-up):
A close-up portrait of the person's face, cropped at the shoulders (bust shot). Face front-facing, looking directly at the camera, neutral expression, natural even lighting. Preserve all facial features and details as accurately as possible: eye shape, nose, lips, eyebrows, face shape, eye color, skin tone and texture, hair shape and color, as well as any unique features (moles, freckles, ear shape). High detail, realism, portrait quality.

LEFT PANEL (back view, full body):
The person standing at full height, front view, in their everyday clothing, arms along the body, feet shoulder-width apart. IMPORTANT: the head and face are completely absent — the figure is cropped at the neck/shoulder line, no head. Only the figure from the neck down is shown. Preserve body build, proportions, and clothing.

MIDDLE PANEL (front view, full body):
The same person, back view, in the same clothing, at full height, including the head and hair (viewed from behind). Preserve the hairstyle, body build, and clothing details.

GENERAL REQUIREMENTS:
Unified style, consistent lighting and figure scale across all panels, consistency of appearance and clothing between panels. Clean background, clear vertical dividers between panels. Realistic photographic style, high resolution.

Для начала скачиваем 1-ую часть ролика: ССЫЛКА

[Video Editing Task] Edit @video1. Replace only the two main characters using @image1 and @image2, including their clothing. Preserve the original video rather than generating a new interpretation. Keep the exact source duration, square aspect ratio, shot order, cuts, camera angles, framing, action timing, original slow-motion segments, and audio. [Character Mapping] @image1 → PERSON_A: The person originally wearing the colorful comic-print zip-up hoodie, initially seated at the cafeteria table, who catches the falling person and the tray. @image2 → PERSON_B: The person originally wearing a white T-shirt and blue jeans, with tousled light hair, who walks through the cafeteria carrying the food tray and slips. These assignments stay fixed throughout the entire clip. Never swap or blend the two identities. [Replace PERSON_A] Replace PERSON_A's appearance with the person in @image1. Preserve the reference facial structure, skin tone, hairline, hairstyle, facial hair if present, and distinctive visible features. Replace the original comic-print hoodie and other visible garments with the clothing shown in @image1. Match the reference garment colors, patterns, cut, and materials. For clothing areas not visible in the photograph, use a simple consistent continuation of the visible outfit. Preserve PERSON_A's exact source performance: seated posture, noticing the approaching person, head turns, supporting the falling person, catching the tray and food, and the final interaction and speech movements. [Replace PERSON_B] Replace PERSON_B's appearance with the person in @image2. Preserve the reference facial structure, skin tone, hairline, hairstyle, facial hair if present, and distinctive visible features. Replace the original white T-shirt and jeans with the clothing shown in @image2. Match the reference garment colors, patterns, cut, and materials. For clothing areas not visible in the photograph, use a simple consistent continuation of the visible outfit. Preserve PERSON_B's exact source performance: walking with the tray, stepping onto the spill, slipping, losing balance, dropping the tray, being caught, and looking at PERSON_A at the end. [Preserve the Original Choreography] Do not change either character's screen position, movement path, pose timing, gaze direction, or physical interaction. Keep the original body blocking and contact points so the supporting arm, falling body, hands, and tray remain correctly connected. Adapt the replacement clothing naturally to the existing movements, with believable folds and fabric deformation. Preserve the exact timing and trajectories of the airborne food and tray, including every original slow-motion moment. Keep all original reaction shots, close-ups, floor inserts, and the final shot of the two characters together with the caught tray. Do not invent new gestures, expressions, dialogue, falls, or catching actions. Transfer each source facial expression and mouth movement to the corresponding replacement face; do not freeze the expression from the reference photograph. [Consistency and Occlusion] Apply both replacements consistently in every shot, including profiles, rear views, partially hidden faces, and motion-blurred frames. Preserve correct foreground occlusion: hands, arms, the other person, the tray, and flying food must obscure the replacement face or clothing exactly when they do in the source. Match the original lighting, shadows, perspective, depth of field, grain, and motion blur. Blend faces naturally around the hairline, ears, jaw, and neck. Maintain stable identities and outfits across cuts. No face flicker, identity drift, original faces reappearing, duplicated limbs, fused hands, floating clothing, or visible mask edges. [Unchanged Elements] Leave all background people unchanged. Preserve the cafeteria, tables, chairs, floor, spill, food, tray, and every other prop. Preserve the original audio unchanged, including dialogue, ambience, music if present, and synchronization. Do not add or remove shots, change playback speed, crop, reframe, or alter camera movement. No new text, captions, logos, or watermarks. Final result: the same cafeteria clip, with PERSON_A replaced by the appearance and clothing from @image1, and PERSON_B replaced by the appearance and clothing from @image2. Everything else remains as in @video1.

Теперь скачиваем 2-ую часть ролика: ССЫЛКА

[Video Editing Task] Edit @video1 using @image1 as the identity reference for the recurring main character. Replace only the main character's facial identity and hair throughout the entire clip. Preserve the source body, clothing, actions, facial performances, locations, props, editing, and audio. Keep the exact source duration, approximately 8 seconds, original 1:1 square aspect ratio, and every original cut. Do not rebuild the montage from scratch. [Identity Reference] @image1 = the replacement identity. Match the person's recognizable facial structure, skin tone, eye color, hairline, hairstyle, and facial hair if present. Do not copy the photograph's clothing, background, pose, lighting, or static expression. Fit the replacement identity naturally to the source head position, perspective, and body proportions. Preserve the source head's approximate scale within the frame; no oversized head or altered body silhouette. [TARGET — Recurring Main Character Only] Replace the principal character wearing the dark blazer, light shirt, and diagonally striped tie in ALL eight shots. Do not replace other people, including the foreground person in the dark-room shot or the background person near the doorway. [Shot Coverage — Follow the Existing Source Cuts] 1. OUTDOORS NEAR THE FENCE AND TRACTOR: Replace the main character's face and hair during the quick backward head turn. Preserve the original alert expression, gaze, outdoor lighting, and motion blur. 2. DARK INTERIOR WITH ANOTHER PERSON IN THE FOREGROUND: Replace the main character's face and hair while preserving the smile and head movement. Keep the foreground person's head unchanged and preserve its original overlap with the main character. 3. BRIGHT ROOM WITH A BLACK MUG: Replace the main character's face and hair. Preserve the original smirk, gaze, mug, hand placement, and window lighting. 4. CORRIDOR, LOOKING BACK OVER THE SHOULDER: Replace the face and hair throughout the turn. Preserve the original shoulder position, walking movement, pleased smile, and camera framing. 5. DOORWAY: Replace the main character's face and hair during the half-turn and smile. Keep the background person, doorway, and all movement unchanged. 6. DARK PROFILE WITH A GLASS CONTAINER: Replace the main character's facial identity and hair in profile, respecting the source silhouette and very low exposure. Keep the replacement naturally dark. Do not brighten the face or invent frontal facial detail in a silhouette shot. Preserve the glass container and hands. 7. GLASS DOOR OR BLINDS: Replace the main character's face and hair during the smiling shoulder-dance movement. Preserve the exact rhythm, body motion, outfit, and background. 8. SEATED AT THE DESK: Replace the main character's face and hair. Preserve the original pointing gesture toward the camera, mischievous squint, smile, desk, props, and final-frame timing. [Preserve Clothing and Body] Keep ALL source garments exactly as shown in each shot: dark blazer, light shirt, diagonally striped tie, trousers where visible, and any existing accessories. Do not transfer any clothing from @image1. Do not change the source torso, limbs, hands, body proportions, posture, or movement. Only blend the replacement head naturally into the existing neck and collar boundary. [Preserve the Source Performance] Transfer each shot's original expression onto the replacement identity: smiles, smirks, squints, alert looks, and gaze changes. Preserve original head rotations, eye directions, blinking, mouth movements, and action timing. Do not impose a single expression across all shots. Do not make the character look toward the camera unless they do so in the source. [Temporal Consistency and Compositing] Maintain the same replacement identity across every cut, angle, and lighting condition. No original face reappearing during turns or partially obscured frames. Match the source lighting, shadows, exposure, color grade, sharpness, grain, depth of field, and motion blur. Preserve realistic skin texture. Respect all original occlusions by foreground people, objects, doorframes, and hands. Keep clean, natural boundaries around the hair, ears, jaw, and neck. No face flicker, identity mixing, doubled features, floating hair, mask edges, stretched faces, or sudden head-size changes. [Unchanged Elements] Preserve the exact shot order, cut points, playback speed, camera movements, composition, locations, and props. Preserve the original audio track unchanged and synchronized. No new scenes, additional gestures, dialogue, music, captions, text, logos, or watermarks. Final result: the same eight-shot source montage with the recurring main character's identity replaced by @image1, while the original clothing, body, performances, editing, and sound remain unchanged.