Парное видео Сжигаем мосты по фото с нейросетью

Seedance 2.5С вашим фотоФормат 9:16
1. Загрузите фотоЛицо хорошо видно, без очков
2. Нажмите «Создать»Цена — на кнопке
3. ГотовоВидео — в «Моих генерациях»
Создать такое же видео

Промпт

TASK & FORMAT Generate a photorealistic identity reskin of @video1, retaining its full 15.033-second duration, 9:16 vertical aspect ratio and original multi-shot sequence. Preserve the scene layout and all source motion; replace acting identities and wardrobe only. Exactly two acting subjects. REFERENCES @video1 = sole motion and camera guide: choreography, timing, trajectories, hand/prop contacts, spatial blocking, shot sequence, head motion and expression timing. Its uncolored MatCap surfaces and facial contours are guides only, never final materials or identity. @image1 = identity and wardrobe for the rear actor in @video1, initially left of the foreground carrier, raising and lowering both hands at the opposite bridge entrance. This same subject appears at the LEFT cliff in the wide burning-bridge shot and remains across the gap in later views. Preserve this action/position binding through every cut and occlusion. @image2 = identity and wardrobe for the foreground bouquet-and-jerrycan carrier in @video1, initially center-right, walking toward the camera with white roses in the right hand and a handled can in the left. This same subject crosses onto the near rock, leaves the props, performs the ignition-related gesture, turns back and later brushes hair away. Preserve this binding throughout; do not assign a new identity when the activity changes. Each image controls only its subject's identity and wardrobe, not location, pose, illumination or props. Each image is the sole authority for that subject's entire facial identity, apparent age, eyes and eyelids, nose, lips, teeth, jaw, cheeks, skin detail, hair and facial hair. Neither the source performer nor the generic proxy may contribute facial anatomy or likeness. Retarget performance to each reference subject's own proportions. SCENE Keep the narrow sagging wooden suspension bridge between two sheer rock cliffs over a deep fog-filled chasm. Retain dark damp wooden planks, thick rope handrails, vertical suspension supports, wet lichen-marked rock ledges and scattered white flower petals. Restore realistic scene materials, not clay rendering. The subject from @image2 carries a recognizable inverted bouquet of white roses with green leaves and long stems and a rectangular handled jerrycan with a circular opening. A thin liquid stream trails beneath the tilted can onto the deck and rock; its exact composition is unknown, so depict liquid consistent with the observed ignition sequence without specifying a fuel or brand. Preserve the small handheld ignition item and its function without inventing its precise design. ACTION / TIMELINE 0.000–2.633s: the subject from @image2 walks toward the frontal camera along the bridge, gripping the bouquet stems in the right hand and the jerrycan handle in the left. The subject from @image1 stays at the rear entrance, lowers both hands and raises them again. Keep the foreground approach and exit toward the right edge. 2.633–5.267s: preserve the source cut to the close rear-following view, present by 2.867s. The subject from @image2 continues across the last planks with the can on screen-left and bouquet on screen-right. The thin liquid stream reaches the deck and rock. Preserve alternating foot plants and the step onto the near ledge. 5.267–7.033s: preserve the cut back to a frontal ledge view, present by 5.733s. The subject from @image1 is still distant across the bridge. The subject from @image2 lowers and releases the bouquet onto the LEFT foreground rock and leaves the can on the RIGHT rock. Preserve the actual release and settling motion in the guide, not invented exact contact times. 7.033–8.800s: the subject from @image2 looks at the small ignition item in the right hand, turns back toward the bridge and performs the source outward release gesture. Preserve the small flame trajectory and first localized deck flame seen by 8.800s. Do not embellish the uncertain mechanism into a detailed lighting procedure. 8.800–10.567s: preserve the cut to the wide side view, present by 9.233s. Orange flames extend along the suspended wooden deck and dark smoke rises. The subject from @image1 remains at the LEFT cliff entrance, not a new actor. Keep the bridge's sag, rail geometry and the same cliff spacing. 10.567–12.333s: preserve the cut back to the near ledge, present by 11.167s. The subject from @image2 looks back toward the fire, then turns forward as burning bridge elements drop into the chasm. Reproduce the guide's destruction timing and downward debris paths. Do not add a jump, fall or rescue involving either subject. 12.333–15.033s: the subject from @image2 faces the camera; visible flames diminish below the ledge, leaving smoke and occasional embers. A slight closed-mouth smile develops. The left hand brushes hair away from the forehead and moves toward the side of the head. Adapt the contact to the reference subject's own hair while preserving arm and head timing. The bouquet stays left and the can right on the near rock. The subject from @image1 remains across the gap, faintly visible where the guide shows that actor. These are sampled action windows; use @video1 for actual transitions and edit points rather than treating sample boundaries as exact events. CAMERA Preserve all source framing and motion: frontal bridge approach, close rear tracking of the carrier's lower body and props, frontal near-ledge view, very wide side view of the burning span, then frontal near-ledge view again. Match source cuts, scale, perspective and handheld movement. No added shots, orbit, zoom or duration compression. FACIAL PERFORMANCE For the subject from @image1, preserve the distant forward-facing head orientation and the observable mouth opening in the opening and raised-hand samples; do not fabricate fine brow or eye detail. For the subject from @image2, retain forward gaze and mostly closed lips with slightly raised mouth corners during 0.000–2.633s; forward gaze and closed lips at 5.733–6.167s; the downward glance toward the right hand and subsequent look back at the bridge during 7.033–8.800s; the head turn back to front and parted lips during 11.167–12.333s; then closed lips, a slight smile and forward gaze during 13.233–15.033s. Hair may briefly occlude the reference subject's own face where source motion indicates. Preserve mouth/eye performance timing from the full guide without inventing speech or exact blink timestamps. Identity images must not freeze expressions into their static pose. LIGHTING / MATERIALS / PHYSICS Use cool diffuse foggy daylight, natural wet wood and rock response, realistic flower petals and a solid handled can. Add the observed warm fire illumination to nearby planks, rock and reference wardrobe as the flames grow, with rising dark smoke and gravity-driven burning debris. The carried props swing with arm motion, hands maintain proper grip, liquid falls downward, and released props settle on the ledge. The bouquet and can do not turn into generic clay blobs. Keep both actors out of direct flame contact where the guide does. No extra explosion or invented total disappearance of the rope system. AUDIO Proposed output only: minimal synchronous close-miked physical Foley. Use brief soft foot contacts during visible walking from 0.000–5.267s, switching from wood to rock at the actual step-off. In the 6.167–7.933s release window, one subdued flower/leaf rustle for bouquet contact and one short can-on-rock contact, each synchronized to the actual guide event. During 9.233–12.333s, use only a few isolated dry fire crackles and a short structural snap or falling-debris sound where a break is visible. Let each event decay naturally and return to near silence. No continuous fire roar, wind, environmental ambience, music, dialogue, bubbling or gurgling. Do not sonify uncertain small-item mechanisms with invented clicks. CONTINUITY LOCKS / CONSTRAINTS Maintain exactly the two distinct reference subjects; no merging, swapping, duplication or added acting characters. @image1 stays bound to the rear entrance/left-cliff actor; @image2 stays bound to the bouquet-and-can carrier and near-ledge actor. Each reference alone determines its subject's full facial identity and wardrobe in every shot. Source faces and MatCap eye/lip contours supply performance timing only, never anatomy, makeup or drawn markings. No source wardrobe transfer, identity blending, face drift or reference-background substitution. Preserve scene geometry, prop functions, liquid trail, ignition consequence, burning and collapse sequence. No extra tools or prop reference slots, no text overlays or watermarks, no invented dialogue, contacts or frame-accurate events between samples. FACE IDENTITY LOCK — REFERENCE PHOTOS ONLY Each subject bound to @image1, @image2 must retain the complete facial identity of its own assigned reference photo. The photo alone defines eye and eyelid shape, nose, lip shape, teeth, jaw and cheek structure, skin detail, hair and facial hair. Use @video1 only for head movement, expression timing and mouth/eye actions, retargeted to the photo subject’s own anatomy. Do not copy the video performer’s facial features or likeness, blend the two faces, or treat neutral guide contours as facial identity, makeup or tattoos. When anatomy conflicts, the assigned reference photo always takes priority. Preserve changing expressions without freezing the photo’s expression.

О шаблоне

Два героя оказываются в напряжённой кинематографичной сцене о решении оставить прошлое позади: движения, смена планов и взаимодействие повторяют драматичный сюжет ролика. Загрузите фото своё и близкого человека — нейросеть перенесёт ваши лица в видео и сохранит динамику сцен и эмоции.

Вопросы

Сколько стоит?

Как обычная генерация видео в Seedance 2.5 на нашем сайте — точная цена показана на кнопке перед запуском. Если нейросеть не справилась, нейроны вернутся автоматически.

Нужно ли своё фото?

Да — загрузите одно чёткое фото лица анфас, без очков и фильтров. Нейросеть сохранит черты.

Можно изменить промпт?

Да. Промпт подставится в поле генератора — поменяйте одежду, фон, надпись или цвет.

Где найти результат?

В разделе «Мои генерации»: оттуда можно скачать фото или сделать ещё вариант.

Похожие шаблоны

Все «ИИ видео» →
Создать такое же