ЧТО ПОНАДОБИТСЯ для генерации трендового танца нейросетью

- Аккаунт в Aigrator

- Нейросеть для видео: Seedance 2.5.

- Четыре файла: фото первого героя (танцора), фото второго героя (того, кто подражает), фото локации и видео с танцем, который нужно повторить (ссылка на него).

ШАГ 1. СОБИРАЕМ МАТЕРИАЛЫ

Сохраните видео-референс танца, подберите фото обоих героев и снимок места, где всё будет происходить. Держите четыре файла под рукой, дальше их нужно будет загрузить в строгом порядке.

ШАГ 2. ЗАПУСКАЕМ ГЕНЕРАЦИЮ

1. Откройте в Aigrator AI раздел «Видео» и выберите модель Seedance 2.5.

2. Задайте качество 720p. Чтобы потратить меньше, подойдёт 480p, а если хочется максимальной чёткости, поставьте 1080p.

3. Выберите формат кадра 9:16 и длительность 14 секунд.

4. Загрузите файлы в правильном порядке:

видео-референс

затем три фото -

сначала танцор

потом локация

потом подражатель.

От порядка зависит, какому файлу достанется номер в промте: видео получит @video1, а фото - @image1 (танцор), @image2 (локация) и @image3 (подражатель).

КАК ПРИВЯЗАТЬ ФАЙЛЫ К ПРОМТУ

В Aigrator AI при вставке промта всё привяжется само, ничего делать не нужно. Если вы запускаете генерацию в другом месте, действуйте вручную: в тексте промта удалите все метки @image1, @image2, @image3 и @video1, а затем впишите их заново. Когда вы набираете символ «@», появляется список загруженных файлов - выберите нужный, и метка привяжется к файлу.

ЧТО ВАЖНО ПОМНИТЬ

- Танцор заменяет исходного героя целиком и повторяет его движения один в один.

- Локация заменяет всё окружение из референса: от улицы до фона.

- Подражатель - отдельный человек со своей внешностью, а не копия танцора.

ПРОМТ ДЛЯ ВИДЕО

TASK — CHARACTER AND LOCATION REPLACEMENT, WITH ONE ADDED CHARACTER Use @video1 as the primary reference for choreography, timing, body motion, camera behaviour and editing. Use @image1 for the dancer's identity, @image2 for the entire new location, and @image3 for one added character who is not present in @video1. @video1 is one single continuous handheld shot, no cuts, about fourteen seconds long, vertical 9:16 frame. CHARACTER MAPPING — ABSOLUTE: @image1 — THE DANCER. Replaces the dancing man from @video1 completely and performs his exact choreography. @image3 — THE IMITATOR. A NEW character added to the scene who does not exist in @video1. He stands next to the dancer, watches him, points at him, tries to copy him and keeps failing. @image2 — THE LOCATION. Replaces the entire environment of @video1. THE DANCER — @image1 defines his complete identity: face, hair, facial hair, glasses, skin, body build, clothing, footwear and accessories. Keep him exactly as in @image1 and invent nothing. Do not transfer the original dancer's white shirt, black vest, black tie, black trousers, black shoes or yellow rubber gloves unless those items genuinely belong to the person in @image1. There is only one dancer in the video, and he is the same person in every frame. THE IMITATOR — @image3 defines his complete identity: face, hair, skin, body build, clothing, footwear and accessories, unchanged and invented nothing. He is a different individual from the dancer, never a copy of him, never a twin, never blended with him. His clothes and look come only from @image3. THE LOCATION — @image2 defines the entire environment: rebuild the whole place exactly as shown in @image2, including its surfaces, ground, walls, architecture, background structures, street furniture, colours, light sources and atmosphere. Every shot is set entirely in the world of @image2. Nothing from the original Paris street survives: no Haussmann buildings, no wrought-iron balconies, no street bonfire, no black smoke, no scattered debris, no transit shelter, no Parisian signage. The location stays spatially consistent and identical across the whole shot, with a believable depth behind the dancer. BACKGROUND PEOPLE — the crowd from @video1 is kept, but they are now ordinary passers-by who belong to the location in @image2: they stand in the middle distance, walk past, glance at the two dancing men and carry on. They wear normal clothing that fits the new location. They are secondary, softly out of focus, never in the foreground, never speaking and never looking into the lens. Never copy the original crowd's outfits, hoodies, face coverings or yellow accents. THE DANCER'S PERFORMANCE — reproduce the choreography of @video1 exactly, beat for beat, with the same energy, the same footwork and the same timing: 00:00–00:02 — he stands centred, looks into the lens, and mimes knocking on an invisible door in the air with one fist while lip-syncing. 00:02–00:05 — he hops forward toward the camera with both arms spread wide and an exaggerated questioning face. 00:05–00:06 — he holds the pose, shoulders bouncing to the beat, still mouthing the words. 00:06–00:08 — he spins in place, kicks one leg up, and mimes another knock above his head while lip-syncing. 00:08–00:11 — he drops into a low bouncy crouch and swings his arms rhythmically side to side. 00:11–00:14 — he shuffles backwards and sideways toward the right of the frame, gesturing with both hands, then does a half turn away and looks back over his shoulder at the lens. His mouth must form the same syllables at the same moments as the original vocal, and his expressions stay playful, confident and exaggerated. Do not simplify, slow down or re-choreograph anything. THE IMITATOR'S PERFORMANCE — this is entirely new performance invented for this video, running in parallel with the dancer: 00:00–00:02 — he stands a little behind and to the side of the dancer, watching him intently, then points at him with one hand as if telling the viewer to look at this. 00:02–00:05 — he steps closer and tries to copy the knock gesture. He does it wrong: he knocks at his own chest instead of the air, looks confused and corrects himself too late. 00:05–00:08 — he tries to copy the spin and the leg kick. He is off the beat, arms flailing, and his kick barely lifts. 00:08–00:11 — he attempts the low bouncy crouch. His knees give out, his feet tangle under him, and he falls over ACCIDENTALLY — a real clumsy slip, arms out, landing on his backside or his side on the ground. Not a planned stunt, not a funny deliberate fall: an embarrassing, weightless, accidental trip. He scrambles back up, embarrassed, dusts himself off, glances at the camera and grins awkwardly. 00:11–00:14 — he tries again, badly: he shuffles the wrong way, out of sync, knocks into his own feet and wobbles, arms waving, while the dancer carries on flawlessly next to him. He never succeeds. His dancing is always clumsy, off-beat and awkward. He stays beside the dancer the whole time, sometimes directly beside him, sometimes half a step in front or behind, always in the same frame, always watching the dancer and glancing at the camera. Both men are always visible in the frame and never block each other: the dancer stays the clear protagonist in the centre, and the imitator moves around him without covering his face or his body for more than a moment. CAMERA — preserve the camera of @video1 exactly: one handheld smartphone shot in vertical 9:16, standard wide-angle phone lens, held at chest to eye level at a medium full-body distance, with the same subtle organic handheld sway and small tracking adjustments following the dancer. No cuts, no zoom, no stabilisation, no cinematic framing, no drone moves, no slow motion. The camera follows the dancer's position in exactly the same way as in the source. PHYSICAL REALITY — both men keep believable contact with the ground, with real weight, real foot placement and real balance. The imitator's fall has real physics: his feet slip, his centre of gravity goes, his body drops hard, and his clothes and hair react. No floating feet, no sliding, no weightless falls, no hands or limbs passing through each other or through objects. No extra fingers, no melting hands, no duplicated limbs, no warped faces. AUDIO — preserve the original audio of @video1 exactly: the same French novelty pop track with the same vocal, at the same timing, plus the same faint background street murmur and crowd noise. The imitator never speaks, never adds words, no voice-over, no added sound effects, no added music. Nothing is re-recorded. STYLE — realistic, natural digital smartphone footage with a natural colour grade and clean, believable skin and fabric texture. No film grain, no cinematic LUT, no plastic skin, no beauty smoothing, no CGI look. CONTINUITY RULES - The dancer is always the person from @image1 and the imitator is always the person from @image3: two distinct identities that never swap, never merge and never transfer features. - The location is always the location from @image2 and never morphs, never changes between moments and never reverts to the original street. - No one from the original video survives in the frame except the background passers-by. - No original costumes, no yellow gloves, no black vest, no Parisian architecture, no bonfire, no smoke. - Exactly two main characters. The background crowd is secondary and never competes with them. - No additional characters. No animals. No text, captions, subtitles, watermarks or logos anywhere in frame. - No identity drift, no morphing, no flicker, no limb duplication, no missing fingers. - No changes to the original camera choreography, no additional cuts, no transitions. FINAL RESULT Produce a seamless fourteen-second vertical 9:16 single-take video in which the original dancer has been completely replaced by the person from @image1 performing the identical choreography, the entire environment has been replaced by the location from @image2, and a second character from @image3 has been added beside him, watching, pointing, trying to copy him badly and falling over accidentally before getting back up and failing again. The result must look as though it was genuinely filmed in the new location with these two people, with real weight, natural skin and seamless integration. PRIORITY: EXACT PRESERVATION OF THE DANCER'S CHOREOGRAPHY. COMPLETE REPLACEMENT OF THE DANCER AND THE LOCATION. ONE ADDED CLUMSY IMITATOR. ZERO REMNANTS OF THE ORIGINAL PERFORMER OR THE ORIGINAL STREET.

Gods Of Steel (Альбом Love Kills 2026 года)

We Are Fire (Новый сингл 2026 года)

Thriller (Кавер Michael Jackson 2026 года)