Как повторить это трендовое видео с переодеванием
В этом видео две большие руки перед камерой сами переодевают девушку: надевают на неё свитер, юбку, обувь и аксессуары.
Всё можно повторить внутри GIVON AI.
- Фото человека, которого хочешь использовать в видео.
- Фото одежды и аксессуаров, которые хочешь на него надеть.
Дальше делаем всего два этапа: сначала создаём стартовый кадр, потом оживляем его.
Зайди в GIVON AI → Image → Nano Banana Pro.
Загрузи своё фото и вставь промпт.
Use the uploaded photo as the exact identity reference for the person.
Create a photorealistic vertical 9:16 starting frame.
The same person stands full-body in front of a large floor mirror inside a bright modern room.
Preserve the exact face, hair, body shape and proportions from the reference.
Exactly TWO female hands with natural manicured nails are already visible in the extreme foreground.
One hand enters from the left foreground.
One hand enters from the right foreground.
The hands are very close to the camera lens and therefore look large because of forced perspective.
The person stands calmly in the center of the frame and looks toward the camera.
Vertical 9:16, natural daylight, realistic smartphone footage, candid, lived-in, slight imperfections, mild grain, smartphone realism.
Тебе понадобится ещё одно изображение, где хорошо видно весь образ, который ты хочешь надеть на человека.
Лучше всего использовать фото, где отдельно и понятно видны:
Зайди в GIVON AI → видео → Seedance 2.0.
@Image1 — готовый стартовый кадр с руками.
@Image2 — фото человека в полный рост и крупно лицо.
@Image3 — фото одежды.
15 секунд · 9:16 · Seedance 2.0
И вставь готовый промпт на видео, который я дам следующим блоком.
same character throughout the entire shot.
Use @Image1 as the exact starting frame, exact room, exact camera position and exact composition.
Use @Image2 as the exact identity reference for the young woman: preserve her face, long wavy brunette hair, body shape and proportions.
Use @Image3 as the exact outfit reference: cream-white openwork knitted sweater, pink floral skirt, pink ballet flats, brown sunglasses and cream shoulder bag.
ONE CONTINUOUS 15-SECOND SHOT.
CRITICAL ACTION RULE:
THE WOMAN NEVER DRESSES HERSELF.
THE TWO LARGE FOREGROUND FEMALE HANDS DRESS HER FROM BEGINNING TO END.
The two large female hands already visible in @Image1 are the main active subjects of the entire video.
They must remain visible and active from the first second.
The woman is passive and only cooperates with necessary body movements.
Her own hands never touch, grab, pull, hold or adjust any clothing or accessories.
[0s–1s]
Start immediately. No waiting and no idle pose.
The same two foreground hands from @Image1 instantly hold the exact cream-white openwork knitted sweater from @Image3 and lift it toward the woman.
[1s–5s]
The two foreground hands physically dress the woman in the sweater.
The woman only raises both empty arms straight upward.
The foreground hands open the sweater, lower it over her head and raised arms, guide both arms through the sleeves, then pull the sweater down over her shoulders and torso.
The woman never touches the sweater.
The hands perform the entire dressing action.
Real physical contact, realistic knitted fabric stretching, cloth inertia, natural folds and realistic physics.
[5s–9s]
The same two foreground hands immediately bring the exact pink floral skirt from @Image3 upward from below.
The hands hold the skirt open in front of the woman.
The woman keeps her own hands away from the clothing and only lifts one leg, then the other.
The foreground hands physically guide both legs into the skirt and pull the skirt upward to her waist.
The hands adjust the waistband and layered fabric.
The woman never grabs or pulls the skirt.
Realistic weight distribution, balance and fabric physics.
[9s–12s]
The same two foreground hands quickly bring the exact pink ballet flats from @Image3.
The woman only lifts one bare foot slightly, then the other.
The foreground hands physically place one pink ballet flat onto the first foot, then the second pink ballet flat onto the other foot.
The woman does not touch the shoes.
Natural balance and realistic physical contact.
[12s–14s]
The same two foreground hands place the exact brown sunglasses from @Image3 onto the woman’s face, then place the exact cream shoulder bag from @Image3 onto her shoulder.
The hands make one quick final adjustment to the sweater and hair.
[14s–15s]
The two foreground hands move slightly away but remain visible near the edges of the foreground.
The fully dressed woman smiles warmly, naturally shifts her weight onto one leg and slightly extends one foot forward in a cute relaxed fashion pose.
STRICT PRIORITY:
The foreground hands perform EVERY dressing action.
The woman performs ZERO dressing actions.
The woman only raises her arms, lifts her legs and maintains balance.
Her own hands remain empty and away from all clothing throughout the entire video.
Exactly two large foreground female hands exist.
Never more than two foreground hands.
No third hand.
No extra arms.
No extra fingers.
No disappearing hands.
No new hands appearing later.
The two foreground hands must be present from the first second and actively manipulate every clothing item.
Every clothing item physically travels through space in the foreground hands before being placed onto the woman.
No instant outfit changes.
No magical transformation.
No teleporting clothes.
ONE CONTINUOUS TAKE.
NO cuts.
NO transitions.
NO montage.
CAMERA FULLY STATIC ON A TRIPOD.
NO camera movement.
NO zoom.
NO pan.
NO tilt.
NO orbit.
Preserve one single woman throughout the shot.
No duplicate woman.
No character drift.
No body duplication.
Photorealistic vertical 9:16 video, natural iPhone Reel look, warm natural daylight, candid, lived-in, slight imperfections, mild grain, smartphone realism, realistic skin texture, authentic human movement.
Realistic cloth physics, realistic knitted fabric stretching, natural fabric folds, realistic weight distribution, hair reacting softly to touch.
Natural room tone and subtle fabric rustle only.
No music.
No voice.
Чтобы эффект получился, руки должны быть уже на первом кадре.
Если просто загрузить фото человека без рук, нейросеть может решить, что человек должен одеваться сам. Именно поэтому сначала мы создаём отдельный стартовый кадр.
В результате получится видео, где две руки перед камерой сами собирают образ на человеке — от первой вещи до финальной позы.