Le prompt HOODIE : ton vêtement qui se recoud sur toi

Le prompt exact, l'image clé et la méthode pour que l'IA ne change pas ton produit.

Ce qu'il te faut

Étape 1 : l'image clé

C'est elle qui fige ton visage (et ton produit). La vidéo ne fera que l'animer, c'est pour ça que le résultat te ressemble.

Modèle : GPT Image 2.5, qualité high, format 16:9. Mets ta photo en image de référence, puis colle :

Cinematic fashion film still, shot on ARRI Alexa 35, 50mm. The exact same young man from the reference photo (same face, same beard, same build) wears a black trucker cap and the exact hoodie from the product photos (same color, same cut, same pocket, same ribbing, no logo). He stands centered, hands relaxed, looking straight into the lens with a slight confident half-smile, in a large minimal photo studio with warm off-white plaster walls and a polished concrete floor. One large soft key light from the left, gentle shadow on the wall, a few fine threads floating in the air around him catching the light. Calm, premium, editorial, realistic skin and fabric texture with visible fleece grain, fine film grain. No text, no watermark.

Étape 2 : la vidéo

Réglages : Seedance 2.5, mode omni reference, 15 s, 16:9, 720p. L'image clé en image de départ, ta photo en référence.

Premium fashion film in a minimal off-white plaster studio, shot on ARRI Alexa, soft warm key light, fine film grain. The man from the reference photo wears the exact taupe hoodie from the product photos: blank washed taupe heavyweight fleece, oversized boxy fit, kangaroo pocket, ribbed cuffs, no logo, no drawstrings. Same face and same hoodie throughout.
[0-3s] He snaps his fingers: instantly the hoodie bursts apart into its flat sewing pattern pieces (front, back, two sleeves, hood, pocket, ribbed cuffs) that freeze floating around his body in slow motion, threads trailing, he stands in a plain black t-shirt.
[3-6s] The camera orbits slowly around him through the floating fabric panels, showing the garment-dyed fleece texture and the stitched edges in sharp detail.
[6-9s] The panels fly back toward him and stitch themselves together in mid-air, taupe threads zipping along every seam, the hoodie rebuilds itself on his body.
[9-12s] Macro close-up: his hand slides over the kangaroo pocket and the ribbed cuff, soft heavy fleece moving naturally.
[12-15s] He pulls the hood up, steps toward the camera with a slight smile, the last loose threads drift down in the light.
Camera: smooth, physical, real motion blur. Sound: a crisp finger snap, fabric whooshes, fast sewing machine ticks, soft room tone. No text on screen.

Les 3 secrets

  1. Découpe en tranches de 3 secondes ([0-3s], [3-6s]…). Le modèle sait exactement quel plan jouer à quel moment, au lieu d'improviser.
  2. Décris le son. Moteur, pas, vent : Seedance et Kling génèrent l'audio, et c'est lui qui rend la vidéo crédible.
  3. Coupe la première demi-seconde au montage : les modèles démarrent presque figés, ta vidéo attaque direct.

Pour payer moins cher