Ayer en el Live-Stream que solemos hacer todos los miercoles @Pablo Santos me propuso poner a prueba Seedance 2.0 de una manera que no se me habría ocurrido nunca! 😅 La idea es darle solo una referencia en video para que la IA tome de ahí la referencia del personaje y la voz, para crear un video totalmente nuevo 🎬 Tras ponerlo a prueba, la verdad es que me sorprendí muchisimo de que se pueda hacer 😳 y dandole tan solo un video de 4 segundos como referencia.. Solo quería compartir con vosotros este pequeño descubrimiento 🫶🏼 --- 🔴 Os dejo por aquí el prompt del video: A 2000s studio-comedy film shot, 16:9, 35mm, moderate film grain, bright American daylight, tracking camera at chest height following a woman along a city sidewalk. Smooth dolly/track with her, no handheld phone, no gimbal whip, no drone, no cuts. The camera stays with her the whole 15 seconds, medium-wide, the street readable behind her. VIDEO REFERENCE @vid1 is Aria's identity AND voice lock: use this reference clip as the source of her exact face, hair (high ponytail), round silver-frame glasses, silver hoop earrings, body, and — critically — her exact real speaking voice, tone, pitch and natural accent as heard in that reference clip. Do not redesign her face. Do not invent a new voice; the voice performing every spoken line in this video must be a direct match to the voice heard in @vid1, same person, same instrument, no drift. No static image reference is used anywhere in this generation — @vid1 alone is the sole source of face, body and voice. She sings out loud, diegetic, the lyrics "I've got the power" in a loud off-key enthusiastic voice, in her own real voice as established by @vid1. No extra score under it. Street air, footsteps, then the gag and the spoken aside. TIMELINE 0-3s: [Tracking medium-wide, opening on @vid1's identity] She is already walking, already singing, mouth wide, chest out, a little strut. The camera slides with her. Daylight, grain. She is having the time of her life.