An AI presenter moves from a snowy plateau into a studio rising from the snow. A finger snap brings the studio to life.

Independent promotional film.

How it was made

  • Model: Wan 3.0 reference-to-video on fal, one unbroken 30-second take at 1920 × 1080 and 30 fps, with sound.
  • Inputs: three pictures, a close-up of the face, a three-view character sheet and the studio, plus one prompt written shot by shot with timings.
  • Cost: $6.00 for the accepted take. A Kling 4.0 take on an earlier prompt cost $7.05 and was rejected. The full breakdown is in the cost guide.
  • Voice: the generated speech was converted to a custom ElevenLabs voice with the word timing unchanged, then placed in open air, a room and a close microphone as he moves.
  • Edit: this version starts at the finger snap and runs 23.6 seconds, with English captions.

Related: how to animate a photo, dialogue and lip sync and AI ads without filming.

Transcript

If you don't like recordings with camera, or never have time to shoot, here we go. Two things. I need two things: your photo, and what you want to say. The rest happens without you.

Have a film in mind?

Tell me what you would like to make, and I will tell you how I would approach it.

Discuss your project ↗