Multi-Reference AI Video Generation

Most AI video generators take one image plus a text prompt. Multi-reference video generation lets you feed the model several references at once — images, clips and audio — and control exactly which traits it borrows from each. Seedance brings multi-reference (reference-to-video) generation to your iPhone.

What is multi-reference video generation?

Multi-reference generation means the model creates a clip from multiple reference inputs instead of a single image. Pin a character from one photo, the camera move from a clip, the lighting from another image and the soundtrack from an audio file — all in one generation, for consistent, multi-shot results you can actually direct.

Reference anything with @mentions

In the Seedance app you point at uploads in plain language using @mentions:

  • “@Image1 as the first frame, then a slow dolly-in.”
  • “Use the camera move and pacing from @Clip1.”
  • “Keep the character from @Image2 consistent throughout.”
  • “Sync the cut to @Audio1.”

What you can reference

  • Up to 9 images — first/last frame, characters, style or scene.
  • Up to 3 video clips (15s combined) — motion, choreography, camera work, effects.
  • Up to 3 audio tracks — music or sound to drive timing and mood.

Why multi-reference beats single-image image-to-video

A single image only tells the model what the first frame looks like. Multi-reference tells it how to move, who to keep consistent, how the camera should travel and what it should sound like — so you get controllable, repeatable results instead of a lucky guess.

What you can make

From social clips and music videos to product ads and short films — Seedance is built for it. Learn what Seedance 2.0 can do, or see how it stacks up in Seedance vs Sora.