How this film was made
The 30 seconds above is AI-generated, not live action. No studio, no camera, no cast.
- Every frame came from Seedance 2.5 at 1080p, as a single 30-second generation — not stitched from parts;
- It packs 21 shots, and the on-screen title cards were generated along with them;
- The whole piece was made on the XinYu AI canvas, off just three character reference images.
We used the model itself to show what the model can do. Every step available to you on the canvas is a step we took ourselves.
What you can do with Seedance 2.5 now
- Up to 30 seconds per clip — a beat that used to get cut short at 15s can now play out in one piece, instead of being split across two generations and stitched back together.
- Change one part, leave the rest alone — someone wants a different prop in the frame? Edit that one element; everything else stays as it was.
- Extend from the end — a clip you already like can keep going. Continue from where it stops rather than starting over.
- 30 images, 10 videos and 10 audio tracks per generation — characters, sets, props, motion references: everything a sequence needs can go in at once.
- Audio-driven — attach a music track and the picture follows the beat.
- Output at 480p / 720p / 1080p.
A path you can follow
- Generate your main 30-second pass;
- Something in the frame is wrong — use video editing to change just that part and leave the rest alone;
- It needs to run longer — use video extension to continue from the end;
- Swap the track and let audio-driven re-time the cuts.
All four steps happen on the same canvas, with your source material, references and finished cut sitting together.
Try it now
Open a canvas, add a video node, pick Seedance 2.5, and wire your reference material into it.
If you have something of your own to advertise — make the film with it.