ByteDance Seedance 2.5
Next-generation video model producing up to 30 seconds of cinematic video with native audio from image, reference, or text.
Studio is in early access; availability resolves per project after sign-in.
- Endpoints
- 6
- Eligible
- 6
- Input schemas imported
- 6 / 6
- Delivery
- fal.ai
- Weights
- Not stated
- Developer
- Not stated in source
Overview
Seedance 2.5 generates up to 30 seconds of cinematic video in a single pass with native audio, real-world physics, and director-level camera control. Image input serves as the first frame, with motion, camera work, and synchronized audio generated from it; an optional end frame sets the conclusion. The reference endpoint combines up to 50 multimodal references, spanning images, video, audio, and style, into one generation. Audio is co-processed in the same latent space as video.
Capabilities
- Up to 30 seconds per single generation
- First-frame plus optional end-frame control
- Up to 50 multimodal references per generation
- Co-processed audio in the video latent space
Best for
- Long-form cinematic clips
- Style-consistent multi-reference video
- Directed start-to-finish shots
Use cases
- 30-second narrative clips
- Reference-stacked brand video
- First-frame-anchored animation
Endpoints
6 eligible. Cataloged is not the same as executable: Studio resolves which endpoints a project can run.
| Endpoint | Task | Catalog status | Inputs |
|---|---|---|---|
bytedance/seedance-2.5/image-to-video | image to video | Eligible | 11 fields · requires image_url |
bytedance/seedance-2.5/reference-to-video | reference to video | Eligible | 14 fields |
bytedance/seedance-2.5/text-to-video | text to video | Eligible | 7 fields · requires prompt |
bytedance/seedance-2.5/us/image-to-video | image to video | Eligible | 10 fields · requires image_url |
bytedance/seedance-2.5/us/reference-to-video | reference to video | Eligible | 13 fields |
bytedance/seedance-2.5/us/text-to-video | text to video | Eligible | 8 fields · requires prompt |
Questions
How long can generations run?
Up to 30 seconds in a single pass.
What references can it combine?
Up to 50 across images, video, audio, and style.