Seedance 2.5 is the video model behind the BytePlus video generator, and the same model is available through the Seedance API. This guide covers what a request can carry, what comes back, and how to plan usage before you integrate.
Three input modes
Seedance 2.5 accepts three kinds of starting point, all through the same API surface:
- Text to video - describe subject, action, camera and pacing in a prompt.
- Image to video - start from a still that already carries the composition you want to keep.
- Video to video - transform or restyle footage you already have.
Reference assets
A single request can carry up to 50 reference assets, mixing images, video and audio. References let the model follow a specific subject, style or motion instead of guessing from text alone, which is what makes iteration predictable: swap one reference, regenerate, and compare.
What comes back
Seedance 2.5 renders native 30-second video in a single request. That removes the stitching step where short clips have to be joined and colour-matched afterwards, and it keeps a whole take inside one generation.
Planning quota and concurrency
Generation is credit based. Token packages and plan tiers set the balance, the quota and how many requests can run at the same time, so size the tier against the number of concurrent renders you need rather than the number of renders per month. Current tiers are listed on pricing.
Prototype before you integrate
The browser workbench runs the same model and spends the same credits, so prompt and reference validation can happen visually first:
- Open the video generator and write the shot.
- Add the reference assets you intend to send through the API.
- Once the output is right, move the same request into your backend.
Where to go next
- Seedance API overview
- Documentation
- Plans and token packages
- Working with stills first? Start with the Seedream API.
