Seedance 2.5
Seedance 2.5 is ByteDance's next AI video model — native 30-second clips and up to 50 multimodal references.
Seedance 2.5
ByteDance’s next AI video model: native 30-second clips in a single generation, guided by up to 50 multimodal references.
Seedance 2.5 is the successor to Seedance 2.0, announced by ByteDance at its Volcano Engine FORCE conference on June 23, 2026. The headline upgrade is duration and control: a single generation now produces a continuous 30-second clip instead of several shorter segments stitched together, and you can guide that generation with up to 50 combined references — text, images, video and audio — in one workflow. ByteDance also describes region-level editing, so you can redraw just the part of a frame that’s wrong instead of regenerating the whole clip, along with native synced audio and support for 4K output.
Seedance 2.5 is currently rolling out in stages. It isn’t generally available yet, official pricing hasn’t been published, and none of ByteDance’s stated specifications have been independently verified. Mitte will add Seedance 2.5 as soon as it reaches general availability — until then, Seedance 2.0 is live on mitte.ai today with native audio, cinematic motion and 4K output.
What’s new in Seedance 2.5
- Native 30-second clips. One continuous generation, no stitching multiple short segments together.
- Up to 50 references per generation. Mix text, images, video and audio to drive a shot with real creative control.
- Region-level editing. Fix a single element in a frame — a hand, a logo, a background detail — without regenerating the whole clip.
- Native synced audio. Sound is generated together with the video, not layered on afterward.
- 4K output. Higher resolution generation for production-ready footage.
These are ByteDance’s own stated specifications as of the June 23, 2026 announcement and are not yet independently verified.
Seedance 2.5 vs. Seedance 2.0
| Seedance 2.0 | Seedance 2.5 | |
|---|---|---|
| Clip length | Up to ~15 seconds | Native 30 seconds, single pass |
| Reference inputs | Fewer, single-type | Up to 50, multimodal |
| Editing | Full regeneration | Region-level editing |
| Resolution | Up to 4K | Up to 4K (claimed) |
| Audio | Native synced audio | Native synced audio |
| Availability | Live now on mitte.ai | Staged rollout, not yet generally available |
Want to pair it with an image model once it lands? Browse the image model collection.
FAQ
What is Seedance 2.5? ByteDance’s next-generation AI video model. It generates native 30-second clips from text, image, video or audio input, guided by up to 50 references in a single generation, with synced audio and claimed 4K output.
Who made Seedance 2.5? ByteDance’s Seed research team, the same team behind Seedance 2.0 and the wider Seed model family.
When will Seedance 2.5 be released? It was announced on June 23, 2026 and has been rolling out in stages since. It’s not yet generally available, and ByteDance hasn’t confirmed a full public release date or official pricing.
How do I use Seedance 2.5 on mitte.ai? It isn’t available on mitte.ai yet. We’ll add it as soon as it reaches general availability. In the meantime, Seedance 2.0 is live and runs directly in your browser, no setup required.
What can Seedance 2.5 create? Continuous 30-second clips guided by up to 50 mixed references, with region-level editing, native audio, and up to 4K resolution, per ByteDance’s announcement.
Can it use my own images, video or audio? Yes. Seedance 2.5 is designed to be multimodal, accepting up to 50 combined text, image, video and audio references per generation.