ByteDance has announced Seedance 2.5, a new video generation model capable of creating single clips lasting up to 30 seconds without manual stitching. The system handles scene changes and tempo shifts while accepting up to 50 different inputs, including reference images and audio tracks. Users may also edit the resulting footage after generation while preserving the original visual style. The model is scheduled for release in early July alongside updates to the existing Seedance 2.0 version, which now supports native 4K resolution and 10-bit colour depth. Volcano Engine, the cloud division responsible for these tools, also introduced Doubao 2.1 Pro, Seedream 5.0 Pro, and Seed-Audio 1.0. The company claims the new language model costs approximately 80 percent less than Claude Opus 4.6.
This development matters because it removes a common technical limitation that previously forced creators to stitch multiple short segments together. Longer clips allow for more natural pacing and complex narratives within a single generation pass. Reducing the cost of high-end language models alongside these capabilities could lower barriers for enterprise adoption and independent production.
- Launch date set for early July
- Seedance 2.0 upgraded to 4K and 10-bit colour
- Doubao 2.1 Pro priced 80 percent below Claude Opus 4.6
