ByteDance unveils Seedance 2.5, generating native 30-second video clips
The model generates single clips up to 30 seconds long, including scene and tempo changes, without stitching together separate shots as earlier video models required.
- Models & capabilities
- Benchmarks & progress
- Minor
ByteDance unveiled Seedance 2.5, a video-generation model, at Volcano Engine’s FORCE conference — a separate product line from Seed 2.1, the language-and-agent models the company showed at the same event. The Decoder reported that the model could generate single clips up to 30 seconds long, “complete with scene changes and tempo shifts,” in one pass, rather than the shorter clips that earlier text-to-video models produced and that creators then stitched together in editing software.
The model could also take in up to 50 additional inputs at once — reference images, audio and other material — and supported post-generation edits that preserved a clip’s established visual style, useful for maintaining continuity across scenes with multiple characters. ByteDance did not publish independent benchmarks or pricing for Seedance 2.5 at the preview stage, and said the model was set to launch more broadly through Volcano Engine in early July 2026.