ByteDance’s cloud division, Volcano Engine, announced the Seedance 2.5 AI video model at a Beijing conference.
The updated model generates continuous 30-second video clips from a single prompt. This capability doubles the 15-second limit found in most existing AI models.
Users can input up to 50 multimodal reference materials such as images and audio. This represents a significant increase from the 12-item limit in the previous version.
Seedance 2.5 supports native 4K resolution. Volcano Engine scheduled the public release for early July 2026.