ByteDance launches Seedance 2.5 video model
ByteDance has released Seedance 2.5, which doubles the length of a single generation to 30 seconds and adds multi-round extensions, pushing its video model toward multi-minute, story-length output.
ByteDance's Seed team announced Seedance 2.5, the next version of its video-generation model. It keeps the unified audio-video joint-generation architecture introduced in Seedance 2.0.
The main change is length. A single generation now runs to 30 seconds of video with audio, up from 15 seconds. Users can extend a clip over several rounds, and ByteDance says characters, settings and pacing stay consistent across extensions, so creators can produce videos several minutes long without splicing clips together by hand. The company says the model can fit several connected shots into one 30-second clip, giving a story a setup, development, a turning point and a resolution instead of stretching out a single moment.
Reference inputs have grown too. A single generation can now take up to 30 images, 10 video clips and 10 audio clips. ByteDance also cites stronger clay-render, motion and "creative" referencing. For editing, the model adds timestamp-level control over audio and video, plus improved green-screen, camera-perspective and reference-based editing. ByteDance is aiming these features at film and advertising work.
Seedance 2.5 is rolling out now on Jimeng AI, Doubao Pro and other platforms. ByteDance says API access through BytePlus ModelArk is "coming soon" but has not given a date.
The announcement includes example prompts and a short film made end-to-end with the model. It does not include benchmarks, independent evaluations or pricing, so all the quality claims come from ByteDance alone.