Now available on CloneViral
Seedance 2.5
is live.
ByteDance’s newest video model, built on a unified multimodal architecture: 30 seconds in a single pass, up to 50 references, native audio, and in-frame multilingual text.
Thirty seconds, generated in one pass.
Seedance 2.5 renders up to 30 seconds of continuous video in a single pass — double the ceiling of Seedance 2.0. The whole clip is planned at once, so lighting, motion, and character identity hold from the first frame to the last instead of drifting the way stitched segments do.
Up to 50 references in one request.
Pack up to 50 multimodal references into a single generation, mixing images, video clips, and audio — four times the capacity of Seedance 2.0. Anchor a character, product, wardrobe, or brand style once and it stays consistent across every shot.
Audio and on-screen text, built in.
Music, dialogue, and sound effects are generated alongside the video rather than dubbed on afterward, so lip sync and sound effects land on the right frames. Titles and multilingual subtitles render directly in frame, in languages including Chinese, English, Spanish, Arabic, Japanese, and Korean.
Three ways in.
Start from a text prompt, from a first frame (with an optional last frame to direct where the shot ends), or from a set of references. All three modes are available now in the video generator at 480p and 720p.