ByteDance launches Seedance 2.5 with 30-second one-take video

ByteDance Seed unveiled Seedance 2.5, a joint audio-video generation model that produces 30-second clips in a single pass. It supports up to 50 reference inputs (30 images, 10 videos, 10 audio clips) and timestamp-level editing. It is live in Dreamina, Jimeng and Doubao, with an API promised via BytePlus ModelArk.

• Single pass generates 30 seconds of video, up from 15 in Seedance 2.0
• Up to 50 references per run: 30 images, 10 videos and 10 audio clips
• Multi-round extension reaches 180 seconds in Dreamina beta
• Timestamp-level editing without regenerating the whole scene