ByteDance's Seedance 2.5 generates native 30-second 4K video clips
Seedance 2.5 claims 30-sec single-pass video with 50 reference inputs—enterprise beta now.
Deep Dive
At the Volcano Engine FORCE conference on June 23, 2026, ByteDance announced Seedance 2.5, its next-gen text- and image-to-video model. It claims native 30-second single-pass clip generation, up to 50 multimodal reference inputs (images, audio, video), and local re-draw editing that changes one element without altering motion or lighting. Currently in global enterprise beta, public launch is targeted for early July 2026. The current Seedance 2.0 leads independent blind-preference tests ahead of Google Veo 3.1 and Kling 3.0.
Key Points
- Native 30-second single-pass 4K clip generation—longest single-shot duration of any current model.
- Up to 50 multimodal reference inputs (images, audio, video) for consistent character, product, and style control.
- Local re-draw editing swaps one element (e.g., product) without altering motion, camera, or lighting.
Why It Matters
Pushes AI video past short clips to native long-form 4K, with unprecedented control for marketing and content teams.