ByteDance Launches Seedance 2.5: 30s Single-Pass Video, 50 Reference Assets
xiaohu · x · 2026-07-31
ByteDance's Seed team has released Seedance 2.5, their latest AI video generation model. The major upgrade is that single-pass video generation has doubled to 30 seconds, allowing a continuous shot to cover a full narrative scene from buildup to conclusion.
Additionally, the limit for simultaneous reference assets has jumped from 15 to 50. The new version introduces second-by-second segmented instructions and timestamp-based editing, enabling precise modifications for green screens and camera angles. To showcase its capabilities, the team released a 4-minute, 22-second short film entirely generated by the model. It is now available on Jimeng and Doubao Pro.
Related event: ByteDance Releases Seedance 2.5 with 30s Video and Multimodal Control(53 posts)→
More from Multimodal
- Gaussian Splatting Generates Volumetric Fog Without Any Proxy Meshes — willeastcott · 2026-08-01
- MiniMax H3 ComfyUI Workflow Now Available — EverythingMacPro · 2026-08-01
- Generative Feedback Loop: AI 'Digests Itself' Over 100 Iterations — mars_santa · 2026-08-01
- Fully Local Generation: Creating a Metal Music Video with LTX 2.3 and Suno — Luzifee-666 · 2026-07-31
- a16z to host AI creative tools meetup in SF next week — venturetwins · 2026-07-31
- H3 Video Model Local Deployment: If Half as Good as API, It Changes Everything — CrasHthe2nd · 2026-07-31