Seedance 2.5 — video takes another leap
ByteDance''s next-generation video model Seedance 2.5 was unveiled at the 2026 Volcano Engine FORCE conference, with public information pointing to a launch around July 2026. It moves AI video from "can generate" to "can deliver" — longer, sharper, more controllable. XinYu AI is working on access; call it right on the canvas once it lands.
🎬 Five highlights
30-second clips · room to tell a story Native 30-second length means fuller scenes and smoother pacing — tell a complete little story in one clip, no more stitching.
Native 4K · sharper by default Native 4K output lifts detail, edges and texture across the board — deliverable, publish-ready footage straight out.
Up to 50 multimodal references · locked across shots Feed up to 50 references at once — image / text / audio / video — to lock characters, products, style and brand assets into every shot. No more face-swaps between cuts.
3D blockout previews · plan before you render Preview layout, blocking and camera moves with a 3D blockout before generating — put control up front, then render.
Precise local edits inside a video Not happy with one spot? Adjust it locally instead of regenerating the whole clip.
What you can build
30-second product-launch and landing-page clips, 4K ecommerce showcases, multi-reference brand storytelling and virtual influencers, AI short films and trailers — a more complete visual story from script, characters and mood board.
Stay tuned
Same as always at launch — just use your credits, no extra subscription required. Exact timing will be announced on-site.
Summarized from public pre-launch information; final specs and performance are subject to ByteDance''s official release.