byteplus/dreamina-seedance-2-5

Seedance 2.5 is ByteDance's next-generation video generation model, unveiled at the 2026 Volcano Engine FORCE conference as the successor to the Seedance 2.0 series. Its headline capability is single-shot native video up to 30 seconds — double the 15-second ceiling of Seedance 2.0 — so creators can produce a complete, continuous scene in one generation instead of stitching shorter clips together. A single prompt accepts up to 50 full-modality reference materials (images, video, and more), giving fine-grained control over characters, style, motion, and composition for highly directed results. It is the first video model in the industry to offer a 3D white-box preview: before committing to an expensive high-fidelity render, the model produces a low-fidelity 3D preview animation so creators can validate camera, blocking, and motion and iterate cheaply — substantially cutting revision cost. The Seedance 2.0 line is concurrently upgraded to native 4K. Supporting both text-to-video and image-to-video, Seedance 2.5 is aimed at commercial advertising, short-drama production, and professional film and television previsualization, where long, controllable, high-resolution shots matter most.