ByteDance: seedance-2.0

Seedance 2.0, released by ByteDance in February 2026, is a professional-grade video generation foundation model built on a Mixture of Experts architecture with a parameter scale of 200 billion. It produces audio and video together natively, synchronizing dialogue, sound effects, and background music with the visuals from the start. The model accepts up to 12 assets in a single generation—up to 9 images, 3 video clips, and 3 audio files—and supports 15-second high-quality multi-shot output with dual-channel audio. The model offers intelligent duration control, where setting duration to a negative value lets it choose the best length, and it performs adaptive aspect ratio selection based on the provided inputs. It also supports video editing and extension while preserving original motion and camera work, and users can turn audio off for silent video. Seedance 2.0 is designed for film and video production, marketing and advertising, e-commerce product showcases, educational training videos, and social media short-form content. Safety and provenance are addressed at the model level: generation of real people's likenesses, copyrighted characters, and protected brand identities is restricted, and every generated video embeds C2PA watermarking that cryptographically signs provenance metadata. Seedance 2.0 ranked first on Arena text-to-video, Arena image-to-video, Artificial Analysis image-to-video with audio, and Tsinghua University's WorldReasonBench among all tested models. It is available through CapCut and Dreamina as well as API providers including Fal and OpenRouter.