ByteDance launches Seedance 2.0 video model
ByteDance has launched Seedance 2.0, a video model that generates audio and video together from mixed text, image, audio and video inputs, putting it up against Sora and Veo.
ByteDance's Seed team announced the official launch of Seedance 2.0. It is the next generation of the company's video creation model and succeeds version 1.5.
The model uses what ByteDance calls a "unified multimodal audio-video joint generation architecture." It accepts four input types: text, image, audio and video. In a single request, users can supply up to nine images, three video clips and three audio clips, plus natural-language instructions. The model can then draw on those assets for composition, motion, camera movement, visual effects and sound. ByteDance says it can also follow text-based storyboards, and its examples tag reference assets inline in the prompt (e.g. "@Image 1").
The model outputs multi-shot clips of up to 15 seconds with dual-channel (stereo) audio. It also supports extending and editing existing video, and ByteDance highlights better instruction following, subject consistency and prompt-driven camera planning.
The company says the main gains over 1.5 are in complex motion and multi-subject interaction. Its examples include pair figure skating with synchronized jumps and lifts, and competitive sports scenes with several participants, which it says earlier models struggled with. ByteDance describes the model's usability in these scenarios as "industry-leading SOTA." That claim rests on ByteDance's own internal evaluations. The post does not cite independent benchmarks.
The announcement does not give details on pricing, API access, availability by region, model size or training data. A product home page is live at seed.bytedance.com/seedance2_0.