ByteDance has unveiled Seedance 2.5, a new AI video model that pushes past a key limit. The model generates single clips up to 30 seconds long without any stitching. It marks a notable step for long-form AI video generation.
ByteDance revealed the model at Volcano Engine’s FORCE conference. Volcano Engine is the company’s cloud arm, which sells its AI models to enterprises. Seedance 2.5 is the centerpiece of the lineup and launches in early July.
The 30-second clips arrive complete with scene changes and tempo shifts. That removes the need to stitch shorter segments together afterward. The result is a smoother, more cinematic output from a single generation.
The model also accepts a large volume of inputs at once. It can process up to 50 added inputs, including reference images and audio. That capacity suits complex film scenes featuring multiple characters. Users can also edit generated videos later while keeping the visual style intact.
ByteDance did not stop at one model, though. The company also upgraded Seedance 2.0 with native 4K support. That version now handles 10-bit color depth for richer visuals.
Four more models joined the announcement at the conference. They include the Doubao 2.1 Pro language model and the Seedream 5.0 Pro image model. ByteDance also introduced the Seed-Audio 1.0 audio model.
Pricing emerged as a major talking point too. ByteDance said Doubao 2.1 Pro costs about 80% less than Claude Opus 4.6. That aggressive pricing underlines China’s push to undercut Western AI rivals.
The release deepens ByteDance’s expanding AI ambitions beyond TikTok. The company is racing OpenAI, Google, and others in generative video. Seedance 2.0 recently rolled out to 100 countries, though not the United States. Seedance 2.5 now raises the stakes again in the fast-moving AI video market.
