First 30 second video generated by Seedance 2.5

ByteDance is nearing the release of Seedance 2.5, the next step in its video model line and the first to produce native 30-second clips in a single pass. Initially, the launch was scheduled for around July 9th, but it was postponed while negotiations with rights holders continued, and the API release is currently targeted for July 16th. Given how Seedance’s rollout has played out so far, the model could surface sooner in some places. Mention of the Seedance model is increasing across partner apps that bundle it with their video generation and editing toolkits, and pricing details have already been posted on China’s Jimeng.

We had early access to the model and the initial results are incontrovertible. All the clips we generated are 30-second one-shot videos, and the consistency of character, lighting, and pacing over that time period is unlike any other clip we’ve tested. The output is very well organized, although some small details may require regeneration. Announced at ByteDance’s FORCE conference in June, the model also accepts up to 50 reference inputs across images, video and audio, and supports region-level editing to change parts of a frame without having to redo the entire clip.

The rollout will run first through Jimeng and Dreamina, then CapCut and the Volcano Engine API, and will put this model in front of advertisers, short-form creators, and e-commerce teams, completely removing the stitching step with a 30-second spot in one pass.

The stakes are as competitive as the technical ones. Seedance 2.0 topped the independent video space earlier this year, even though Hollywood’s cease-and-desist letter delayed its global rollout. ByteDance has since previewed its copyright licensing platform with filter after filter. Google’s Gemini Omni was a close second in video editing, but Omni Flash’s top speed is about 10 seconds per clip, so a consistent 30 minutes of one-shots would reset that gap again. We should know by mid-July.

sauce