Dreamina says Seedance 2.5 makes AI video longer and easier to edit, but access is uneven

ByteDance is trying to move AI video beyond short clips. Dreamina says Seedance 2.5 can generate 30-second videos, extend them to three minutes in a beta mode and repair selected areas without rebuilding an entire scene.

That would expand on Seedance 2.0, but access is muddled. Dreamina’s launch announcement said the model was available to subscribers across Europe, Southeast Asia, the Middle East, Africa and South America. Its list did not include the US. One user outside the listed regions said it was not yet live for him, and Dreamina says access can vary by account, region and product surface.

More time and more control

Seedance 2.0 supported audio-video generation lasting four to 15 seconds. Its open platform accepted up to three video clips, nine images and three audio clips as references, according to the model’s research paper.

Dreamina says Seedance 2.5 generates 30 seconds in standard mode, compared with the 15-second maximum described in the Seedance 2.0 paper. It also says creators can combine as many as 50 inputs, including prompts, scripts, photos, videos, music and style guides.

Dreamina says users can select and refine only the parts of a video that need adjustment instead of regenerating the whole scene. That could make revisions easier if the feature works reliably.

The company also claims better continuity, realism and readiness for production. The only accounts of those claims in the coverage reviewed come from Dreamina’s own announcement and product pages; no independent testing appears in it. The three-minute option is a beta extension mode, not evidence that Seedance can consistently produce one coherent three-minute generation.

A launch that is still unfolding

The release does not appear to be a completed global rollout. Dreamina’s announcement said subscribers in five broad regions had received access, but it did not establish that every account in those regions could use the model.

Dreamina’s official access guide says that “account-level availability can still depend on rollout, region, and product surface.” Its official Seedance 2.5 model page also still said “Seedance 2.5 is Coming Soon!” when fetched on July 31, 2026. Dreamina’s own AI video hub, dated the same day, says the model “is now available on Dreamina.”

Seedance 2.5 is ByteDance’s bid to make AI video longer, easier to direct and easier to revise, but access and its longer-form performance remain uneven at launch.