Short answer: Seedance 2.5 is not live on EmpirioLabs yet, but it is the video model to watch right now. Public reporting says ByteDance's Volcano Engine has introduced Seedance 2.5 with longer clips, richer reference inputs, and a staged release window. EmpirioLabs is preparing support so customers can try it soon after access, pricing, quality, and safety checks are ready.
What has been reported
The Decoder reported on June 23, 2026 that Seedance 2.5 is designed for single clips up to 30 seconds, scene changes and tempo shifts inside one generation, and up to 50 multimodal reference inputs such as images and audio. Treat those as reported launch details until ByteDance publishes final public API documentation.
That matters because many AI video workflows still depend on stitching several short shots together. A longer native clip could make product demos, UGC ads, music-video beats, and narrative tests easier to draft without breaking the flow between shots.
How it fits next to Seedance 2.0
Seedance 2.0 already pushed the category forward with cinematic motion, stronger prompt following, synchronized audio, and higher-resolution output. The Verge's coverage of Seedance 2.0 noted ByteDance's push into consumer and API video generation, while EmpirioLabs already exposes Seedance 2.0 tiers through the playground and API.
Seedance 2.5 looks like the next step in that same line: less clip stitching, more reference control, and more room for the model to carry a complete idea in one pass. If the final release matches the early reports, the main question for builders will be how much longer clips cost, which inputs are available through the API, and how fast jobs complete under real production load.
Planned on EmpirioLabs soon
Seedance 2.5 is planned for EmpirioLabs soon. Once the model is available for reliable API integration, we will add it to the model catalog, pricing page, playground, and the video generation API. If final model behavior differs from the launch-window reports, the EmpirioLabs model page will reflect the live version rather than stale rumor-window specs.
In the meantime, you can use Seedance 2.0 Pro for cinematic video with synchronized audio, or Seedance 2.0 Mini when speed and cost matter more than maximum resolution.
What to prepare before it lands
- Write prompts as scenes, not loose adjectives. Longer clips need clear subject, camera, setting, and timing.
- Collect reference material you have the right to use. If a model accepts many references, the legal and creative quality of those references matters.
- Decide your target length before prompting. A 30-second clip should have a beginning, middle, and end.
- Benchmark the same brief across Seedance 2.0 and Seedance 2.5 once both are available, then choose on quality, cost, and time to result.
Try Seedance today
Open the EmpirioLabs playground to test the current Seedance models, or call the same video workflow from POST /v1/videos/generations. Watch the model catalog for the Seedance 2.5 page once support goes live.
