What is officially confirmed
On 31 July 2026, ByteDance Seed introduced Seedance 2.5 as a model for longer audiovisual storytelling. The official announcement describes clips up to 30 seconds, multi-round extension, multimodal references, and controlled editing.
What Seedance 2.5 can do
Coherent sequence
One request can describe a longer dramatic episode with a clear progression of action.
References
Images, video, and audio can help preserve characters, environments, motion, and atmosphere.
Editing
You can direct changes at the level of a time range, camera, transition, or visual element.
Sound and image
The model connects speech, effects, music, motion, and editing logic in one result.
Where to check access and documentation
Official Seedance 2.5 announcement
Capabilities and constraints stated by ByteDance Seed.
Seed / announcement ↗Seedance 2.5 model page
A concise overview of references, editing, and the model workflow.
Seed / model page ↗Dreamina workspace
Check actual availability, limits, and interface behavior in your account.
Dreamina / product ↗Practice: direct one scene
Start with the dramatic core rather than a list of effects: who wants what, what changes, and which sound helps the viewer understand it.
Define the action
Describe the subject, goal, obstacle, and visible change at the end of the clip.
Divide the time
Set the opening, turn, and ending. For each part specify action, camera, and sound.
Secure the rights
Use your own or permitted references, and do not upload other people faces, voices, or brands without a valid basis.
Cinematic scene “The Last Train”: a rainy evening platform, one subject realizes the train is leaving without them. 0–4s: wide shot, puddle reflections, sparse announcements. 4–10s: camera moves closer as the subject notices a light in the last carriage. 10–16s: a short burst forward, then a stop; footsteps, brakes, and rain stay synchronized with the motion. Restrained indigo-and-amber palette, realistic light, a clear arc from waiting to decision. No text, logos, or random passers-by.
How to evaluate the result
Meaning first
Is it clear what happened and why the scene ends where it does?
Then consistency
Do the face, objects, motion direction, and light remain stable between shots?
Sound after that
Does the audio support the edit, or does it feel layered on top of the picture?
Cost, access, and limits
Access, limits, resolution, queueing, and price depend on the specific product and account. Do not promise a fixed tariff or universal limit; verify them on the current service page before publication.