Generate, edit, extend
Text, images, video, and audio go in; new footage, edited footage, or continued footage comes out. One multimodal surface for all three moves.
Model family · Video
Bring footage. It keeps the thread.
Seedance takes any mix of text, images, video, and audio, and generates, edits, or extends from it — sound arriving with the picture, characters holding from reference clips. In MML ONE it answers your storyboard: five tiers, routed shot by shot.
By ByteDance (Seed team)
What it's best at
One model, four kinds of input.
What Seedance actually does with material — per its public API releases, not the launch-week hype.
Sound with the picture
Native audio generated together with the frame — 4 to 15 second clips, and since June 2026, native 4K at 10-bit color.
Reference-clip filmmaking
Feed real footage and stills to hold characters, props, and look across takes — the announced 2.5 generation scales this to roughly fifty simultaneous inputs.
A real fast tier
Seedance 2.0 Fast is a first-party tier with the same input surface at lower latency — not a reseller invention.
Provenance built in
Output ships C2PA-labeled. For commercial delivery, that audit trail is a feature, not a footnote.
In MML ONE
Five Seedance tiers, ready to route.
Exactly what our catalog serves today — vendor names, newest first.
Seedance 2.0The current generation: multimodal input, native audio, 4–15 second clips.
Seedance 2.0 FastSame input surface, lower latency — the first-party speed tier.
Seedance 2.0 FaceThe verified real-person material route — consented likeness only, through the liveness-verified flow.
Seedance 2.0 Fast FaceThe verified-likeness route at fast-tier latency.
Seedance 1.5 ProThe previous generation, still served for continuity.
Tier lineup as served in MML ONE on the day this page was written. Seedance 2.5 is announced for early July 2026 — the catalog inside the app is the live truth.
In the MML ONE flow
Where Seedance earns its place.
Takes land in the storyboard as versioned assets. Reference-heavy work — a product that must stay itself, a performer who must stay themselves — is what you route to Seedance.
The honest part
Know before you route.
Not available in the US, and the family is under active copyright-litigation pressure from major studios. If your delivery footprint includes the US, plan around it.
Realistic human faces can't be uploaded as source material. Real-person likeness requires the verified-consent flow or ByteDance's virtual-human library.
Capacity and channel gaps are real: launch-period queues ran hours, and international API resolution has trailed the Chinese channel.
Public Alpha
Bring one film into the graph.
Start with a premise, a screenplay, or a folder of references. We'll set up your provider keys and walk through the first scene with you.