Model comparison
Seedance 2.5 vs Seedance 2.0
ByteDance previewed Seedance 2.5 at its Force conference on June 23, 2026 — and on paper it's a real generational jump over Seedance 2.0, the model Renoise runs today. Longer native clips, far more reference inputs, and frame-level editing are the headline changes. Here's exactly what's different, what ByteDance has actually confirmed versus what's still a rumor, and which model makes sense for your work right now.
Seedance 2.5 vs Seedance 2.0 at a glance
| Capability | Seedance 2.5 | Seedance 2.0 |
|---|---|---|
| Max clip length | 30s native, one shot | 5–15s |
| Reference inputs | Up to 50 multimodal | 9 images + 3 video + 3 audio |
| Localized editing | Edit a region, no full re-render | — |
| 3D blockout (白模) input | Yes | — |
| Resolution | Expected 4K | Native 4K (live) |
| Availability | Expected early July 2026 | Live in Renoise today |
The short version: Seedance 2.5 is built for longer, more controllable, production-grade work, while Seedance 2.0 is the model you can actually generate with today. Everything below the announced 2.5 specs comes from ByteDance's own preview; the 4K and launch-date rows are still hedged (more on that at the end).
What's actually new in Seedance 2.5
30-second native clips
This is the change that matters most. Most AI video models — Seedance 2.0 included — top out around 15 seconds, so anything longer means stitching clips together on a timeline. Stitching invites character drift and continuity breaks at every cut.
ByteDance says Seedance 2.5 generates a single, continuous 30-second shot natively, in one pass. A complete scene, ad, or story beat can hold from the first frame to the last without a single cut — which is a different kind of tool than a 15-second clip generator.
Up to 50 multimodal references
Seedance 2.0 already accepts references — up to 9 images, 3 video clips, and 3 audio tracks per generation. Seedance 2.5 raises that ceiling dramatically: up to 50 full-modal reference materials in a single generation, combining images, video, and audio.
That capacity is aimed squarely at brand films and episodic short dramas, where the model has to lock onto consistent brand elements, characters, and a long list of instructions at once. More references means tighter control over what actually shows up on screen.
Localized editing and 3D blockout
Two control features round out the upgrade. Localized editing lets you change a region of a clip — add a prop, fix a detail — without regenerating the whole video. And 3D blockout (白模) input lets a director pre-stage camera angles and composition with rough 3D geometry, so the generated shot follows an intended layout instead of a best guess.
Neither has a direct equivalent in Seedance 2.0. Together they push AI video toward the kind of repeatable, frame-level control that professional pipelines expect.
Resolution: both reach 4K
A common question is whether Seedance 2.5 will output 4K. It's expected to — but it's worth knowing that Seedance 2.0 already supports native 4K in Renoise today (100 credits per second; Seedance 2.0 Fast stays at 720p). So 4K isn't the reason to wait for 2.5; the length, reference capacity, and editing controls are.
See the difference
A single continuous 30-second clip generated by Seedance 2.5 — one character tracked across six rooms, each in a different art style, held consistent by eight image references with no stitching:
Frame-precise editing — starting from an existing clip and adding a single element while keeping the character, camera move, and environment intact:
And here's what Seedance 2.0 produces today in Renoise — one sentence to a 1080p shot with physically accurate motion:
Which one should you use today?
Use Seedance 2.0 — because it's the one that exists. It's live on the Renoise Canvas right now: text-to-video and image-to-video, 5–15 second clips, 720p / 1080p / native 4K, with a Fast mode for quick iteration. For most short-form work — product videos, music videos, social clips — 2.0 is more than capable, and you can stitch clips on the Canvas timeline for longer pieces.
Save Seedance 2.5 for when your project genuinely needs a single unbroken 30-second take, dozens of references for brand consistency, or region-level editing — and when it's actually available.
When will Seedance 2.5 be available?
Seedance 2.5 is not generally available yet. As of late June 2026 it's in a late closed beta, with a global enterprise beta recruiting — there's no public, self-serve access anywhere. Reports point to a public rollout in early July 2026, but ByteDance has not officially confirmed an exact date or public pricing, so treat that as a target rather than a guarantee.
Renoise runs Seedance 2.0 today and integrates new models the moment they reach general availability — so the plan is to add Seedance 2.5 as soon as it ships publicly. You don't need to wait to start producing video now.