Seedance 2.5: What ByteDance Revealed at FORCE (and What You Can Use Today)

At its FORCE conference on June 23, 2026, ByteDance gave the AI-video world its first look at Seedance 2.5 — and the headline is a big one: a single 30-second clip generated natively, plus up to 50 reference inputs in one shot. For a field where most models top out around 15 seconds, that's a real jump.

The catch: it's not out yet. ByteDance previewed 2.5 on stage and is targeting an early-July 2026 launch. So here's what's coming — and, since you can't generate with it today, what you can run right now to get most of the way there.

What's New in Seedance 2.5

ByteDance's pitch for 2.5 is a set of upgrades aimed at longer, more controllable shots:

• 30-second native clips. A single continuous generation, not several short clips stitched together — up from the ~15-second ceiling on today's models.
• Up to 50 reference inputs. One generation can take in as many as 50 references — images, video, and audio — for locking characters, products, style, and motion across a long shot.
• Region-level editing. Change one part of a frame — swap a background, replace a product, adjust a single subject — while the rest of the shot stays put.
• 3D blockout previsualization. A rough, untextured 3D layout (a "white model") that guides the generated scene — useful when the camera and composition need to be planned in advance.

ByteDance is aiming 2.5 at production and industrial work — commercial video, embodied AI, manufacturing — with the public launch expected in early July 2026.

Seedance 2.0 Today vs 2.5 This Summer

The leap in 2.5 is really about two things: duration and reference count. On both, today's Seedance 2.0 is already further along than most people assume:

Seedance 2.5 figures are from ByteDance's preview; final details land at launch.

So if you've been waiting on 2.5 for multimodal references or in-place editing, you mostly don't have to — Seedance 2.0 already does both today. What 2.5 adds is more: longer takes and far more references, not a brand-new capability class.

What You Can Use Today: Seedance 2.0 on Pixo

While 2.5 finishes up, Seedance 2.0 is live — and it's no consolation prize. It currently sits at the top of the Artificial Analysis video leaderboards. On Pixo you get:

• Multimodal reference fusion — up to 9 images, 3 video clips, and 3 audio tracks in one generation, to lock a face, a location, a motion reference, and a voice at once.
• Native audio — dialogue, sound effects, and music generated in the same pass.
• Clips up to 15 seconds at up to 1080p, with director-level camera control.

New to it? Our Seedance 2.0 guide walks through the reference system and prompt grammar, and our Seedance vs Veo vs Kling comparison shows where it wins shot to shot.

When 2.5 launches, we'll cover it in full — and add it to Pixo as soon as it's available.

Start creating with Seedance 2.0 in Pixo — free credits to start. New to AI video? Begin with our getting-started tutorial.

Frequently Asked Questions

When is Seedance 2.5 coming out?

ByteDance previewed Seedance 2.5 at its Volcano Engine FORCE conference on June 23, 2026, and is targeting an early-July 2026 launch. It isn't publicly available yet.

Is Seedance 2.5 available on Pixo?

Not yet — it hasn't launched. Pixo runs Seedance 2.0 today, and we'll add 2.5 once it's available.

What's new in Seedance 2.5 vs Seedance 2.0?

ByteDance previewed native 30-second clips (up from ~15 seconds), up to 50 reference inputs (up from 15 on 2.0), region-level editing, and a 3D blockout previsualization feature. It's expected to launch in early July 2026.

Does Seedance 2.5 generate 4K?

ByteDance hasn't detailed 2.5's output resolution yet. Seedance 2.0 already outputs up to 4K, so it's a reasonable expectation — but it's not confirmed until launch.

Can I try Seedance right now?

Yes. Seedance 2.0 is available on Pixo with multimodal references and native audio, in clips up to 15 seconds at up to 1080p.