Seedance 2.5: release date, what's new, and how to be ready on day one
ByteDance's Seedance 2.5 brings native 30-second clips, 50 reference inputs, and joint audio-video generation. What's confirmed, what's not, and how to have a workflow ready the day it ships.
ByteDance announced Seedance 2.5 on June 23, 2026 and opened its API on BytePlus on July 16. The three confirmed upgrades: native 30-second single-segment video, up to 50 reference inputs, and audio generated in the same pass as the picture. It is not on Western hosting platforms yet. If you build your ad workflow on Seedance 2.0 today, switching to 2.5 when it lands is a one-node swap.
What did ByteDance actually announce?
Seedance 2.5 was announced at the Volcano Engine FORCE Conference on June 23, 2026, went through an enterprise beta, and its API went live on BytePlus on July 16, 2026. Coverage from TechTimes and BigGo confirms the same three headline features:
| Feature | Seedance 2.0 (today) | Seedance 2.5 (confirmed) |
|---|---|---|
| Clip length | Short segments, stitched for longer cuts | Native 30-second single segment |
| Reference inputs | A handful of images | Up to 50 full-modal reference materials |
| Audio | Added after generation | Generated with the video in one latent space |
| Consistency | Good within a segment | Holds character, lighting, and motion style across the full clip, per ByteDance |
| Pricing | Per second on hosting platforms | Not disclosed |
The audio point is the sleeper. ByteDance describes the architecture as joint audio-video generation, meaning the model decides the sound and the picture together rather than bolting a voiceover onto finished footage. If it works as described, lip sync stops being a separate step that can fail.
When can you actually use it?
Right now, only through the BytePlus API, which is aimed at enterprise accounts. The hosting platforms most Western teams use have announced intent but not shipped: fal has said Seedance 2.5 will sit next to Seedance 2.0 in its catalog once ByteDance releases it to partners.
That gap between "the API exists" and "you can use it where you already work" is normal. Seedance 2.0 followed the same path before landing on fal in April 2026.
What will it cost?
Nobody outside ByteDance knows yet. Every Seedance release so far has been priced per second of output on hosting platforms, and there is no sign 2.5 breaks the pattern. For planning: a 30-second native clip will cost meaningfully more per run than the 5-to-10-second clips most teams generate today, which makes it a finishing tool, not a drafting tool. Draft short and cheap, then commit to the long take.
How do you prepare without access?
The practical answer: build the workflow now on Seedance 2.0, which behaves like a shorter-clip version of the same family.
In Orisu, a video ad is a workflow, not a prompt box. The product, script, actor, scene, and model each sit on their own node. When 2.5 reaches our providers it joins the catalog, and upgrading an existing workflow means changing the model on the video node. The script, brand kit, references, and every downstream step stay put. Teams that already run Seedance 2.0 workflows in Orisu will be generating with 2.5 the day it arrives, with nothing rebuilt.
That is also our general position on model releases. Video models leapfrog each other every few months, so the durable asset is the workflow around the model, not loyalty to any one of them.
What we still don't know
Three open questions worth tracking before you plan campaigns around 2.5:
- Resolution at launch. Reporting mentions 4K on the roadmap, but launch resolution is unconfirmed.
- Rate limits and regional availability outside BytePlus.
- Real-world consistency. "Holds character across 30 seconds" is ByteDance's claim, and every video model demo overperforms its median output. Judge it on your own product, not the sizzle reel.
We'll update this post as partners ship it.
Judge it on paper.
The free tier takes an email and a minute. Paste your URL, build a brand kit, and compare the output yourself.
Common questions.
Is Seedance 2.5 out yet?
Partly. ByteDance announced it on June 23, 2026 and opened the API on BytePlus on July 16, 2026. Most Western platforms that host Seedance models, including fal, have said they will carry it but have not shipped it yet.
What is new in Seedance 2.5 compared to 2.0?
Three confirmed changes: native 30-second single-segment clips instead of stitching shorter ones, support for up to 50 reference inputs, and audio generated together with the video in one pass instead of being added afterward.
How much will Seedance 2.5 cost?
ByteDance has not published pricing yet. Seedance models have so far been priced per second of generated video on hosting platforms, and 2.5 will most likely follow the same pattern.
Can I use Seedance 2.5 in Orisu?
The moment our providers ship it, it joins the model catalog and any existing workflow can switch to it by swapping the model on the video node. Seedance 2.0 is available in Orisu today, so you can build and test the workflow now.
