Seedance 2.5
ByteDance's 30-second, 50-reference audio-video model. Generate with it here — and see what is actually confirmed about it, and what isn't.
Up to 30 seconds in a single generation at 480p or 720p, with dialogue, music and effects rendered in the same pass as the picture.
What ByteDance has confirmed about Seedance 2.5
Every figure below comes from a ByteDance-owned page — the Seed launch post or the official model page — or from the shipped API surface we generate against. Nothing here is secondary reporting.
4 to 30 seconds per generation
The model page states "Create videos up to 30 seconds in a single generation, with the option to extend twice," and the published API takes any whole second from 4 to 30. Seedance 2.0 documented a 4–15 second range.
Up to 50 reference assets
Images, video clips and audio in a single pass, addressed in the prompt as [Image1], [Video1], [Audio1]. Seedance 2.0's technical report documented 9 images, 3 video and 3 audio — 15 in total, not the 12 you'll see quoted.
Joint audio-video generation
Picture and sound are generated together rather than dubbed on afterwards, so dialogue in double quotes comes back lip-synced. ByteDance describes 2.5 as "building on" the unified architecture it introduced with Seedance 2.0.
480p or 720p output
The two tiers the API exposes. Not 4K: the native-4K figure circulating for 2.5 was announced for Seedance 2.0 at the same June 2026 event, and coverage merged the two releases.
Official Seedance 2.5 demos
ByteDance published these alongside the model. They are ByteDance's own selections, not independent tests — and not generated on Flixly.
Source: ByteDance Seed — Seedance 2.5 · © 2026 ByteDance Seed
Seedance 2.5 showcase reel
720p demo published alongside the model announcement.
Reference and camera-control demo
Illustrates reference-video interpretation and camera movement.
These are ByteDance's selects, not a typical result. Run your own prompt through Seedance 2.5 to see what it does with yours.
What ByteDance has not published
Most pages ranking for Seedance 2.5 state these as specifications. They are not in any ByteDance source.
Frame rate
No fps figure has been published for 2.5. The 24 fps you'll see traces to third-party documentation for Seedance 2.0 endpoints — and to the billing formula, which is not the same claim as a rendered frame rate.
Benchmarks
ByteDance published none, and Seedance 2.5 appears on no independent video leaderboard. Every quality claim in circulation is a vendor statement or a single reviewer's anecdote.
Native 4K
Still unpublished and still absent from the API, which offers 480p and 720p only. Any page listing a 4K figure for 2.5 has carried it over from the 2.0 announcement.
Seedance 2.5 vs Seedance 2.0
Both are live on Flixly, and neither replaces the other: 2.5 buys length and references, 2.0 still owns the high-resolution tiers.
| Spec | Seedance 2.5 | Seedance 2.0 (live on Flixly) |
|---|---|---|
| Max duration, one generation | 4–30 seconds | 4–15 seconds |
| Reference assets | 50, across images / video / audio | 15 (9 image / 3 video / 3 audio) |
| Audio | Joint audio-video | Joint audio-video |
| Output resolution | 480p or 720p | Up to 1080p, 4K on some routes |
| Frame rate | Not published | 24 fps in third-party endpoint docs |
| Benchmarks | None published | Listed on Artificial Analysis |
| Available on Flixly | Yes — T2V, I2V and R2V | Yes — Standard, Fast and Mini |
Three ways to run it
All three Seedance 2.5 modes are live, alongside the Seedance 2.0 family, Sora 2, Veo 3.1, Kling 3.0 and Wan 2.7.
Text to Video
A prompt in, up to 30 seconds out, with dialogue and score generated in the same pass.
Image to Video
Animate a start frame — and give it an end frame when the shot needs to land on a specific composition.
Reference to Video
Up to 50 image, video and audio references in one generation, each addressable from the prompt.
Frequently Asked Questions
Can I generate with Seedance 2.5 right now?
Yes. Seedance 2.5 is live on Flixly in all three modes — text to video, image to video, and reference to video. Pick it from the model selector on any of those pages, set your resolution and duration, and generate.
Does Seedance 2.5 support 4K?
No. The API offers 480p and 720p only. The widely repeated native 4K figure was announced for Seedance 2.0 in June 2026, not for 2.5. If you need higher resolution, Seedance 2.0 reaches 1080p and 4K on some routes and is also live here.
How long can a Seedance 2.5 video be?
Any whole second from 4 to 30, set with a slider. That is double Seedance 2.0's 15-second ceiling. Because the model is billed by output length, a 30-second clip costs roughly six times a 5-second one — the credit estimate on the page reflects your exact settings before you generate.
How much does Seedance 2.5 cost on Flixly?
It is priced by output length and resolution, and 480p costs less than half of 720p — so drafting at 480p and re-running the prompt you like at 720p is the cheapest way to work. One thing to watch in reference to video: a video reference adds its own length to what you are billed for, so a 20-second reference on a 6-second output is billed on 26 seconds. Image and audio references do not do this.
How many reference images does Seedance 2.5 accept?
Up to 50 references in total across images, video clips, and audio in one generation — roughly a 3.3x increase over Seedance 2.0 across every modality. You address them in the prompt as [Image1], [Video1], [Audio1]. More references does not automatically mean better output; dense reference sets have produced cross-modal errors in testing.
Should I use Seedance 2.5 or Seedance 2.0?
Use 2.5 when you want length, synchronized audio across a longer take, or a large reference set. Use 2.0 when you need resolution above 720p, since 2.5 does not offer it. Both share the same joint audio-video architecture and both are live on Flixly.