Model Overview · Updated August 29, 2026
Seedance 2.5: Everything Confirmed So Far
Seedance 2.5 launched globally on July 31, 2026 and is live now on Dreamina with free daily credits — native 30-second single-segment video, up to 50 multimodal references, and audio generated together with the image. Here is what is confirmed, what changed versus 2.0, and how to start.
Seedance 2.5 at a Glance
Confirmed details from ByteDance’s announcement and the July 31, 2026 global launch — no speculation.
| Announced | June 23, 2026 — Volcano Engine FORCE conference |
| Developer | ByteDance (Seed team) |
| Max clip length | 30 seconds standard, single segment — up to 180 seconds in beta long-video mode |
| Reference inputs | Up to 50 full-modal reference materials |
| Audio | Joint audio-video generation in one latent space |
| Editing & control | Region-level editing without regenerating · R2V (reference-to-video) motion control |
| Status | Live — launched July 31, 2026 on Dreamina |
| Public access | Dreamina (global) · free daily credits; OpenArt hosts it; Runway announced coming soon |
| API pricing | OpenRouter: ≈ $0.2311/sec at 720p · ≈ $0.1028/sec at 480p (verified Aug 29) |
Details as announced at the Volcano Engine FORCE conference (June 23, 2026) and confirmed on the official Dreamina page at the July 31, 2026 global launch.
The Three Headline Changes
Seedance 2.5 is not an incremental quality bump — each of these changes how creators plan a generation.
Native 30-Second Video
Previous Seedance versions generated short clips that creators stitched together for longer stories. Seedance 2.5 outputs a continuous 30-second single-segment video in one generation — one narrative arc, consistent characters and lighting throughout, no cut-point artifacts.
50 Multimodal References
Seedance 2.5 accepts up to 50 reference materials — images, video clips, and audio — in a single generation. This is a step change for brand consistency: product shots, character sheets, style frames, and voice references can all steer one output.
Unified Audio-Video Generation
Instead of generating video first and adding synchronized audio afterwards, Seedance 2.5 co-processes visual and audio signals inside the same latent space. Dialogue, ambient sound, and music are generated together with the moving image.
Seedance 2.5 vs Seedance 2.0
What actually changes for creators, side by side.
| Dimension | Seedance 2.0 | Seedance 2.5 |
|---|---|---|
| Max single-generation length | ≈ 5–10 seconds per clip | 30 seconds, single segment |
| Reference materials | Limited image references | Up to 50 full-modal references (image / video / audio) |
| Audio | Video-first output | Native joint audio-video generation |
| Resolution | Native 4K (June 2026 upgrade) | 4K video output |
| Availability | Public (Dreamina, Jimeng, API providers) | Public (Dreamina, Jimeng, OpenRouter, fal.ai) |
| Pricing | Published via platforms/API | Dreamina free credits · OpenRouter ≈ $0.2311/sec at 720p |
What 30-Second Native Video Changes About Prompting
With 5–10 second clips, a prompt describes one shot: one subject, one camera move, one lighting setup. A 30-second single-segment generation is closer to a scene — it can hold an establishing moment, a development, and a payoff inside one continuous take.
That rewards prompts written in beats: what happens in the opening seconds, what shifts in the middle, and how the shot resolves — while keeping subject, style, and color grade constant so the model holds consistency across the full duration. The structured prompt formula (subject, action, camera, lighting, mood, style, duration, aspect ratio) still applies; the action element simply becomes a short sequence instead of a single motion.
PromptMotion’s generator already builds structured prompts in this format — describe your idea in plain language and it produces the full structure. See the Seedance 2.5 how-to guide for beat-by-beat prompt examples.
Seedance 2.5 Resources
Use Seedance 2 Free Today
The doors that are open right now — free daily credits and what they include.
Read →
Try Seedance 2.5 Free
Every legitimate free access path — official platforms and in-site generation.
Read →
Seedance 2.5 Release Date
Confirmed timeline: announcement, enterprise beta, and the July 31 global launch.
Read →
Seedance 2.5 API Pricing
Live endpoints, current per-second cost, supported inputs, and integration flow.
Read →
How to Use Seedance 2.5
Step-by-step workflow and prompt structure for 30-second generations.
Read →
Gemini Omni Flash 1.1
Compare the new Google model, native audio, live API, current pricing, and in-site generation.
Read →
Camera Movement Prompts
Named camera techniques and copy-ready motion language for more controlled video.
Read →
Seedance Prompt Guide
The structured prompt formula that carries over to 2.5.
Read →
Frequently Asked Questions
What is Seedance 2.5?
Seedance 2.5 is ByteDance's next-generation AI video model, announced on June 23, 2026 at the Volcano Engine FORCE conference. Its headline capabilities are native 30-second single-segment video generation (no stitching), support for up to 50 full-modal reference materials, and a unified architecture that generates audio and video jointly in the same latent space instead of synchronizing them after generation.
Is Seedance 2.5 available to the public?
Yes. Seedance 2.5 launched globally on July 31, 2026 and is live on Dreamina with free daily credits. Jimeng serves China. Public developer access is available through OpenRouter, fal.ai, and first-party ByteDance channels; OpenRouter and fal.ai pricing was rechecked August 29, 2026.
What is the difference between Seedance 2.0 and Seedance 2.5?
Seedance 2.0 generates shorter clips (typically 5–10 seconds per generation) with video-first output. Seedance 2.5 generates up to 30 seconds in a single pass, accepts up to 50 reference materials across image, video, and audio modalities, and produces synchronized audio natively. Seedance 2.0 also received a native 4K upgrade at the same June 2026 announcement.
Do Seedance 2.0 prompts work with Seedance 2.5?
The structured prompt principles carry over: subject, action, camera movement, lighting, style, duration, and aspect ratio remain the core elements. Seedance 2.5's longer 30-second format rewards multi-beat prompts that describe a sequence of moments rather than a single shot. PromptMotion's prompt generator builds this structure automatically.
Can I try Seedance-style video generation on PromptMotion?
Yes. PromptMotion offers three free structured prompts per day with no credit card. Pro users can render in-site with Seedance 2.5 Quality or Seedance 2.0 Fast through a third-party API provider. PromptMotion is an independent tool and is not affiliated with ByteDance.
Get the Most Out of Seedance 2.5
Master the structured prompt format now — the same formula carries straight into 2.5’s 30-second generations. 3 free prompt generations per day. No credit card required.
PromptMotion is an independent third-party AI prompt optimization tool and is not affiliated with, endorsed by, or associated with ByteDance Ltd. “Seedance” is a trademark of ByteDance Ltd. Model details on this page reflect public announcements as of the July 31, 2026 global launch and may change as the rollout continues. PromptMotion provides optional Seedance 2.5 and Seedance 2.0 rendering through a third-party API provider.