Seedance 2.5 Release Date: When It Launched and Where to Use It

Seedance 2.5 was released on July 31, 2026, when ByteDance's Seed team launched it in the Jimeng AI and Doubao apps (Seed blog: Introducing Seedance 2.5). ByteDance first announced the model on June 23, 2026 at its Volcano Engine FORCE conference in Beijing (The Next Web). BytePlus then made it available to developers on ModelArk in supported markets in early August, in a post dated August 6, 2026 (BytePlus). So yes, Seedance 2.5 is out, and you can use it today.

Seedance 2.5 release date at a glance

At the June announcement, coverage said an enterprise beta was live and public launch was targeted for early July (The Next Web). The actual public release came on July 31, about four weeks after that target.

The full Seedance timeline

Seedance 2.5 is the fourth major generation of ByteDance's video model in about 14 months. Here's how the family got here, based on Wikipedia's Seedance history, Seed and BytePlus announcements.

1. June 2025: Seedance 1.0. ByteDance's first Seedance model generated 1080p video from text and image prompts, including multi-shot clips.
2. December 2025: Seedance 1.5 Pro. Audio and video were generated together in one model, with multilingual lip-sync and sound effects.
3. February 12, 2026: Seedance 2.0. Clips up to 15 seconds with image, video and audio references.
4. March 26, 2026: international rollout of Seedance 2.0 through CapCut in several markets, excluding the US, with real-face blocking and invisible watermarks.
5. April 2026: Seedance 2.0 on BytePlus ModelArk for enterprise and developer access in supported markets.
6. Mid-June 2026: Seedance 2.0 Mini joined Seedance 2.0 and 2.0 Fast as the series' lowest-cost tier (BytePlus docs).
7. June 23, 2026: Seedance 2.5 announced in Beijing.
8. July 31, 2026: Seedance 2.5 released in Jimeng AI and Doubao.
9. Early August 2026: Seedance 2.5 on BytePlus ModelArk for developers.

What shipped in Seedance 2.5

Here's what the released model does, from Seed's launch post and the BytePlus developer documentation (BytePlus Seedance 2.5 tutorial).

• 30-second generations. Up to 30 seconds per request, up from 15 seconds in Seedance 2.0. Seed says you can "extend twice" for longer stories (Seedance 2.5 page).
• 50 omni references. Up to 30 images, 10 video clips and 10 audio clips in one request, including audio-only references.
• Joint audio and video. Sound effects, ambience and speech are generated with the picture, with multilingual speech including English, Chinese, Spanish, Japanese, Korean, Arabic and Portuguese.
• Timestamp-level editing. Replace subjects, add or remove objects, or repair part of a clip at specific moments.
• Extension. Extend an existing clip forward or backward.
• White-model and green-screen workflows for previz and compositing.
• Draft mode. A quick 480p draft you can turn into a 1080p final within seven days.

Announced vs shipped: the 4K question

The June announcement coverage described Seedance 2.5 as native 4K with 10-bit color (The Next Web). The BytePlus API documentation for the released model lists 480p, 720p and 1080p output, with 1080p in 10-bit color and H.265 encoding. In the Seedance 2.0 series docs, BytePlus says 4K output is "supported only by Seedance 2.0." If a 4K master matters to you, check your platform's current resolution list before planning around it.

The reference count changed in the same way. The announcement cited 50 multimodal inputs including 3D models and style references. The shipped spec defines those 50 as 30 images, 10 videos and 10 audio clips, with 3D blockouts handled through white-model control.

Where can you use Seedance 2.5 now?

• Jimeng AI and Doubao (China): Seed's launch post points to Jimeng Web and Doubao Pro, under Video Generation.
• BytePlus ModelArk: API access in supported markets globally. BytePlus notes it isn't available in the United States.
• Third-party platforms: many creative apps now integrate Seedance 2.5. On Kenerate, it's in the Video Studio next to Seedance 2.0, Ken 1, Veo 3.1, Kling v3, Wan 3.0 and Sora 2 Pro.

Try Seedance 2.5 in the Video Studio

Safety measures at launch

BytePlus lists several safeguards for Seedance 2.5 (BytePlus):

• Visible watermarks and C2PA Content Credentials for AI transparency.
• Intellectual property protection for API and Experience Center users.
• Technical restrictions on generating video from real faces, with verification for authorized likeness use.
• A library of more than 10,000 virtual human assets for character work.

For creators, that means you can't upload a photo of a real person as a reference. Use original characters, virtual humans or stylized designs instead.

How to get ready for Seedance 2.5 projects

Seedance 2.5 rewards preparation more than earlier models did, because it can take in so many references at once.

1. Build a reference kit. Character sheets from several angles, product photos, a style frame, a motion clip and a voice or music sample.
2. Write in timed beats. A 30-second window needs structure: what happens at 0–5 s, 5–15 s and 15–30 s.
3. Name reference roles. Tell the model what each file controls: face and outfit, camera movement, or voice.
4. Draft, then finish. Use a short or low-resolution pass to check motion before rendering the final.

A starter prompt:

AI Prompt RecipeFlux.1 Pro
16:9
“A 25-second one-take scene. 0–6 s: wide shot of a lighthouse on a cliff at dusk, waves crashing below. 6–16 s: the camera glides up the tower to the lantern room as the beam switches on. 16–25 s: the beam sweeps across a stormy sea and the camera pulls back to the full coastline. Audio: wind, surf, a low cello drone building slowly. Cinematic, cool blue grade with warm lantern light.”
Copy PromptTry in Image Studio

AI Prompt RecipeFlux.1 Pro
16:9
“Use Image 1 for the mascot character's design and colors and Image 2 for the cafe interior. The mascot waves from behind the counter, slides a cup across to camera and gives a thumbs-up. Static camera, bright morning light, cafe ambience and a cheerful ukulele melody, 12 seconds, 9:16.”
Copy PromptTry in Image Studio

Seedance 2.5 or Seedance 2.0: which to use now?

Now that both are available, pick by the job rather than by version number.

Many creators combine them: rough out the shot list with a fast model, render long hero shots on Seedance 2.5, and use Seedance 2.0 when a deliverable has to be 4K. If you're new to the category, our AI video generator overview and text-to-video guide explain the basics, and the reference-to-video guide covers the multi-reference workflow Seedance 2.5 is built around.

What's next after Seedance 2.5?

Seedance 2.5 is the newest version on Wikipedia's Seedance timeline, and no release date for a Seedance 3.0 or other successor has been published. Seed's launch post acknowledges "room for improvement" in the physical plausibility of complex motion and in scenes with several interacting subjects, which hints at where future versions will focus. Until something new is announced, Seedance 2.5 is the current flagship, and Seedance 2.0 (with its 4K option) and Seedance 2.0 Mini (for quick previews) remain in the lineup.

For a deeper feature breakdown, read our Seedance 2.5 AI video generator guide. To compare it with other models, see Seedance vs Kling vs Veo vs Wan vs Runway, or start from the Seedance 2.5 overview. New Kenerate accounts get 25 free credits on sign-up, and credits never expire. Plan details are on the pricing page.

Bottom line

Seedance 2.5 was announced on June 23, 2026 and released on July 31, 2026, with developer access on BytePlus ModelArk following in early August. The shipped model delivers 30-second takes, 50 references, joint audio and timestamp editing. Plan around 1080p through the API rather than the 4K mentioned at the announcement. It's available now in ByteDance's apps, on ModelArk in supported markets, and on platforms like Kenerate.