Seedance 2.5 from ByteDance is officially available on Apiframe.

Seedance 2.5: The Complete Guide (2026)

ByteDance's Seedance 2.5: 30-second clips, 50 references, up to 720p, and synced audio. Now live on Apiframe, what changed from the original announcement, and how to get access.

Renaud Last updated November 1, 2026 July 2, 2026 · 8 min read Beginner
Seedance 2.5: The Complete Guide (2026)

Seedance 2.5 is ByteDance's next-generation AI video model, and it is now live: single clips up to 30 seconds generated in one pass, up to 50 references in a single generation, resolution up to 720p, and audio created together with the picture instead of added afterward. This guide covers what it is, what changed compared with Seedance 2.0, how pricing and access work, and how to get started. If you landed here from a comparison post, a tutorial, or a search for "Seedance 2.5," this is the page those all point back to.

Current status

Seedance 2.5 is live. ByteDance announced it on June 23, 2026, at the Volcano Engine FORCE conference in Beijing, opened wider API access through BytePlus in mid-July, and launched the model on July 31, 2026. Apiframe added Seedance 2.5 to its unified API on August 7, 2026, so you can reach it today through the same endpoint, key, and billing as every other model on the platform.

One correction worth flagging: ByteDance's original announcement promised native 4K output, but the model that shipped tops out at 720p, with 480p and 720p as the two available resolutions. We have corrected the resolution claims throughout this guide to match what the API actually offers rather than the pre-launch announcement. Everything else ByteDance announced held up as advertised: 30-second clips in a single pass, up to 50 references, audio generated in sync with the picture, editing that targets one part of the frame, and shot-by-shot camera control.

What is Seedance 2.5?

Seedance 2.5 is ByteDance's next-generation multimodal AI video model, and it is a bigger change than the version number suggests. ByteDance went from 2.0 straight to 2.5 rather than shipping a small 2.1 update, and the feature list explains why. If you are weighing it against the earlier version, our Seedance 2.0 guide covers what that model offers on its own, and our explainer on how text-to-video generation works covers the basics behind both.

The model is served through Doubao and Volcano Engine in China, and through BytePlus ModelArk and the Dreamina app elsewhere. Through an API, it handles text-to-video, image-to-video, and reference-to-video behind a single endpoint. Generation runs in the background: you send a request, then either check on it or wait for a webhook to tell you the result is ready. Our walkthrough on setting up webhook callbacks for media generation shows how to wire that up.

What's new: full capability breakdown

30-second clips in a single pass

Most longer AI videos today are really several short clips joined together, and every join introduces small problems: a character's face drifts, the lighting jumps, the motion stutters. Seedance 2.5 generates one continuous clip of up to 30 seconds in a single pass, with scene changes and shifts in pace built into that same generation, so there are no joins for those problems to show up at. If you want a consistent way to judge whether that holds up on your own footage, see our guide to testing the quality of AI-generated video.

Up to 50 references in one generation

A single generation can take up to 50 references, mixing images, short video clips, and audio. That is a large jump from around 12 on the previous model. You can tag each reference with a role, so the model knows which input should fix a character's identity, a product's look, a brand's colors, a camera style, or a voice. The same thinking applies to still images, which we cover in our guide to using reference images in AI image generation.

Up to 720p resolution with synchronized audio

Seedance 2.5 renders at up to 720p, with 480p and 720p as the available options. That is lower than ByteDance's original announcement promised (native 4K and 10-bit color), a gap worth flagging, because it trips people up if they build around the announcement instead of the shipped spec. Audio is created in the same pass as the video rather than dubbed on afterward, so music, sound effects, dialogue, and lip-sync all land in sync with the picture by default. For a wider look at which models do this, see our roundup of APIs that support synchronized audio with generated video.

Camera direction and multi-shot control

You can describe a sequence of shots and the camera moves you want (pans, dollies, orbits, focus changes) directly in the prompt, and the model keeps the subject consistent across those cuts. ByteDance reports roughly 20% better prompt adherence than Seedance 2.0, which shows up most clearly in how reliably the camera does what you asked for. For more on how wording changes the result, see our piece on how prompt writing affects photorealistic output.

Editing one part of the frame

You can change a specific area or element of a clip while the rest of the frame stays put, or use an existing clip as a reference and change only what you name, keeping its character, setting, camera movement, framing, pacing, and length intact. Fixing one detail no longer means regenerating the whole clip and hoping for the best. Our AI video editing API guide covers this kind of workflow in more depth.

Rough 3D previews before the full render

Seedance 2.5 can also block out a scene as a plain, untextured 3D pass before you commit to a full render, so you can settle layout, framing, and motion early rather than discovering problems at the end.

Seedance 2.0 vs Seedance 2.5

ByteDance has been fairly specific about what changed. Here is how the two models line up side by side.

CapabilitySeedance 2.0Seedance 2.5
Max clip length4 to 15 seconds4 to 30 seconds, in one continuous pass
Reference inputsUp to around 12Up to 50 (images, video, and audio combined)
ResolutionUp to 1080p480p or 720p (ByteDance's announcement promised native 4K; that did not ship)
AudioGenerated in sync, over shorter clipsGenerated in sync, sustained across the full 30 seconds
EditingRegenerate the whole clip to change any partChange one part of the frame and keep the rest
Camera controlPrompt-based, more limitedMulti-shot direction (pans, dollies, orbits)
Prompt adherenceBaselineAbout 20% better, per ByteDance

This table originally reflected ByteDance's pre-launch announcement. It now matches the specs Seedance 2.5 actually shipped with, and resolution is the one meaningful correction. We will keep it current as more real-world testing comes in.

Pricing

Seedance 2.5 is billed per second: cost rises with duration and resolution, and you only pay for successful generations. To work out what that means at your volume, see our guide to estimating the monthly cost of AI video generation.

ByteDance has not published an official per-second rate for Seedance 2.5, as it did for Seedance 2.0 (around $0.06 per second on standard tiers, and roughly $0.022 per second on faster, lower-cost tiers through third-party providers). Through Apiframe, the confirmed rate is 15 credits ($0.15) per second at 480p and 34 credits ($0.34) per second at 720p. Since Seedance 2.5 caps at 720p while Seedance 2.0 reaches 1080p, check the current numbers rather than assuming the newer model automatically costs more. For a broader comparison across video models, see our AI video API pricing breakdown.

How to access Seedance 2.5

There are two practical routes, and both are live now.

The official route runs through BytePlus ModelArk or Volcano Engine directly. That is the right path if you already work inside ByteDance's ecosystem or need a direct enterprise relationship.

The quicker route is a unified API: one key, no separate ByteDance or BytePlus account, with background jobs and webhooks built in. Apiframe's Seedance 2.5 API is live now, and because billing and authentication are shared across every model on the platform, you are not managing a separate account just for this one. Get an API key to start generating today, or read the Seedance 2.5 API docs for the full request and parameter reference. If you want a general primer on how video APIs work before going deeper into Seedance, our AI video API guide is a good place to start.

What Seedance 2.5 is good for

A 30-second story in one pass, a large allowance for references, and audio generated with the picture make Seedance 2.5 a strong fit for a few specific jobs.

Longer stories and consistent characters benefit most directly. Holding a character's identity steady through movement, angle changes, and scene changes matters for short films, brand spots, and serialized content, and that is exactly what the reference system and single-pass generation are built for. Content with synchronized audio (dialogue, voiceover, music, and sound effects made in the same pass as the picture) removes a whole production step. Product video at scale is another strong fit: a set of references built from product photos, brand colors, and a voice sample can drive several on-brand variants without a film crew, and editing one part of the frame lets you swap a label or a color per market without regenerating the clip. Our AI product photography API guide covers the still-image side of that workflow, and our walkthrough on building a cost-effective video ad with AI APIs shows how the pieces fit together. Localization works the same way, since audio and on-screen text can be generated in many languages from a single setup.

For a closer look at how Seedance 2.5 stacks up against the rest of the field, our AI video generation API guide and our roundup of the best AI video generation APIs compare capability and cost. You can also read it against Veo 3.1 in our Veo 3 API guide and against Kling in our Kling 3.0 guide.

More on Seedance 2.5

This page is the starting point for everything Seedance 2.5 on Apiframe. A few related resources, some live now and some still to come.

  • Live: a text-to-video API tutorial in Python, with working code you can adapt to Seedance 2.5
  • Live: an image-to-video API tutorial in Python, if you are starting from a still image
  • Coming: a migration-focused Seedance 2.0 vs 2.5 comparison for teams already building on 2.0
  • Coming: a dedicated Seedance 2.5 API pricing page, now that the model has shipped
  • Coming: Seedance 2.5 for product video and brand content at scale

Until those go live, the sections above cover the same ground.

FAQ

What is Seedance 2.5?

Seedance 2.5 is ByteDance's next-generation multimodal AI video model, announced on June 23, 2026, and launched on July 31, 2026. Its headline features are 30-second clips in a single pass, up to 50 references, resolution up to 720p, editing that targets one part of the frame, multi-shot camera direction, and audio generated in sync with the picture.

Is Seedance 2.5 out yet?

Yes. Seedance 2.5 launched on July 31, 2026, and Apiframe added it to its unified API on August 7, 2026.

How is it different from Seedance 2.0?

The core differences are clip length (15 seconds up to 30, in one continuous pass), the number of references (about 12 up to 50), and editing (regenerating the whole clip versus changing one part of the frame). Resolution moved the other way: Seedance 2.0 tops out at 1080p, while Seedance 2.5 currently caps at 720p, because ByteDance's original 4K announcement did not make it into the shipped model. See the comparison table above for the full breakdown.

How much does the Seedance 2.5 API cost?

It is billed per second of generated video: 15 credits ($0.15) per second at 480p and 34 credits ($0.34) per second at 720p, confirmed through Apiframe. Check the Seedance 2.5 API page for current pricing.

How do I get API access?

Through BytePlus ModelArk or Volcano Engine directly, or through a unified API like Apiframe, which gives you a single key and endpoint without a separate ByteDance account. Both routes are live now.

Does it generate audio?

Yes. Audio is created in the same pass as the video, so music, sound effects, dialogue, and lip-sync arrive in sync with the picture rather than as a silent clip you score afterward.

💡
Seedance 2.5 is live on Apiframe now: 30-second clips in a single pass, up to 50 references, and audio generated in sync, at up to 720p. Get your API key and start generating today.

Power your next AI product with Apiframe.

Instant access to 70+ media models through a single API. Start free and scale when you're ready.