Status checked: August 20, 2026.
Short answer: Seedance 2.5 is officially released and available through LinkModel. LinkModel users should select the public model ID seedance-2-5; the integration follows the same authentication, task submission and result-handling pattern as Seedance 2.0. ByteDance also has a live Seedance 2.5 model page, while Volcano Engine documents its own API access.
| Access path | Current status | What it means |
|---|---|---|
| LinkModel API | Available now | Call Seedance 2.5 with model ID seedance-2-5 |
| ByteDance model page | Released | The official page presents Seedance 2.5 as the current generation of the model |
| Dreamina | Available | Seedance 2.5 is live for consumer creation through Dreamina |
| Volcano Engine API | Available | Volcano Engine provides its own asynchronous video-generation API |
When was Seedance 2.5 released?
ByteDance previewed Seedance 2.5 at the Volcano Engine FORCE conference on June 23, then officially launched Seedance 2.5 on July 31, 2026. Access dates still vary by product and API channel, so the formal model release and platform availability should remain separate events.
| Milestone | Confirmed status |
|---|---|
| June 23, 2026 | Seedance 2.5 previewed at FORCE |
| July 31, 2026 | ByteDance officially launched Seedance 2.5 |
| August 20, 2026 status check | Official model, Dreamina, Volcano Engine API and LinkModel access are live |
| LinkModel model ID | seedance-2-5 |
This distinction matters for the query “Seedance 2.5 release date.” The model is released, but access still depends on which product or API platform you intend to use.
What is new in Seedance 2.5?
ByteDance describes Seedance 2.5 as a new-generation audio-video model focused on longer narratives, precise multimodal references and editable output. The release turns three preview features into documented capabilities.
1. Up to 30 seconds in one generation
Seedance 2.5 can generate a single video between 4 and 30 seconds. When duration is set to -1, the model chooses a suitable length from the prompt and reference material. ByteDance also highlights multi-round extension for building longer sequences while preserving the source video's aspect ratio.
<figure> <video src="https://static.linkmodel.ai/hero-asset/seedance/seedance2.5-4.mp4" autoPlay muted loop playsInline className="w-full rounded-2xl" ></video> <figcaption> ByteDance's Seedance 2.5 demonstration shows a 30-second narrative generated as one continuous result. This is an official demonstration, not a LinkModel test. </figcaption> </figure>
2. Up to 50 multimodal reference assets
The official API limits make the “50 references” headline more concrete. A full-modal reference task can combine up to 30 images, 10 videos and 10 audio files. Reference video and audio totals are each capped at 30 seconds, and every file still has its own format and size requirements.
This larger input budget supports ensemble scenes, product references, storyboards, music cues and production assets in one request. Seedance 2.5 uses the prompt to determine whether those inputs describe reference generation, editing or extension.
<figure> <video src="https://static.linkmodel.ai/hero-asset/seedance/seedance2.5-5.mp4" autoPlay muted loop playsInline className="w-full rounded-2xl" ></video> <figcaption> ByteDance demonstrates multi-asset control for ensemble scenes and film previsualization. This is official demonstration footage, not a LinkModel benchmark. </figcaption> </figure>
3. Precise video editing and extension
Seedance 2.5 can replace or remove subjects, redraw a local region, change audio and extend an existing video. The official documentation adds task-level controls such as omni_reference_task_type, which can explicitly identify an edit or extend request and move some parameter errors to request time.
Editing is not a completely free-form request. The API keeps the input video's aspect ratio, requires adaptive duration, and accepts an editing source between 4 and 30 seconds. These constraints are important for developers migrating an existing video pipeline.
<figure> <video src="https://static.linkmodel.ai/hero-asset/seedance/seedance2.5-7.mp4" autoPlay muted loop playsInline className="w-full rounded-2xl" ></video> <figcaption> ByteDance demonstrates changing a local element while preserving the broader shot. This is official demonstration footage rather than a hands-on LinkModel result. </figcaption> </figure>
Is the Seedance 2.5 API available?
Yes. Seedance 2.5 is available through LinkModel. If you are already integrated with Seedance 2.0, keep the same API workflow and select this LinkModel model ID:
seedance-2-5You can review availability and platform-specific details on the LinkModel Seedance 2.5 model page. The model itself also introduces task-specific constraints for editing, extension, duration and aspect ratio, so production requests should be checked before migrating.
Vendor note: Volcano Engine's official Seedance 2.5 API tutorial uses doubao-seedance-2-5-260628. That identifier belongs to Volcano Engine; do not use it as the LinkModel model ID.
| Volcano Engine API setting | Seedance 2.5 value |
|---|---|
| Output duration | 4–30 seconds, or -1 for adaptive length |
| Output resolution | 480p and 720p in 8-bit; 1080p in 10-bit H.265/HEVC |
| 4K API output | Not currently documented |
| Aspect ratio | 16:9, 4:3, 1:1, 3:4, 9:16, 21:9 or adaptive |
| Output format | MP4 or MOV |
| Individual account limits | 180 RPM and 3 concurrent tasks |
| Enterprise account limits | 600 RPM and 10 concurrent tasks |
| Task record retention | 7 days |
| Generated video URL | Available for 24 hours, up to 100 downloads |
The resolution row deserves special attention: Volcano Engine added 1080p to its August 17 documentation update, but LinkModel's current public pricing table lists 480P and 720P. Upstream support does not prove that every gateway exposes the same request field, and no reviewed Seedance 2.5 endpoint documents 4K output.
How do I use Seedance 2.5 on LinkModel?
Seedance 2.5 is live on LinkModel. The integration uses the same overall shape as Seedance 2.0: existing authentication, asynchronous task submission and result handling remain familiar. Select seedance-2-5 as the model when creating the task.
Use the Seedance 2.5 model page as the source of truth for LinkModel pricing, availability and current request options. Platform model names are not interchangeable: copy seedance-2-5 from LinkModel rather than a vendor-specific identifier from another API.
Seedance 2.5 release and API FAQ
When was Seedance 2.5 released?
ByteDance officially launched Seedance 2.5 on July 31, 2026, after previewing it at FORCE on June 23. Product and API access dates vary by platform; by this article's August 20 status check, LinkModel and Volcano Engine API access were live.
Is the Seedance 2.5 API available?
Yes. Seedance 2.5 is available through LinkModel with the model ID seedance-2-5; Volcano Engine also documents its own asynchronous video-generation API. The official sources cover text-to-video, multimodal reference generation, video editing and extension.
Which Seedance 2.5 model ID should I use?
On LinkModel, use seedance-2-5. Model identifiers are platform-specific, so copy the value shown on the LinkModel model page rather than reusing an identifier from another provider.
Does the Seedance 2.5 API support 4K output?
No documented Seedance 2.5 endpoint supports 4K. Volcano Engine currently lists 480p and 720p in 8-bit plus 1080p in 10-bit H.265/HEVC, while LinkModel's public pricing table currently lists 480P and 720P.
Is Seedance 2.5 available through LinkModel?
Yes. Seedance 2.5 is available on LinkModel with the public model ID seedance-2-5. Existing Seedance 2.0 integrations can retain the same overall workflow and switch the selected model.
The bottom line
Seedance 2.5 is now a released model available through LinkModel and Volcano Engine, not a future announcement. Its practical gains are longer 30-second generations, up to 50 multimodal references, precise editing and extension. For API teams, the equally important facts are the platform-specific model ID, the current documented resolution limits and the task-specific migration constraints.
To start on LinkModel, open the Seedance 2.5 model page and use seedance-2-5 in your existing video-generation workflow.
For the next step, use Seedance 2.5 vs 2.0 to decide whether to migrate, then turn the selected workflow into a timestamped brief with the Seedance 2.5 prompt guide.
Primary sources: ByteDance's official Seedance 2.5 launch, Seedance 2.5 model page, and Volcano Engine's Seedance 2.5 API tutorial.

