Seedance 2.5: What’s New, Pricing, and API Access

TL;DR

Seedance 2.5 is ByteDance Seed’s audio-video generation model for longer, reference-controlled video. It adds native clips up to 30 seconds, up to 50 image/video/audio references at the model level, multi-round extension, and more precise editing. The API is available through both CometAPI and BytePlus. On CometAPI, create an asynchronous task with POST https://api.cometapi.com/v1/videos, save the returned task ID, then poll the retrieval endpoint for the result.

As of September 3, 2026, CometAPI lists Seedance 2.5 as available for text-to-video and image-to-video. BytePlus also documents dreamina-seedance-2-5-260628 as a supported Seedance 2.x model, so the earlier “coming soon” status is no longer current. The two services use different credentials, endpoints, model IDs, and billing.

What Is Seedance 2.5?

Seedance 2.5 is a next-generation joint audio-video model from ByteDance Seed. It can generate a complete clip from text, animate a reference image, and use image, video, and audio references to guide identity, motion, camera language, atmosphere, and sound. Its main appeal is not one isolated benchmark score; it is the ability to keep a longer creative sequence inside one generation workflow.

The model is aimed at production work that needs more continuity and control than a short concept clip: campaign creative, product storytelling, social video, previsualization, education, and other reference-led workflows. Developers should still test prompt adherence, subject consistency, audio sync, physical plausibility, queue time, and cost per accepted clip on their own material.

What Is New in Seedance 2.5?

Up to 30 Seconds in One Generation

Seedance 2.5 supports clips from 4 to 30 seconds. Thirty seconds matters because a creator can hold several story beats, camera moves, and transitions in one render instead of stitching many short clips and repairing continuity between them.

Up to 50 Multimodal References

ByteDance documents up to 30 images, 10 videos, and 10 audio clips at the model level. In a real production workflow, those references can represent product angles, character sheets, approved brand imagery, motion examples, camera references, music, or voice cues. API exposure differs by platform, so confirm which reference types the endpoint accepts before designing a pipeline.

More Precise Editing and Extension

The model adds multi-round extension plus reference-based, timestamp-level, green-screen, and camera-perspective editing controls. These capabilities are useful when a team needs to revise a shot without rebuilding the creative direction from scratch, although not every platform exposes every model-level control through the same endpoint.

ByteDance also highlights clay-render reference, motion reference, stronger scene continuity, and multi-round extension. These are meaningful workflow capabilities, but they are not a universal quality score. A team should still test subject consistency, physical plausibility, prompt adherence, audio sync, and failure rate on its own footage.

Is the Seedance 2.5 API Available?

Yes. CometAPI lists Seedance 2.5 for text-to-video and image-to-video and documents 4–30 second output at 480p or 720p. BytePlus also documents dreamina-seedance-2-5-260628 in its enhanced Seedance API. Availability is therefore confirmed on both routes as of September 3, 2026, although access requirements and endpoints differ.

Model ID note. CometAPI’s live model page and request example use seedance-2-5, while one support table in the API reference displays seedance-2-5-260628. BytePlus uses dreamina-seedance-2-5-260628. Query GET /v1/models with your CometAPI key before pinning a production configuration.

How Do I Access the Seedance 2.5 API?

Use the CometAPI base URL

https://api.cometapi.com/v1.

The core request is

POST /v1/videos

with multipart form data. Start with a four-second 720p test, save the returned task ID, then poll

GET /v1/videos/{id}

or use a webhook-supported workflow to receive the result.

Is API available?

Yes. As of September 2026, Seedance 2.5 is officially released, and you can access it through API providers rather than relying only on the Jimeng/Doubao interfaces. ByteDance announced the model on July 31, 2026, with API access initially described as coming through BytePlus ModelArk.

For developers who want the simplest integration, CometAPI is one practical route.

How to access Seedance 2.5 API with CometAPI

Create a CometAPI account and get an API key.

1. Use the CometAPI video endpoint: POST /v1/videos
2. Set the model ID to the currently documented Seedance 2.5 identifier. CometAPI documentation currently shows both seedance-2-5 and seedance-2-5-260628 in different places, so for production you should verify the ID returned by GET /v1/models.
3. Submit a video-generation task.
4. Save the returned task ID and poll GET /v1/videos/{id} until the task is completed.

A basic request looks like this:

curl https://api.cometapi.com/v1/videos \
  -H "Authorization: Bearer $COMETAPI_KEY" \
  -F 'model="seedance-2-5"' \
  -F 'prompt="A continuous cinematic shot through a rain-lit night market."' \
  -F 'seconds="4"' \
  -F 'size="1280x720"'

The API is asynchronous: the creation request returns a task ID, which you then use to retrieve the generated video. CometAPI Supports 4–30 second generation and 480p/720p output for its Seedance 2.5 integration.

A successful create call returns an id and an initial status such as queued. Save that ID. Poll GET /v1/videos/{id} every 10–20 seconds until the task reaches completed, failed, or error. When completed, download or re-host the signed video_url before it expires.

For image-to-video, attach a reference image with input_reference and describe its role in the prompt.

How Much Does Seedance 2.5 Cost?

The table below lists Seedance's official list prices, which are higher than CometAPI's actual charges.

Budget for attempts, not only accepted clips. At the model-page rates, a 10-second render is about $1.03 at 480p or $2.31 at 720p. If a workflow averages three renders before approval, generation cost is roughly three times the single-render estimate, before storage, moderation, delivery, or human review.

As of September 3, 2026,The Seedance 2.5 model in CometAPI gives resolution-specific default-group billing of $0.412 for 5 seconds, $0.824 for 10 seconds, and $1.24 for 15 seconds. The estimates below use the resolution-specific model-page rates; confirm the rate shown for your account before production. Details see also Seedance 2.5 API Pricing Guide.

What Can You Build with Seedance 2.5?

Brand Campaigns

Use approved product, character, color, motion, and audio references to explore campaign concepts that stay closer to an established visual system. A 30-second ceiling gives a team room for setup, product reveal, and closing beat in one generation.

Short Films and Previsualization

Longer continuous generations help teams block multi-shot sequences, test camera moves, and explore pacing before committing to full production. Multi-round extension can develop a sequence while preserving its existing direction.

Social Video

For trailers, explainers, and narrative social clips, the model can combine several beats without forcing every idea into a five-second shot. Test vertical sizes and subject continuity before scaling a template across a content calendar.

Product Visualization

Reference images can anchor recognizable product form, color, and placement while prompts control environment, motion, and camera behavior. Review every output for unauthorized design changes before publishing.

How to Test Seedance 2.5 Before Production

There is no industry-standard benchmark that can declare Seedance 2.5 the top choice across all production scenarios. A useful launch-day test should therefore separate vendor-published capability from what the current CometAPI route exposes and from what your team has actually measured.

Run each prompt at least three times with the same duration and size. Record success rate, queue time, render time, cost per accepted clip, and reviewer score. Do not compare a sngle attractive Seedance result with an average result from another model.

When Should You Use Seedance 2.5?

Best fit. Seedance 2.5 is most interesting when a longer clip, a controlled visual reference, and multi-shot continuity matter more than the lowest possible draft cost. Product spots, story-driven ads, previsualization, educational sequences, and reference-led creative development are natural starting points.

Not the first choice. Use another route when you require a currently documented 1080p or 4K output, need every advertised video/audio reference type exposed through one API today, or only need inexpensive short drafts. It is also premature to standardize on advanced editing features unless the exact endpoint you plan to use documents them.

Seedance 2.5 Alternatives

Wan 3.0

Consider Wan 3.0 when high-volume drafts and lower generation cost matter most. CometAPI’s public catalog lists text-to-video, image-to-video, and video editing, with a discounted starting rate of $0.04 per second as of September 3, 2026. Test it against Seedance 2.5 on motion quality, audio needs, reference fidelity, and continuity rather than choosing on price alone.

MiniMax H3

Consider MiniMax H3 when native audio, flexible reference control, and output up to 2K are more important than Seedance 2.5’s 30-second ceiling. CometAPI lists a discounted starting rate of $0.064 per second as of September 3, 2026. Compare both models with identical prompts and measure cost per accepted clip, since a lower per-second price can be offset by retries.

The practical selection rule is simple: compare cost per accepted clip, not cost per generated second alone. A cheaper model becomes expensive if it needs many retries; a premium model is wasteful if a short draft is all the workflow needs.

All availability, identifiers, parameters, sizes, and prices in this article were checked on September 3, 2026.

Conclusion

Seedance 2.5 is strongest when longer continuous clips, reference fidelity, and multi-shot continuity matter more than minimum draft cost — product spots, story-driven ads, previsualization, and reference-led creative work.

The API is available through both CometAPI and BytePlus as of September 3, 2026; CometAPI's asynchronous POST /v1/videos workflow is the simpler integration route, though you should verify the live model ID and confirm your account's rate before production.