TL;DR
BytePlus lists Seedance 2.5 API pricing at $10.70 per 1 million tokens for requests without video input and $6.40 per 1 million tokens for requests with video input. Its published five-second, 16:9 examples cost $0.514 at 480p and $1.156 at 720p without video input.
Seedance 2.5 was officially launched on July 31, 2026, but ByteDance said API access through BytePlus ModelArk was still coming soon. Treat the model as launched and priced, but not yet confirmed as a generally available production API. Before integration, verify the account, region, callable model ID, endpoint parameters, and returned billing data.
Seedance 2.5 API Pricing at a Glance
| Item | Published information |
|---|---|
| Provider | ByteDance Seed / BytePlus ModelArk |
| Official BytePlus model ID | dreamina-seedance-2-5-260628 |
| Rate without video input | $10.70 per 1M tokens |
| Rate with video input | $6.40 per 1M tokens |
| Published output tiers | 480p and 720p |
| Five-second example without video input | $0.514 at 480p; $1.156 at 720p |
| Five-second example with video input | $0.553–$2.152 at 480p; $1.244–$4.838 at 720p |
| Failed-generation billing | BytePlus says only successfully generated videos are charged |
| API status | Pricing and model information are published; broad API access remains unconfirmed |
The token rate is only one part of the bill. Video-input requests use a lower rate, but the input-video duration is also included in estimated token consumption, so the final request can still cost more.
What Is Seedance 2.5?
Seedance 2.5 is ByteDance Seed’s multimodal video-generation model for longer storytelling, reference-driven creation, and targeted editing. ByteDance says it can generate up to 30 seconds of audio-video content in one pass, use up to 30 image references, 10 video references, and 10 audio references, and support timestamp-level editing and multi-round extensions.
For a broader feature overview, see What Is Seedance 2.5?. This guide focuses specifically on API pricing, billing, availability, and production budgeting.
Is the Seedance 2.5 API Available Now?
Seedance 2.5 was officially launched by ByteDance Seed on July 31, 2026. The official launch announcement says the model is rolling out through Jimeng AI, Doubao Pro, and other products, with API access coming soon through BytePlus ModelArk.
BytePlus has already published the model ID and pricing. However, a published rate card does not prove that every account or region can create a working endpoint. The current CometAPI Seedance 2.5 model page also describes enterprise API access as coming soon.
Before building a production dependency, confirm that:
- The model is selectable in the intended account and region.
- A real video-generation request completes successfully.
- The endpoint exposes the required duration, resolution, input types, references, and editing controls.
- The response returns usage and billing data that match the planned cost model.
Until those checks pass, use Seedance 2.5 for evaluation planning rather than as a guaranteed production route.
How Much Does Seedance 2.5 Cost?
The BytePlus ModelArk pricing page lists two online rates for Seedance 2.5:
| Request type | Published output resolutions | Price per 1M tokens |
|---|---|---|
| No video input | 480p and 720p | $10.70 |
| Includes video input | 480p and 720p | $6.40 |
These are usage rates, not flat prices per clip. Final cost depends on input-video duration, output duration, output dimensions, frame rate, and actual returned token usage.
How much does Seedance 2.5 cost without video input?
For a five-second, 16:9 output, BytePlus publishes these examples:
| Resolution | Output duration | Cost per video | Approximate cost per second* |
|---|---|---|---|
| 480p | 5 seconds | $0.51 | $0.10 |
| 720p | 5 seconds | $1.16 | $0.23 |
*The per-second figures are simple divisions of the official five-second examples. They are useful for comparison but should not be treated as guaranteed linear pricing for longer videos.
In this example, moving from 480p to 720p increases the cost by about 2.25×. Resolution therefore has a much larger effect on the final clip price than the headline token rate alone suggests.
How much does Seedance 2.5 cost with video input?
Requests with video input use the lower $6.40-per-million-token rate, but input duration is included in the usage calculation. For a five-second, 16:9 output, BytePlus publishes the following ranges:
| Resolution | Input-video duration | Output duration | Published price range |
|---|---|---|---|
| 480p | 2–30 seconds | 5 seconds | $0.553–$2.152 |
| 720p | 2–30 seconds | 5 seconds | $1.244–$4.838 |
At 720p, the top of the video-input range is about 4.18× the $1.156 no-video-input example. A lower per-token rate therefore does not guarantee a cheaper request.
BytePlus currently publishes Seedance 2.5 prices only for 480p and 720p. Do not create a 1080p or 4K production budget until those tiers and their endpoint availability are documented.
How Does Seedance 2.5 Billing Work?
BytePlus describes estimated video token consumption with this formula:
Estimated token consumption =
(input video duration + output video duration)
× output width
× output height
× output frame rate
÷ 1024
The estimated price is then:
Estimated price = token unit price × token consumption
The formula is useful for planning, but the final cost should be based on the usage returned by the API. Store usage.completion_tokens together with the model ID, input type, duration, resolution, frame rate, terminal status, and charged amount.
BytePlus also states that only successfully generated videos are charged and that failures such as moderation failures are not billed. Internal quality rejection is different: a technically successful clip that your team rejects is still a generated and charged output.
For production reporting, separate these outcomes:
- Provider failure or moderation failure
- Successful and accepted
- Successful but fixable
- Successful but rejected
This distinction makes it possible to calculate both provider reliability and the real cost of usable content.
What Does a Usable Seedance 2.5 Clip Cost?
List price measures one successful generation. Production teams usually care more about the cost of one accepted asset after retries and review.
Cost per accepted clip =
generation cost ÷ acceptance rate
+ review and revision cost
Consider a planning example using the five-second 720p price without video input:
| Assumption | Value |
|---|---|
| Published generation cost | $1.156 per successful output |
| Assumed acceptance rate | 70% |
| Assumed review cost | $0.30 per accepted clip |
| Generation cost per accepted clip | $1.156 ÷ 0.70 = $1.65 |
| Total cost per accepted clip | About $1.95 |
| Estimated budget for 100 accepted clips | About $195 |
The 70% acceptance rate and $0.30 review cost are planning assumptions, not Seedance benchmarks. Replace them with results from your own prompts, reviewers, and production standards.
This metric also prevents a common buying mistake: selecting the model with the cheapest attempt even when it requires more retries or manual correction.
Is Seedance 2.5 More Expensive Than Seedance 2.0?
Yes. Based on BytePlus’s comparable five-second, 16:9 examples without video input, Seedance 2.5 costs more at both published resolutions.
| Five-second output | Seedance 2.0 | Seedance 2.5 | Increase |
|---|---|---|---|
| 480p | $0.35 | $0.51 | 46.90% |
| 720p | $0.76 | $1.16 | 52.10% |
The premium is worthwhile only when Seedance 2.5 improves the economics of the full workflow. Compare both models with the same prompts and references, then measure acceptance rate, continuity, reference fidelity, revision time, and cost per accepted clip.
For short or high-volume draft generation, the lower-cost Seedance 2.0 model may remain the more practical route while Seedance 2.5 access and production performance are being validated.
Who Should Test Seedance 2.5?
Seedance 2.5 is most relevant when longer continuous scenes, many reference assets, or targeted edits matter more than the lowest possible generation price.
| Workload | Why test Seedance 2.5 | Main risk to measure |
|---|---|---|
| Narrative ads and short films | Up to 30-second single-pass storytelling and extensions | Character and scene continuity across transitions |
| Reference-heavy campaigns | Up to 30 images, 10 videos, and 10 audio references | Reference fidelity and input-driven cost |
| Film and advertising edits | Timestamp, green-screen, camera, and reference editing | Whether the callable API exposes each control |
| Product and ecommerce videos | More consistent campaign variants from brand assets | Product accuracy and reviewer workload |
| Synthetic training video | Longer controlled scenes and scenario variation | Physical plausibility and multi-subject stability |
ByteDance acknowledges that complex motion and interactions among multiple subjects still need improvement. Those cases should be part of the first evaluation set rather than assumed to be solved by launch demos.
How Should Teams Test Seedance 2.5 Before Production?
Use a small evaluation matrix that changes one cost variable at a time:
- Run the same text-led prompts at 480p and 720p for five seconds.
- Repeat one video-reference task using both a short input and a 30-second input.
- Include complex motion and multi-subject scenes.
- When available, test 30-second generation, extension, multimodal references, and timestamp editing separately.
- Record returned token usage, provider charge, queue time, generation time, moderation status, and accepted/fixable/rejected labels.
- Compare Seedance 2.5 with Seedance 2.0 using cost per accepted clip rather than cost per submitted request.
Video generation is asynchronous, so production systems should use task IDs, polling or callbacks, durable result storage, timeouts, and clear retry rules. See How to Add AI Video Generation to a SaaS App for the surrounding workflow architecture.
FAQ
What is the Seedance 2.5 API price?
BytePlus lists $10.70 per 1 million tokens without video input and $6.40 per 1 million tokens with video input for 480p and 720p output. Actual request cost depends on duration, dimensions, frame rate, input type, and returned token usage.
What is the official Seedance 2.5 model ID?
The BytePlus model ID is dreamina-seedance-2-5-260628. CometAPI lists the catalog identifier doubao-seedance-2-5. Use the identifier required by the provider or gateway you are calling.
Is the Seedance 2.5 API available now?
Seedance 2.5 is officially launched and its pricing is published, but ByteDance’s July 31 announcement said BytePlus ModelArk API access was still coming soon. Confirm a successful request in the intended account and region before integration.
Can Seedance 2.5 generate 30-second videos through the API?
ByteDance documents up to 30-second generation as a Seedance 2.5 model capability. Whether a specific API route exposes that duration must be confirmed from the live endpoint and a real test request.
Are failed Seedance 2.5 generations charged?
BytePlus says only successfully generated videos are charged and failures such as moderation failures are not billed. A successful clip rejected by your internal reviewer remains a charged generation.
Check Seedance 2.5 Availability Through CometAPI
The CometAPI Seedance 2.5 model page currently lists the model while describing enterprise API access as coming soon. Check the live model page and API documentation before implementation.
Once the route is enabled and verified, begin with a five-second 480p and 720p baseline. Record the returned usage, actual charge, latency, and acceptance result before adding longer outputs or video-reference inputs.