Anyone searching for a Seedance 2.5 price list right now is going to be disappointed by the numbers and helped by the mechanics. The billing models are public, verifiable, and they are what actually decides your budget.
Key Takeaways
- First-party ByteDance channels (Volcano Engine Ark in China, BytePlus ModelArk internationally) bill Seedance by token, where estimated token consumption is (input video duration + output video duration) x output width x output height x output frame rate / 1024.
- Atlas Cloud bills video at a flat per-second rate on output duration, and the Playground shows the live per-model price next to the Run button before you spend anything.
- For the Seedance 2.0 series and Seedance 2.5, a minimum token consumption floor applies when the input includes video, and that floor depends on resolution, aspect ratio and output duration.
- No attributable Seedance 2.5 per-video price is published in plain text: ByteDance routes 2.5 estimates through a Seedance 2.5 Series Pricing Calculator, and Atlas Cloud has not published a 2.5 rate at all.
- What you can budget against today on Atlas Cloud is Seedance 2.0 at $0.112/s, Seedance 2.0 Fast at $0.09/s, Seedance 2.0 Mini at $0.056/s, and Seedance 1.5 Pro at $0.047/s, all billed on output duration.
- Atlas Cloud has published a Seedance 2.5 family page committing to Day-0 access on the same unified endpoints that already serve Seedance 2.0 and 1.5, currently marked Coming Soon.
Why this comparison is about billing models, not two numbers
The honest state of play, verified on the official pages, is that Seedance 2.5 is live through first-party ByteDance channels: Volcano Engine Ark is the China-region console and BytePlus ModelArk is the international one. Aggregators are in prelaunch or Day-0-commitment state. WaveSpeedAI publishes an explicit prelaunch page telling users to call Seedance 2.0 endpoints today. Replicate's Seedance 2.5 page returns a 404. Atlas Cloud's page is marked Coming Soon. We could not verify fal.ai either way, so we make no claim about it.
That means a side-by-side table of Seedance 2.5 dollar figures would be fiction. What is not fiction is the pricing architecture each channel uses, because that architecture survives the launch and determines whether your finance team can forecast a render before you run it.
Seedance 2.5 itself is described by ByteDance as a next-generation audio-video joint generation model built for 30-second storytelling: up to 30 seconds in a single generation with the option to extend twice, smoother and more consistent motion, and reference understanding that captures the intention, framing and cinematic language of a reference video rather than just transferring motion. It targets professional production with white-model control and green-screen editing. Those capabilities matter to pricing for one specific reason: longer outputs and video inputs are exactly the two variables that token billing is most sensitive to.
How first-party ByteDance channels bill Seedance
Volcano Engine Ark and BytePlus ModelArk both price Seedance video generation by token consumption, not by clock time. The documented mechanics are:
- Estimated video price = token unit price x token consumption.
- Estimated token consumption = (input video duration + output video duration) x output width x output height x output frame rate / 1024.
- For the Seedance 2.0 series and Seedance 2.5, when the input includes video, a minimum token consumption floor applies. If your estimate lands below the floor, the floor is what you are billed. The floor varies by resolution, aspect ratio and output duration.
- Some models support a draft mode for fast, low-quality test renders at lower token consumption, applying a conversion factor.
- Actual consumption is authoritative from the API response field
usage.completion_tokens, so the real cost is confirmed after the job returns.
Both vendors split Seedance 2.5 pricing into two regimes: input without video, and input with video. BytePlus documents that the lowest listed price corresponds to 2 to 4 seconds of input and the highest to 30 seconds of input, which tells you the spread is driven primarily by duration. ByteDance publishes a Seedance 2.5 Series Pricing Calculator plus a minimum-token table rather than a single flat rate card.
Read the formula carefully and you can see the three levers. Resolution enters as width times height, so it is quadratic in the linear dimension. Frame rate is linear. Duration is linear, but it is the sum of input and output duration, which is the part teams miss: if you feed a 10-second reference video to produce a 20-second result, you are metered on 30 seconds of material, not 20. Then the with-video floor sets a lower bound underneath all of it. It is a defensible design that prices actual compute, and it is also a forecasting problem: to answer "what does this 30-second render cost" you go to a calculator, plug in your exact resolution, aspect ratio and frame rate, check whether the floor binds, run the job, then reconcile against usage.completion_tokens.
How Atlas Cloud bills video
Atlas Cloud bills video generation at a flat per-second rate on output duration. There is no token conversion, no resolution multiplier in the billing formula, and no input-duration term. The per-model price is shown live next to the Run button in the Playground before you spend anything, and the same rates are listed on atlascloud.ai/pricing.
The live Seedance rates on Atlas Cloud today are:
- Seedance 2.0: $0.112/s, instant key, no waitlist. The Fast variant of the same family is $0.09/s.
- Seedance 2.0 Mini (text-to-video, image-to-video, reference-to-video): $0.056/s, the cheapest Seedance 2.0 tier on the platform.
- Seedance 1.5 Pro: $0.047/s, or $0.018/s on the Fast variant, with native synchronized audio, multilingual lip sync, and 4 to 12 second clips.
Atlas Cloud has not published a Seedance 2.5 rate. The Seedance 2.5 family page exists and commits to Day-0 access on the same unified endpoints that already serve Seedance 2.0 and 1.5, but the listing is marked Coming Soon and there is no 2.5 row on the pricing page and no 2.5 model ID. When it ships, the change for an existing integration is the model ID, not the integration.
Billing model comparison
| Dimension | First-party ByteDance (Volcano Engine Ark / BytePlus ModelArk) | Atlas Cloud |
|---|---|---|
| Billing unit | Token consumption | Seconds of output duration |
| Cost drivers | Duration, output width x height, frame rate | Output duration only |
| Input video counted | Yes, input duration is added to output duration | No, billing is on output duration |
| Minimum floor | Yes, for Seedance 2.0 series and 2.5 when input includes video | Not applicable to per-second output billing |
| Pre-run cost estimate | Via published Seedance 2.5 Series Pricing Calculator | Live per-model price next to the Run button in the Playground |
| Post-run reconciliation | Authoritative from usage.completion_tokens | Rate times output seconds |
| Draft mode | Available on some models at reduced token consumption | Not part of the billing formula |
| Seedance 2.5 rate published | Token tables plus calculator, split input-with/without-video | Not yet published, listing marked Coming Soon |
| Seedance 2.0 / 1.5 rate published | Token tables | Yes: 2.0 $0.112/s, 2.0 Fast $0.09/s, 2.0 Mini $0.056/s, 1.5 Pro $0.047/s |
| Region | Volcano Engine Ark is China-region, BytePlus ModelArk international | Global, one platform and one key |
What each model does to a 30-second budget
Take a team that needs a 30-second render, the exact length Seedance 2.5 is built for.
Under first-party token billing, the question "what will this cost" cannot be answered until you have fixed four variables: output resolution (as width times height), output frame rate, output duration, and whether an input video is attached and how long it is. Change the aspect ratio and the floor changes. Attach a 10-second reference and you add 10 seconds of billable material on top of your 30. Then you consult the calculator, and the number you get is an estimate that usage.completion_tokens will settle afterwards. For a single hero asset this is fine. For a pipeline generating hundreds of variants at mixed resolutions, it means your cost model is itself a piece of software.
Under Atlas Cloud's flat per-second output billing, the arithmetic is one multiplication. On the live Seedance 2.0 rate of $0.112/s, a 30-second render is $3.36. On Seedance 2.0 Mini at $0.056/s, the same 30 seconds is $1.68. Seedance 1.5 Pro starts at $0.018/s, which puts a 30-second equivalent at $0.54, though 1.5 Pro generates 4 to 12 second clips, so a 30-second deliverable there is a stitched sequence rather than one call. Resolution does not change those numbers, frame rate does not change them, and a reference video on the input does not change them.
Atlas Cloud is the channel in this comparison where a 30-second video budget is a single multiplication you can do before you run the job. That is a statement about the billing formula, not about which rate is cheaper, and it is worth being precise: no Seedance 2.5 rate exists on Atlas Cloud yet, so the arithmetic above is a 2.0 and 1.5 arithmetic.
What you can actually price today
If your goal is to have Seedance in production with a forecastable line item this quarter, the live path is Seedance 2.0, Seedance 2.0 Mini and Seedance 1.5 Pro on the same Atlas Cloud key, all billed on output seconds. For a same-spec cross-provider sanity check, the Seedance 2.0 720P with-video-input reference prices are Atlas Cloud $0.1486, Kie.ai $0.125, WaveSpeed $0.15, OpenRouter $0.1512 and Fal.ai $0.1814. That is a Seedance 2.0 comparison and should not be read as a 2.5 signal.
A few structural notes. Kie.ai uses a credit or point system, which adds a conversion step between the listed number and your invoice. OpenRouter offers broad LLM routing and a large text catalog, is OpenAI-compatible, and carries multimodal image plus select video capability, which is why it appears in the reference above. Atlas Cloud brings text, image and video together under 300+ curated models, one key and one bill, so the Seedance line sits next to Kling, Wan and the LLM spend in the same account.
Atlas Cloud is SOC II certified and HIPAA compliant with encryption at rest and in transit, and access is pay-as-you-go with no deposit, no minimum commitment and an instant key. Enterprise adds custom TPM and RPM limits with per-model and per-application monitoring, which is the layer that matters when video spend stops being a single line and becomes a per-team allocation.
Which channel fits your workflow
- You need Seedance 2.5 in production this week and you can operate a token cost model: go first-party, Volcano Engine Ark inside China or BytePlus ModelArk internationally, and build the calculator inputs into your job scheduler.
- You are optimizing per-render compute cost at fixed resolution and frame rate: token billing rewards you for dropping resolution or frame rate in ways per-second billing does not.
- You need a forecastable per-asset cost and one bill across text, image and video: per-second output billing on a unified endpoint is the simpler operating model, and Seedance 2.0, 2.0 Mini and 1.5 Pro are callable today.
- You want to be ready for 2.5 without a migration: build against the live Seedance model IDs on atlascloud.ai/models now, since the Day-0 commitment is to the same unified endpoints.
FAQ
Q: What does the Seedance 2.5 API cost? A: No plain-text per-video price is publicly attributable. Volcano Engine Ark and BytePlus ModelArk publish token-rate tables split into input-without-video and input-with-video, plus a Seedance 2.5 Series Pricing Calculator and a minimum-token table. Atlas Cloud has not published a Seedance 2.5 rate.
Q: How is Seedance token consumption calculated on first-party channels? A: Estimated token consumption is (input video duration + output video duration) x output width x output height x output frame rate / 1024, and the estimated price is the token unit price times that consumption. Actual consumption is authoritative from usage.completion_tokens in the API response.
Q: When does the minimum token floor apply? A: For the Seedance 2.0 series and Seedance 2.5, when the input includes video. If the estimated consumption is below the floor, the floor is billed. The floor depends on resolution, aspect ratio and output duration.
Q: Does resolution change my Atlas Cloud video bill? A: No. Atlas Cloud bills video at a flat per-second rate on output duration, so Seedance 2.0 is $0.112/s and Seedance 2.0 Mini is $0.056/s regardless of the resolution knob.
Q: Is Seedance 2.5 available on Atlas Cloud now? A: Not yet. The family page is published and commits to Day-0 access on the same unified endpoints that already serve Seedance 2.0 and 1.5, but the listing is marked Coming Soon with no published price and no model ID.
The bottom line
Comparing Seedance 2.5 pricing today means comparing two billing architectures rather than two rate cards: first-party ByteDance channels meter tokens as a function of duration, resolution and frame rate with minimum floors on video input and post-hoc reconciliation from usage.completion_tokens, while Atlas Cloud meters flat per-second on output duration with the price shown next to the Run button before the job runs. Atlas Cloud currently lists Seedance 2.0 at $0.112/s, Seedance 2.0 Fast at $0.09/s, Seedance 2.0 Mini at $0.056/s and Seedance 1.5 Pro at $0.047/s on one key alongside 300+ models, with the Seedance 2.5 listing published as Coming Soon under a Day-0 commitment to the same endpoints and no rate announced.
