Seedance 2.5 API on Atlas Cloud: Pricing, Availability, and Quickstart

A new ByteDance video model lands and the engineering questions arrive in the same order every time. Can I call it yet? What will a minute of output cost me? And how much of my pipeline has to change when I switch? Seedance 2.5 is at the awkward stage where the model is real and documented by its creator, the demand is real, and the answers you find on aggregator pages are a mix of live listings, prelaunch trackers, and outright guesses.

This page is the unglamorous version. It states exactly where Seedance 2.5 stands on Atlas Cloud today, explains the two billing models you are actually choosing between, gives you live per-second numbers you can budget against right now, and hands you an integration you can run this afternoon that will not need rewriting when 2.5 arrives.

What Seedance 2.5 actually is

Per ByteDance's own model page, Seedance 2.5 is a next-generation audio-video joint generation model built for 30-second storytelling, with precise reference control and strong editing capability. The headline points from the first-party source:

• Up to 30 seconds in a single generation, with the option to extend twice for longer, more complete narrative arcs
• Smoother, more consistent motion and more realistic visuals than the previous generation
• Smarter reference handling: it reads a reference video for its intention, framing, and cinematic language rather than merely transferring motion
• More reliable editing that responds to a wider range of audio and visual editing requests
• Production-oriented features such as white-model control and green-screen editing, plus professional camera movement and performance blocking

That is the whole verified capability story. You will see richer spec sheets circulating on third-party pages. Treat anything beyond the list above as unconfirmed until ByteDance publishes it, because a spec that turns out to be wrong is the kind of thing that quietly breaks a production plan two sprints later.

Availability: where you can call Seedance 2.5 today

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. At the time of writing, that listing is marked Coming Soon, there is no Seedance 2.5 row on the Atlas Cloud pricing page, and no Atlas Cloud price or model ID for 2.5 has been published yet. Anyone telling you otherwise is reading a marketing page as a changelog.

The broader landscape, checked directly rather than inferred:

So the honest global answer is that Seedance 2.5 access today runs through first-party ByteDance channels, Volcano Engine Ark inside China and BytePlus ModelArk internationally, while aggregators sit in prelaunch or Day-0-commitment state. If you need to touch the model this week, that is your route. If you need it inside an existing multi-model application, the preparation work described below is what pays off.

Pricing: two billing models, one of them you can budget today

The reason nobody can hand you a clean cost-per-clip number for Seedance 2.5 is that the two channels bill on fundamentally different units.

First-party token billing

Volcano Engine Ark and BytePlus ModelArk bill Seedance by token, not by second. The published mechanism works like this:

• Estimated video price equals token unit price multiplied by token consumption
• Estimated token consumption equals (input video duration plus output video duration) times output width times output height times output frame rate, divided by 1024
• Some models support a draft mode for fast, low-quality test renders at reduced token consumption, applying a conversion factor
• Actual, authoritative consumption comes back on the API response in usage.completion_tokens
• For the Seedance 2.0 series and for 2.5, when the input includes video a minimum token consumption floor applies: if your estimate falls below the floor, the floor is billed. The floor varies by resolution, aspect ratio, and output duration
• ByteDance publishes a Seedance 2.5 series pricing calculator alongside a minimum-token table, and both vendors split the rate card into input without video and input with video, with the lowest listed price corresponding to 2 to 4 seconds of input and the highest to 30 seconds

The practical read: your cost driver is duration times resolution times frame rate, and reference-video workflows carry a floor you cannot estimate away. It is a defensible model, but it is not arithmetic you do in your head before a standup.

Atlas Cloud per-second billing

Atlas Cloud bills video per second of output duration, at a flat published rate, with the live price shown next to the Run button in the Playground before you spend anything. The Seedance rates live on the platform right now:

For context on the same key: Kling V3.0 Turbo runs $0.112 standard or $0.095 discounted, Gemini Omni Flash text-to-video $0.125, MiniMax H3 $0.14, Wan 2.7 Spicy $0.10, and Grok Imagine Video v1.5 $0.08. Full current rates sit on atlascloud.ai/pricing.

To be completely clear, since the H1 of this page promises pricing: the Atlas Cloud rate for Seedance 2.5 has not been published. The numbers above are the live Seedance family rates, and they are the correct anchor for a budget you build today. A 30-second render at the current Seedance 2.0 rate is $3.36 of output; at the Mini rate it is $1.68. That is the order of magnitude a per-second video pipeline operates in, and it is the baseline against which any future 2.5 rate will be judged.

Which model is easier to plan around

Neither is dishonest. They optimize for different things. Token billing prices the compute you actually consumed; per-second billing prices the artifact you actually shipped. If your finance model is built around cost-per-finished-second of video, the second one requires no translation layer.

Quickstart: build the integration now, swap the model string later

Here is the part worth doing today. Video generation on Atlas Cloud is a two-step asynchronous REST flow on the same key that covers the rest of the catalog: you submit a job, then you poll it. There is no vendor SDK to install.

Step one, get a key. Sign up at console.atlascloud.ai. There is no deposit, no minimum commitment, and no waitlist. The key is issued immediately and billing is pay-as-you-go.

Step two, submit a render and poll for the result:

pythonCopy
1import os, time, requests
2
3BASE = "https://api.atlascloud.ai/api/v1/model"
4HEAD = {"Authorization": "Bearer " + os.environ["ATLAS_CLOUD_API_KEY"],
5        "Content-Type": "application/json"}
6
7job = requests.post(f"{BASE}/generateVideo", headers=HEAD, json={
8    "model": "bytedance/seedance-2.0-mini/text-to-video",
9    "prompt": "Slow push-in on a rain-streaked shopfront window at dusk.",
10    "duration": 5,
11    "resolution": "720p",
12}).json()
13
14pid = job["data"]["id"]
15
16while True:
17    r = requests.get(f"{BASE}/prediction/{pid}", headers=HEAD).json()
18    if r["data"]["status"] in ("completed", "succeeded", "failed"):
19        break
20    time.sleep(3)

Step three, pick a live Seedance model and keep its identifier in one place. Copy the exact identifier from the Seedance family page at atlascloud.ai/models/seedance2 rather than typing it from memory, then hold it in a single constant:

pythonCopy
1VIDEO_MODEL = os.environ.get("VIDEO_MODEL", "bytedance/seedance-2.0-mini/text-to-video")  # live today

Step four, run one generation from the Playground before you write the request in code. The Playground shows the live per-second price beside the Run button, so you can confirm both the output quality and the exact cost of your target duration and settings, then mirror those settings in your request. Optional parameters beyond the four above, including ratio, bitrate_mode, generate_audio and watermark, are documented per model at atlascloud.ai/docs; take them from there rather than guessing, because video models vary more in their accepted fields than text models do.

That is the whole preparation exercise. When Seedance 2.5 ships on Atlas Cloud, the thing that changes is the value of VIDEO_MODEL. Your authentication, your polling loop, your retry logic, your logging, your billing account, and your cost-tracking code all stay exactly as they are. That is the concrete meaning of a Day-0 commitment on unified endpoints: the new model arrives as a string, not as a migration.

Two practical notes while you wait. First, benchmark on Seedance 2.0 Mini at $0.056/s rather than on the flagship tier at $0.112/s, because iteration cost dominates during prompt development and the cheaper tier tells you most of what you need about your prompt structure. Second, keep reference assets and prompts in a versioned store, since Seedance 2.5's stated strength is reading the intention and framing of a reference video, and a clean reference library is the asset that transfers directly to it.

FAQ

Q: Can I call the Seedance 2.5 API on Atlas Cloud right now? A: No. 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, but at the time of writing the listing is marked Coming Soon and no Atlas Cloud model ID for it exists. Seedance 2.0, 2.0 Mini, and 1.5 Pro are live on that same key today.

Q: How much does Seedance 2.5 cost on Atlas Cloud? A: No Atlas Cloud price for Seedance 2.5 has been published, and there is no Seedance 2.5 row on the pricing page. The live Seedance rates are 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 or $0.018/s on its Fast variant, all billed on output duration. Those are the numbers to budget with today.

Q: Why can't I get a single dollar figure for a Seedance 2.5 clip from the first-party channels? A: Because they bill by token rather than by second. Consumption is estimated from input plus output video duration multiplied by output width, height, and frame rate, divided by 1024, with a minimum token floor when the input includes video. ByteDance publishes a pricing calculator for the series, and the authoritative figure is the usage.completion_tokens value returned by the API after the call.

Q: How much work is it to switch my OpenAI SDK app to Atlas Cloud? A: For text models, change the base URL to the Atlas Cloud v1 endpoint and swap in your Atlas Cloud API key, and the SDK and request patterns stay the same. Video is a separate REST flow rather than a chat completion, so a video feature is a small piece of new code rather than a port, which is also why adding a new model later is a one-line change.

Q: What happens to my integration when Seedance 2.5 launches? A: If you built against a live Seedance model on Atlas Cloud, you change the model identifier. Authentication, endpoint, billing account, and monitoring are unchanged, because 2.5 is committed to the same unified endpoints.

The bottom line

Seedance 2.5 is live through first-party ByteDance channels and is a Coming Soon listing with a Day-0 commitment on Atlas Cloud, where no price or model ID has been published yet. What is live on Atlas Cloud today is the Seedance family at $0.112/s for 2.0, $0.09/s for 2.0 Fast, $0.056/s for 2.0 Mini, and $0.047/s for 1.5 Pro, billed per second of output on one API key that also carries 300+ other text, image, and video models under a single bill, on a platform that is SOC II certified and HIPAA compliant.