Seedance API Pricing: Key, BytePlus Access and Docs

If you are about to wire Seedance 2.0 into a product, the first question is rarely about quality. It is about the invoice. ByteDance's video model does not bill by the second the way most people assume, the price shifts with resolution and reference footage, and Seedance 2.0 alone ranges from about 11 cents to 30 cents per second depending on where you call it. This breakdown lays out the real numbers, shows how to get BytePlus ModelArk access without guesswork, points to the docs pages worth bookmarking, and gives you a no-code route for projects where an API integration would be overkill.

💡 Price check window: the figures below come from public pricing references published between August and September 2026. Rates move often, so confirm on the BytePlus pricing page before you lock a budget.

What Seedance Costs Per Second

The short version: the official BytePlus rate for Seedance 2.0 at 720p works out to roughly 15 cents per second of output when you send text or images, and about 23 cents per second for the newer Seedance 2.5. Everything else in this article is a variation on those two numbers.

Tokens, Not Seconds

BytePlus ModelArk bills video by token consumption, not by clock time. The formula in the pricing docs reads:

tokens = (input video seconds + output video seconds) × width × height × frame rate ÷ 1024

Copy Prompt

When a request carries only text or images, the input video term is zero and only the output length counts. At 24 frames per second, a 720p clip (1280 × 720) consumes about 21,600 tokens per second, and a 1080p clip (1920 × 1080) consumes about 48,600 tokens per second. Multiply by the per-million-token price and you have the cost of the clip.

Two details change the bill. First, when a request includes a reference video, a minimum token floor applies, and an estimate that lands below it is billed at the floor instead. Second, the reference video's own duration is counted, so a lower per-token rate for video input can still produce a bigger invoice than a plain text prompt.

💡 The final charge is reconciled from the usage.completion_tokens field in the API response. Log that field on every call and a monthly statement will never surprise you.

Official Price Table

Independent reviewers report that BytePlus lists Dreamina Seedance 2.0 rates per million tokens, split by resolution and by whether the request includes video input. The numbers below follow that reporting. The ModelArk pricing doc is the authority, so check it before you commit money.

Seedance 2.5, which produces single generations of up to 30 seconds, launched on ModelArk at $10.70 per million tokens without video input. BytePlus keeps the full 2.5 matrix behind a pricing calculator, and the minimum floors depend on resolution, aspect ratio, and output duration. The older Seedance 1 Pro is listed at $2.50 per million tokens, well under every 2.0 rate.

Cost of Common Clips

Here is what the published formula produces for clips with no reference video. These are estimates at 24 frames per second, and your invoice follows the token count in the API response.

Add a reference video and the math shifts. A 5 second clip with a 5 second reference at 720p on Seedance 2.0 counts 10 seconds, or 216,000 tokens, at the $4.30 rate: about $0.93, which is more than the $0.76 text-only clip. A full 30 second Seedance 2.5 take at 720p lands near $6.93.

Official Rates Versus Reseller Prices

BytePlus is the source price, but many teams meet Seedance first through an aggregator. Those platforms add convenience, simpler billing, and sometimes a discount. They can also add a large markup, and the spread is wider than most people expect.

The Reseller Spread

Published per-second prices for Seedance 2.0 and its smaller siblings, as listed by each provider:

The resolution and input type behind each number differ, so read this table as a spread, not a ranking. Still, the pattern is clear: the Seedance 2.0 family ranges from roughly 6 cents to 30 cents per second.

Seedance 2.5 Snapshot

The newer model shows the same pattern. These 720p text-to-video prices come from a price tracker snapshot dated August 22, 2026.

The cheapest option undercuts the official rate by about 13 percent, and the most expensive charges roughly double. Check what each tier actually renders before you chase the lowest number, because a discount that costs you motion quality is not a discount.

Retries Are the Hidden Bill

Published per-second prices describe accepted clips. Real projects also pay for rejected takes: a hand that morphs, a lip sync that drifts, a camera move that ignores the prompt. One September 2026 price tracker (anikuku.com) recommends budgeting on accepted clips only. If one take in three survives review, your true cost per usable clip is triple the table figure.

These levers move the total most:

• Resolution: 1080p uses 2.25 times the tokens of 720p for the same duration.
• Duration: cost scales in a straight line, so a 10 second take costs twice a 5 second take.
• Reference video: its seconds are counted on top of the output seconds.
• Prepaid packs: they lower the unit rate but carry an expiry window, and one price tracker cites 90 days.
• Region and account terms: both can change the effective rate.

Get BytePlus ModelArk Access

BytePlus ModelArk is ByteDance's international model platform, with US dollar billing and an API that mimics the OpenAI style. For developers outside China, it is the official first-party route to Seedance.

Create the Account

1. Register a BytePlus account and sign in to the ModelArk console.
2. Add a payment method. Pay-as-you-go works, and prepaid resource packs are the alternative if you want a lower unit rate.
3. Find the model in the console's model list, either Dreamina Seedance 2.0 or 2.5, and activate it if the console asks.
4. Pick your region before you create credentials, because the next step ties them to it.

Generate an API Credential

The credential lives under the permission management section of the ModelArk console. The official workflow page walks through the clicks. Good habits matter more than the clicks:

• Copy the credential once, at creation, and store it in an environment variable.
• Never commit it to Git or paste it into a client-side script.
• Use one credential per project so you can revoke a single one without breaking everything else.
• Send it as a bearer token in the Authorization header.

Mind the Region

Credentials are isolated by region. The documented international endpoint for Seedance video is the Asia-Pacific base URL:

https://ark.ap-southeast.bytepluses.com/api/v3

Copy Prompt

A credential created in one region will not authenticate against another, and at the time of writing, SDK documentation notes that Seedance video runs on the Asia-Pacific endpoint only.

Where the Docs Live

The Pages Worth Bookmarking

BytePlus keeps everything under docs.byteplus.com. These are the five pages you will reopen most:

• ModelArk quick-start: account, console, and a first request.
• Video generation API reference: request fields, task states, and response shape.
• Dreamina Seedance 2.5 tutorial: model behavior and 2.5 specifics.
• ModelArk pricing: rates per million tokens, tiers, and packs.
• Credential workflow: creating and storing credentials safely.

A First API Call

Seedance is an asynchronous task API. You create a task, receive an ID, poll until the status settles, then download the finished MP4.

curl https://ark.ap-southeast.bytepluses.com/api/v3/contents/generations/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $ARK_TOKEN" \
  -d '{
    "model": "dreamina-seedance-2-0-260128",
    "content": [
      {
        "type": "text",
        "text": "A lone fishing boat crossing a calm sea at sunrise, slow dolly-in, natural light"
      }
    ]
  }'

Copy Prompt

Resolution, duration, aspect ratio, and audio are request fields described in the video generation API reference above. When the task succeeds, the response carries the video URL and the usage.completion_tokens figure you are billed on.

💡 The dated suffix in a model ID changes between releases. Copy the exact ID from your console's model list instead of trusting a blog post, this one included.

Pick the Right Seedance Version

Which Version Fits Which Job

The workflow that keeps invoices small is simple. Draft on Mini or Fast, finish on standard 2.0, and reserve 2.5 for shots that genuinely need the extra length.

How to Use Seedance 2.0 on PicassoIA

If an API integration is more than your project needs, Seedance 2.0 on PicassoIA runs in the browser with no BytePlus account, no token math, and no code. It generates dialogue, sound effects, and music in the same pass as the picture.

Write the Prompt

1. Open the Seedance 2.0 model page.
2. Describe the shot in cinematic, chronological order: subject, motion, camera move, lighting.
3. Put spoken dialogue in double quotes so the model voices it.
4. Optionally attach up to 9 reference images, 3 reference videos, and 3 audio clips, and call them in the prompt as [Image1], [Video1], or [Audio1]. Audio needs at least one image or video alongside it.

💡 A first frame and a last frame cannot be combined with reference images in the same run. Pick one approach per generation.

Set Duration, Resolution, Audio

Free Drafts With Seedance 2.5 Lite

Seedance 2.5 Lite renders at 480p or 720p in 5 or 10 second clips, from text or from a starting image, with synchronized audio. Wonder members can run it without limits, which makes it the right place to burn through ten prompt variations. Prompts carry over between models, but a seed only repeats a take on the model that produced it, so move the winning wording to Seedance 2.0 and re-render the final take there.

Budget Habits That Save Money

• Write the shot list first. Ten planned shots cost less than thirty trial ones.
• Log usage.completion_tokens. Costs per accepted clip become visible within a day.
• Cap retries in code. Three attempts per shot is a sensible ceiling.
• Skip reference video when a still image gives the same direction, since its seconds are billed.
• Fix instead of re-roll. Trimming, color matching, and effects often rescue a good take. The video editing and effects models in the PicassoIA model library can handle those repairs without paying for a fresh generation.
• Buy packs in small steps until you know your monthly volume.

Three mistakes show up again and again:

1. Comparing per-second prices across different resolutions. A 720p price and a 1080p price are not the same product.
2. Forgetting that reference seconds count. The cheaper token rate misleads here.
3. Hard-coding a dated model ID. Releases change the suffix, and the call fails.

Try Your Own Seedance Clip

The fastest way to find out what a usable clip costs is to make one. Open Seedance 2.0 or the unlimited Seedance 2.5 Lite on Picasso IA, paste a single sentence describing your shot, and render a five second take. Then change one variable, whether the camera move, the light, or the seed, and render again. After a handful of runs you will know exactly how many takes a keeper needs, and that number is the only pricing figure that matters for your budget. Browse the rest of the lineup at picassoia.com/en/all-models and start experimenting.