Seedance API pricing compared (updated monthly)

In shortAs of 2026-10-13, Seedance 2.0 Fast costs between about $0.12 and $0.24 per second of 720p video depending on where you buy it, and Seedance 2.5 between about $0.20 and $0.47. BytePlus ModelArk sells it direct in tokens; resellers such as fal, Replicate and Dawnwood Models sell it per second. The table below lists each provider's public 720p rate with how it bills, so you can compare like with like. Check each provider's own page before you commit; prices move.

How to read this table

Every price is USD per second of generated video at 720p, taken from each provider's public pricing page on 2026-10-13. Where a provider bills in tokens or credits we say so in the last column, and where we converted, the number is marked with ≈. A dash means the provider does not list that model or we could not find a public per-second rate. We sell Seedance ourselves (Dawnwood Models), so we are in the table; we have tried to describe every provider the way it describes itself.

Prices as of 2026-10-13. Every provider here changes prices, runs promotions and adds tiers; check each provider's own pricing page before you build a budget on these numbers. If a number here is out of date, email support@dawnwood.ai and we will correct it in the next monthly update.

Why the same model has five prices

Seedance is ByteDance's model and BytePlus ModelArk is the first-party way to buy it outside China. ModelArk bills in output tokens, and a second of 720p video is roughly 21,700 tokens, so its per-second price depends on resolution and length. Everyone else in the table is a reseller or an aggregator, and the price differences come from four places:

• Margin and billing model. Some resellers mark the list rate up, some pass it through and charge a fee on top-ups instead, some sell credits bundled into creator plans.
• Volume terms upstream. BytePlus offers volume discounts and higher concurrency to larger contracts. A reseller with those terms can sell below the public list rate; that is not by itself a sign of anything unusual.
• What a "second" includes. Replicate prices requests with a video input separately because the input video is also tokenized. Most per-second prices assume text and image inputs only.
• Promotions. Launch and limited-time discounts are common, especially on Mini.

What else to compare besides price

• Failed generations. Ask whether you pay for a failed or moderated request. On Dawnwood Models you do not: credits are reserved when a task is created and returned in full if it fails.
• Concurrency. Individual first-party accounts are limited in how many tasks run at once. If you render many shots in parallel, the concurrency a provider gives you matters as much as the rate.
• Request shape. If you already have a ModelArk integration, a ModelArk-compatible provider is a two-line change. See Migrating Seedance from fal or ModelArk.
• Invoicing and terms. For a business, who the seller is, what jurisdiction it sits in and what invoice you receive can matter more than a cent per second.
• Content policy. Requests go through the upstream model's moderation wherever you buy, and each provider has its own terms on top. Read them before you build on a provider, especially if your work involves real people's likeness.

Cost of a five-second clip

At 720p, a five-second clip of Seedance 2.0 Fast costs $0.75 on Dawnwood Models; multiply any per-second rate in the table by five to compare. Billing is on the duration you request, not the length you keep, so in shot-by-shot production the number that matters is per-second rate × seconds requested × attempts per keeper.

How we keep this page current

We re-check every provider's public pricing page once a month and update the table and the date at the top. Our own prices on this page come from the same source as our model pages, so they cannot drift from what the API charges.