Seedance 2.5 Flash API Pricing: What Gets Cheaper and What You Give Up

Seedance 2.5 Flash has a lower published token price than standard Seedance 2.5 on BytePlus ModelArk, but it stops at 720p. BytePlus now documents Flash as available for API calls and its online experience. That does not establish access for every account.

This is a pricing and integration-planning guide based on the published documentation, not a generation test. The useful decision is whether Flash fits the output you actually need, rather than whether its name implies a free upgrade to your existing video workflow.

BytePlus Flash prices versus standard Seedance 2.5

The official price table lists these online rates in USD per million tokens for 480p and 720p:

Input	Seedance 2.5 Flash	Standard Seedance 2.5
Without video	$7.00	$10.70
With video	$4.30	$6.40

Flash’s model ID is dreamina-seedance-2-5-flash-261115. Its pricing row marks 1080p, 4K and offline inference unsupported. Its five-second, 16:9 estimates without video input are $0.34 at 480p and $0.76 at 720p.

Actual usage comes from usage.completion_tokens; video-input requests can also have a minimum billable token count. Adding a reference clip is not a cost-saving shortcut.

The practical consequence is straightforward: decide the delivery resolution before selecting the model. If the brief requires native 1080p, Flash’s lower rate does not solve that requirement. Adding another processing step should be a separate quality and cost decision, not a hidden assumption in the estimate.

Confirm availability separately from pricing

Updated November 15, 11:30 UTC (5 p.m. IST): BytePlus’s API documentation now lists Flash as fully available, replacing the earlier scheduled-launch notice. It also specifies Flash’s reference-input limits. No Flash request was executed for this guide.

When you integrate, check the model selected in the request, the account’s activation status and the returned task status separately. A price row establishes what the provider lists; it does not prove that access is enabled for a particular account. Do not silently send the standard model when Flash is unavailable unless your application explicitly permits that fallback and its cost.

For omni reference-to-video, Flash accepts up to 10 reference images, five reference videos and five reference audio clips. Standard Seedance 2.5 allows 30, 10 and 10 respectively. That matters before a request reaches billing: a standard-model brief with 12 reference images cannot move unchanged to Flash.

Review inherited request templates against the Flash-specific limits rather than assuming that the cheaper model is a drop-in replacement. Reducing the references changes the creative instructions too, so make that an explicit editorial choice instead of silently dropping assets.

Do not mix BytePlus and Runway billing rules

A provider comparison needs the same model variant, resolution, input assets and failure policy. Runway’s standard Seedance 2.5 guide describes per-second credit billing, additional video-reference charges and a minimum charge. Those are Runway’s rules for the standard model, not BytePlus Flash prices.

Even task failures can differ. Runway says its standard-model input-moderation failures are charged, while BytePlus’s pricing page says unsuccessful video generations, including content-moderation failures, are not billed. Keep those policies attached to their provider rather than merging them into a generic “Seedance price”.

For an internal estimate, write down the provider and model ID beside every number. Otherwise, a spreadsheet can look precise while combining one provider’s output price with another provider’s input or retry assumptions.

Budget for a usable clip, not a successful request

Suppose a team needs one approved product clip. Before generating, define what makes a result usable: the product remains recognisable, motion is acceptable and the final format fits the placement. A technically successful output that fails that brief still consumes the creative budget.

Keep a simple trial record with requested settings, task outcome, billed usage and whether the reviewer accepted the clip. Set a stop condition before running variations, and compare Flash with standard on the same brief rather than comparing unrelated showcase videos. This is a proposed evaluation workflow, not evidence that either model needs a particular number of attempts.

The cheaper model is the one that meets the brief at a lower total cost. Flash’s published rates make it worth evaluating for lower-resolution work, but neither a price table nor a launch demo establishes its usable-output rate for your product.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help technical teams explain product launches and pricing trade-offs in content developers and founders can actually use.

Book a 30-minute call