Seedance 2.5 Flash API opens November 15 with a 34% lower 720p rate

BytePlus announced Dreamina Seedance 2.5 Flash on November 12, giving video creators a lower-priced option for 720p generation. CapCut, Higgsfield and Lovart announced access in their apps. BytePlus has scheduled online trials and direct API calls on ModelArk for November 15 at 14:00 UTC+8, or 06:00 UTC. Volcano Engine’s Ark platform in China lists the same start time.

BytePlus put the price of generating 720p video from text or an image at $0.152 per second. The flagship Seedance 2.5 costs $0.231 per second on the same basis, making Flash about 34% cheaper. That difference applies to each clip a creator makes at the same resolution while trying different versions of a shot. BytePlus also says Flash generates nearly twice as fast. Its published comparison gives no workload or queue conditions for that claim.

Two output resolutions and three kinds of reference media

Flash initially outputs 480p or 720p. ModelArk’s price table lists online 1080p generation separately for the flagship. BytePlus also shows a distinction in how drafts work. The flagship’s Draft mode can take a selected 480p draft into a final 1080p render. Flash’s 480p output is a regular generation and the model does not offer that draft-to-final feature.

A Flash request can combine reference images, video and audio. The limits are 10 images, five video clips and five audio clips, compared with 30 images, 10 video clips and 10 audio clips for the flagship. A production that submits a few product photos, an example of the intended motion and a sound reference can still supply all three kinds of media. The lower limits matter for jobs that bundle assets from several shots into a single request.

A video reference changes the bill

ModelArk’s examples put a five-second 720p clip in 16:9 at $0.76 on Flash and about $1.16 on the flagship when the request contains no input video. At 480p, the five-second estimates are $0.34 and about $0.51 respectively. Ten successful 720p clips at those settings would cost about $7.60 on Flash and $11.56 on the flagship, using BytePlus’s estimates for one clip. Higher output resolution uses more tokens at the same duration. For requests without input video, ModelArk lists the same Flash rate of $7 per million tokens at 480p and 720p. The output dimensions change the number of tokens a clip uses.

The calculation changes when the request includes a video. ModelArk counts input video duration as well as output duration. Its five-second 720p Flash example ranges from $0.84 to $3.25, with the low end corresponding to a two- to four-second video input and the high end to a 30-second input. Requests with video also have minimum usage thresholds that depend on resolution, aspect ratio and output length. ModelArk says it charges for successfully generated videos.

Each app sets its own access and price

Higgsfield announced early access on November 12 and said the United States was outside that rollout. Lovart announced availability the same day and said its annual Pro plan includes five Flash generations, while annual Ultimate includes 20.

CapCut advertised Flash in its app at $0.01 per second in a November 12 post. The offer is separate from the ModelArk API rate. CapCut said it based the price on its U.S. app and that the amount can vary with the market, plan, platform and generation settings. These app rollouts precede ModelArk’s direct API opening.

ModelArk’s video API creates a job and lets the calling app check its status to retrieve the finished clip. BytePlus lists a dedicated Flash model for those jobs and plans to open both its web trial and API access at 14:00 UTC+8 on November 15.