ByteDance Opens Seedance 2.5 Flash to API Users
The cheaper Flash variant of ByteDance's video model is now available through the BytePlus ModelArk API, with some capabilities removed.
Updated on Nov. 15, 2026 in Artificial Intelligence
Live Poll
Do you trust AI-generated video to be reliable enough for everyday use?
ByteDance has opened its Seedance 2.5 Flash video model to API users on BytePlus ModelArk. Flash costs $7.0 per million tokens at 480p and 720p, compared with $10.70 per million tokens for the full Seedance 2.5 model.
Why it matters
The lower price gives developers a cheaper option for generating short AI video clips, with a trade-off in output quality. Flash drops 1080p, 4K and offline inference support, which limits its use for higher-resolution production work.
Flash accepts up to 20 reference files, including 10 images, 5 videos and 5 audio clips, compared with 50 for the full model, and both produce clips up to 30 seconds long. Flash does not support 1080p, 4K or offline inference according to the ModelArk documentation, and the source does not report how its quality compares with the full model.
The players
ByteDance
ByteDance is the Chinese technology company behind TikTok and the Doubao chatbot, and it develops the Seedance family of AI video models.
BytePlus
BytePlus is ByteDance's international cloud and enterprise technology arm, which operates the ModelArk platform used to deliver its AI models to developers.
The details
Flash keeps the 30-second clip length and built-in audio of the full Seedance 2.5 model while dropping 1080p support. The full model launched July 31 with native 1080p output, and consumer access to Flash began November 12 through Dreamina, LibTV and Updream.
Timeline
July 31, 2026: The full Seedance 2.5 model launches with native 1080p output.
November 12, 2026: ByteDance announces Flash and consumer access begins through Dreamina, LibTV and Updream.
November 15, 2026: The Flash API opens on BytePlus ModelArk.
The Tech Race
Flash extends the Seedance 2.5 line launched on July 31 with native 1080p output by offering a cheaper variant that drops that resolution. The move follows the pattern of AI providers splitting model tiers by price, with Volcano Engine's 720p text-to-video pricing of about 0.99 yuan per second sitting roughly 34% below Seedance 2.5.
Developers building video features can now access a lower-cost option at roughly $0.15 per second of 720p output, though 1080p and 4K output require the full model. Anyone generating clips through the consumer apps Dreamina, LibTV or Updream gains access to Flash starting November 12.
The takeaway
Lower-priced AI video tiers make short-form generation cheaper, but resolution limits mean buyers should match the variant to the output they need. Compare the specification list before choosing between Flash and the full model.
Further reading
For more on the models behind this release, see our Artificial Intelligence coverage.
Source note: This article includes information reported by Startup Fortune.
Live Poll
Do you trust AI-generated video to be reliable enough for everyday use?
