Seedance 2.5 Flash cuts 720p video generation price by 34%

ByteDance's Volcano Engine announced Seedance 2.5 Flash on November 12, 2026, a cheaper, faster tier of its Seedance 2.5 video model that the company says generates video about twice as fast.

The Chinese launch report from IT Home puts the price for 720p text-to-video and image-to-video at about 0.99 yuan a second, roughly 34% below Seedance 2.5. BytePlus, ByteDance's international cloud arm, lists the same model at $7.00 per million tokens for 480p and 720p without video input, against $10.70 for the flagship. Its worked example prices a five-second 720p clip at $0.76, or $0.15 a second, and a five-second 480p clip at $0.34. The flagship's five-second 720p clip is $1.156, or $0.231 a second.

Seedance 2.5 is ByteDance's joint audio-and-video model, detailed this summer, built around 30-second single generations and multimodal references. Flash keeps the 30-second ceiling and the same two lower resolutions. It takes up to 10 images, five video clips and five audio clips per request, and it supports video editing. The flagship accepts up to 50 reference assets in one run, BytePlus's product page says, 30 images, 10 videos and 10 audio files, and it carries the 1080p, 4K and Draft-mode features that Flash's pages do not list.

That splits the Seedance 2.5 family into two tiers. The flagship stays the choice for the highest audiovisual quality and for 1080p and 4K production, with Draft mode letting a user pick a preview clip before committing to a high-resolution render. Flash is pitched at high-frequency work: testing ad creative, producing regular footage, and iterating fast.

For scale, Google's Vertex AI price list, read on November 12, 2026, quotes Veo 3.1 at $0.40 per generation and Veo 3.1 Fast at $0.10 per 720p generation. Google bills those per generation rather than per second, so they do not map cleanly onto Flash's per-second rate.

ByteDance's speed claim and its claim that Flash comes close to the flagship both rest on the company's own announcement. The BytePlus documentation gives no generation time, and tpsreport.news reports that ByteDance published no benchmark or side-by-side comparison with rival models. Speed also depends on prompt, duration, references and queue time, so the claim is worth timing once the API is open.

Seedance 2.5 Flash is rolling out now on Jimeng (Dreamina), LibTV, Updream, Creao AI and TapNow; the API opens on BytePlus ModelArk and Volcano Ark on November 15, 2026.