Faster video previews with Seedance 2.5 Flash

The Flash variant generates up to 30 seconds at 480p or 720p. BytePlus plans to launch its API on November 15 at $0.152 per second for 720p.

Fifteen seconds of 720p video costs $2.28 at the announced rate for Dreamina Seedance 2.5 Flash. BytePlus is introducing this accelerated variant of ByteDance’s video model at $0.152 per second for text-to-video and image-to-video generation, approximately 34% less than Seedance 2.5. The API is scheduled to launch on ModelArk on Sunday, November 15, 2026.

At that same rate, a thirty-second video costs $4.56. This amount applies to the announced text and image generation modes at 720p; it does not establish the cost of a request using a video reference. It also concerns BytePlus’s API offering, which is separate from Dreamina subscriptions and credits. Visual and audio references to guide the shot

The limits communicated by the provider allow up to ten images, five videos, and five audio clips per request. Generated videos can reach thirty seconds at 480p or 720p. References help define the intended result beyond a written description alone.

In the Flash overview on Dreamina, the workflow distinguishes between a multiple-reference mode called Omni Reference and guidance using first and last frames. Aspect ratio, resolution, and duration are set before generation begins. The page also describes output with synchronized audio.

These limits differ from those of the standard Seedance 2.5 model, whose guide lists support for up to fifty references per request. Reported acceleration without an absolute generation time

The provider claims generation is nearly twice as fast as the standard model, with results close to its flagship version. The announcement provides neither an absolute generation time per video nor a protocol detailing the comparison conditions. It therefore cannot be translated into a guaranteed turnaround time for a fifteen- or thirty-second shot.

The claim of comparable quality also remains the provider’s assessment, without detailed comparative scores in the launch materials. Dreamina’s FAQ presents the two versions as complementary: Flash prioritizes frequent experimentation, while the standard model targets higher-quality final renders, with 1080p output. Flash’s announced ceiling remains 720p.

The positioning focuses primarily on motion tests, framing exploration, and variations on advertising concepts. A rotating product shot or animated poster can help compare several directions before selecting those worth developing further. Moving from a preview to the standard model

The documented workflow involves reviewing the Flash clip, then reusing the prompt and references in Seedance 2.5 for a higher-quality version. This step triggers a new generation using the same starting material.

Within the interface, the accelerated variant must be selected manually from the model list. The page describes its use in the video generator, in a canvas organized into nodes, and in Agent workflows for producing several options. The guide recommends concise instructions centered on an identifiable action, with the subject, its movement, and camera direction clearly described.