Skip to main content
Post

Seedance 2.5 Flash now available on AI Gateway

Vercel added ByteDance's Seedance 2.5 Flash to AI Gateway. It's a faster, lower-cost version of Seedance 2.5. Vercel says its output is close to the standard model in human realism, reference-based editing, and prompt adherence.

You can generate video with synchronized audio from a text prompt, or from image, video, and audio references. Clips run 4 to 30 seconds at 480p or 720p and 24 fps, in landscape, square, or portrait. You can also edit or extend an existing video, or supply images to set the first and last frames.

In the AI SDK, call generateVideo with the model bytedance/seedance-2.5-flash. To guide a result with references, pass them in inputReferences and point to them in the prompt as [Image 1], [Video 1], and so on. One example prompt is "The lion from [Image 1] walks through a sunlit meadow, filmed like [Video 1]." The model page lists up to 10 images (30 MB each), 5 videos (200 MB each, 2 to 30 seconds), and up to 20 reference inputs total. Audio references can be mp3 or wav, up to 15 MB.

The model page lists output at $7 per 1M generated tokens. The release note's examples use AI SDK 7, and the model page says AI SDK 6 or later works. You can also try it in the model playground on Vercel.

Loading replies