Seedance is a family of AI video generation models made by ByteDance Seed, ByteDance's group for frontier model research. A video generation model is a neural network that produces a video clip from a text description (text-to-video) or from a starting picture (image-to-video). Seedance 1.0 makes 1080p videos with several connected shots and keeps the main subject and the visual style the same across the shot changes.
Seedance 2.5 is an audio-video joint generation model: it produces the sound together with the picture. It accepts text prompts and image, video or audio references, up to 50 reference files in one prompt, which helps keep the same character across clips. It generates videos of 4 to 30 seconds in 480p, 720p or 1080p at 24 frames per second, as MP4 or MOV files, and it can extend a video or edit one part of it.
Businesses call the model through ModelArk on BytePlus, a cloud platform for AI models, and pay per token used, or buy prepaid resource plans valid for three months; when a plan is used up, each further use is billed separately by actual usage.