MAGI-1 is an open-source video generation model built on an autoregressive diffusion approach. You guide it with text prompts, and it produces video in a streaming fashion, which allows clips to be extended to longer lengths than many fixed-length generators.
The site lets you upload a JPG, PNG or WEBP image as a starting point and try the model in the browser. It is a good fit for creators, hobbyists and developers who want to explore text-driven video generation with an openly available model.
