Genmo Mochi

Genmo Mochi is an open-source AI video model that the AI community noticed quickly because its motion quality was meaningfully better than earlier open-source options. It uses an asymmetric diffusion transformer architecture that produces smooth, temporally consistent movement — human motion, fluid dynamics, and basic physics interactions hold up reasonably well. The model is available under an Apache 2.0 license, so it can be deployed locally, fine-tuned, and incorporated into production pipelines. A web interface on Genmo's site makes it accessible without local setup, though running it locally requires 24GB+ of VRAM. Output resolution tops out at 848x480, and clip length is short. It has become a popular base for community fine-tunes. The documentation is still sparse in places, and the project remains in an early stage.