Pixelle-Video

daVinci-MagiHuman is an open-source audio-video generative foundation model for creating human-centric videos with synchronized speech, facial expression, and body motion. It uses a single-stream transformer architecture to deliver multilingual generation and fast inference from reference images and text.