daVinci-MagiHuman

DaVinci MagiHuman is an open-source AI model that creates talking-head videos from a single portrait photo. You provide a clear image of a face and either a text script or an audio file. The model then generates a video where the person in the photo appears to speak, with lip movements that are synchronized to the provided audio. It does this in one unified process, producing both the video frames and the matching audio together, instead of using separate systems for speech and animation.

The tool is a 15-billion parameter model developed by Sand.ai and the GAIR Lab at Shanghai Jiao Tong University. Its main technical distinction is its single-stream Transformer architecture, which denoises video and audio tokens simultaneously. This approach aims for better lip-sync quality and efficiency. The model is released under the Apache 2.0 license, meaning the weights are free to download, inspect, and use for commercial purposes. You can run it via a hosted web demo, or self-host it by downloading the checkpoints from Hugging Face or cloning the GitHub repository.

DaVinci MagiHuman is useful for developers and researchers experimenting with generative video AI, as well as creators who need to produce simple, lip-synced avatar videos without complex editing. It's a practical option for generating explainer videos, personalized messages, or prototype content where having an open-source, self-hostable model is a priority over cinematic quality. For best results, use a clear, front-facing portrait with good lighting.