DaVinci MagiHuman

DaVinci MagiHuman is an advanced open-source AI model with 15 billion parameters, developed by Sand.ai and the GAIR Lab at Shanghai Jiao Tong University. It turns a single portrait image plus a script or audio file into a high-quality, lip-synced talking video, so you no longer need a studio, actors, or complex editing workflows to bring a face to life.

Unlike traditional approaches that stitch together separate text-to-speech and video pipelines, DaVinci MagiHuman uses a unified single-stream Transformer that jointly denoises video and audio tokens at the same time. This keeps lip movements and speech naturally in sync and avoids the mismatches that often come from combining independent tools.

Released under the Apache 2.0 license, the model gives you full freedom to inspect its weights, run inference locally on your own hardware, and use the technology for commercial purposes. Whether you are prototyping avatars or shipping a product, you stay in control of your data and your costs.

Speed is another highlight. DaVinci MagiHuman is optimized for fast generation and can produce short clips in just seconds on professional-grade hardware such as the NVIDIA H100, making rapid iteration and large-scale content creation far more practical.