Duix-Avatar

Why we included this project

Duix.Avatar turns a person's face and voice into a talking digital human on a Windows machine, cloning appearance and speech from reference footage and then driving the result with text or audio so lips and expressions stay in sync. The whole pipeline runs offline, which is the main reason it stands out. There is no cloud round-trip, so footage never leaves your machine, and finished video comes out the other end on a computer with an NVIDIA GPU. That makes it a useful reference for privacy-conscious creators and for teams that cannot send recordings of people to a remote service: the project shows how speech recognition, voice cloning, and face/vision components fit together into a working product. It ships as an Electron application with Docker-based deployment, and the interface is built for non-technical users rather than researchers, so it feels like a production tool instead of a paper demo. One thing to check before adopting it: the license is Duix.com's own community agreement, not a permissive open-source license, and it includes attribution and branding obligations. Teams that plan to redistribute the code or build products on top of it should read those terms first.