DoaCam

DoaCam is an open-architecture SDK for building real-time 3D AI avatar experiences in the browser. It connects Google Gemini's native audio model to a WebGL avatar with lip-synced speech, computer vision, and persistent client-side memory. Built with vanilla JS + Vite, FastAPI backend, and Google ADK. Features include AudioWorklet-based audio pipeline (16kHz capture / 24kHz playback), MediaPipe face tracking for expression mirroring, a 97-motion animation engine, and privacy-first IndexedDB memory. No sign-up required, works on legacy hardware.