W-Okada Voice Changer

Live voice conversion for streamers and gamers is what w-okada's Voice Changer delivers: a realtime client that transforms a microphone signal through AI voice models with latency low enough for conversation. It supports RVC, Beatrice v1 and v2, MMVC, so-vits-svc, and DDSP-SVC models, and can run standalone or in a server-client split so a second machine handles inference while a gaming PC stays unburdened; a REST API supports building custom clients in other languages. Prebuilt packages for Windows and Apple Silicon Macs are distributed through Hugging Face, so setup is download-and-run, while Linux users build from the repository; the v2 series, currently at v2.2.2 beta, adds a lightweight dedicated client for Beatrice v2. NVIDIA, AMD, and Intel GPUs are all usable for acceleration, and CPU-friendly models like Beatrice keep it workable without a discrete GPU. The project is MIT licensed with roughly 20,700 GitHub stars, making it the standard front end for running RVC-family models live in streaming, gaming, and online meeting setups.