w-okada Voice Changer, released as VCClient, is open-source software that changes your voice in real time with AI. Speak into a microphone and the converted audio comes out with a different voice character, using models such as RVC and Beatrice v2. It is free to download from GitHub and Hugging Face, with editions for Windows, Apple Silicon Macs and Linux, plus notebooks for Google Colab. The project has more than 21,000 stars on GitHub, and a REST API lets developers drive it from their own programs or from curl.
The main audience is streamers, gamers, VTubers and content creators who want to sound like a different character during live sessions or recordings. Hobbyist developers and machine learning enthusiasts who like to try voice models, merge speakers or edit a Beatrice model will also feel at home. Some comfort with downloading builds, choosing the correct edition for a GPU and troubleshooting audio devices helps, because the interface and most documentation are in Japanese, with other languages machine-translated. Users of Mac builds should expect a manual step to run unsigned software.
Typical uses include talking as an anime character in a voice chat, hiding your natural voice on a stream, or recording narration with a converted voice. Because VCClient supports both standalone and network setups, voice conversion can run on a second computer while a demanding game uses the main one. Developers can wire the REST API into custom clients, Docker setups or notebooks on Colab and Kaggle. Newer releases add shortcut keys, buffer visualization, Beatrice speaker merging and automatic pitch shifting, and the sibling projects TTSClient and ASRClient cover reading text aloud and live speech recognition.
