Voice Changer (w-okada)
An AI voice changer that converts speech live on Windows, Apple Silicon Macs and Linux, with Beatrice and RVC support and optional remote processing.
Voice Changer (w-okada), also called VCClient, converts your voice as you speak using AI voice models. It's for people who want live voice conversion on their own computer, including those recording gaming commentary while running demanding software. Processing can stay local.
You can also run voice conversion on a separate computer over a network. This moves the processing load away from the machine running your game or other applications. The app supports Windows, macOS on Apple Silicon M1, and Linux, including x86-64 and ARM systems.
Model support depends on the build you choose. Standard builds support Beatrice, while CUDA builds for Windows and ONNX builds for Windows and M1 Macs also support RVC (Retrieval-based-Voice-Conversion). Beatrice controls include speaker merging, automatic pitch shifting, and pitch and formant adjustments. You can export edited Beatrice models.
For developers building voice conversion into another application, a REST API provides access from custom clients. Google Colab is another supported way to run the processing, but that uses cloud hardware rather than keeping conversion on your own machine; the documented Colab setup requires an ngrok account.
Voice models and included character voices have their own usage terms. Some restrict what converted speech can say or how it can be distributed, and Amitaro requires attribution that makes clear the speech is a converted voice.
