W-Okada Voice Changer: Windows setup and voice models

This tutorial covers W-Okada Voice Changer setup for live voice conversion on Windows. The presenter describes it as free and contrasts it with an earlier VoiceMod workflow that used recorded or uploaded audio. He demonstrates character voices, but notes that the results need more adjustment.

The walkthrough selects the Windows CUDA download, described as almost 4 GB, and points out a separate Mac file. It then installs VB-Cable as administrator. The presenter shows how to restore the usual audio output if installation changes it and restarts after an installer prompt. Launching dist/main/start_http downloads models and starter voices on the first run, so the setup requires an internet connection at that stage.

Audio routing uses the microphone as input and Cable Input as output. A monitor lets users hear the converted voice. The tutorial covers noise gate, gain, pitch, noise suppression and recording controls. The presenter recommends selecting a dedicated GPU when available and describes chunk size as a tradeoff between response speed and voice quality.

To add a voice model, he extracts its files, opens Edit and assigns the PTH file to Model and the index file to Index. He then selects the imported voice and tests it with a recording.