RVC

RVC (Retrieval-based Voice Conversion) is the open-source voice-conversion framework behind the AI-cover wave: train a model of any voice on as little as ~10 minutes of audio, then convert any speech or singing into that voice. The project's classic Gradio WebUI (RVC-Project/Retrieval-based-Voice-Conversion-WebUI, 38.7k stars, actively maintained) covers model training, inference, and real-time voice changing; the core repo scored here is being rewritten as a pip-installable rvc library/API (pip install git+..., rvc init scaffolds assets/ and .env). Base models are trained on open VCTK data. MIT licensed. Note the obvious misuse surface: voice cloning enables impersonation — check consent and local deepfake law before publishing converted voices.

Library form (verified from README of RVC-Project/Retrieval-based-Voice-Conversion):

```bash

pip install git+https://github.com/RVC-Project/Retrieval-based-Voice-Conversion

mkdir my-voice && cd my-voice

rvc init # creates assets/ (models) and .env in an empty directory

```

Then follow the repo README/docs for the train and infer commands (the library is under active development — expect API churn).

Classic WebUI form (the 38.7k-star product most tutorials cover):

```bash

git clone https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI

# follow its README (Python 3.10, PyTorch, then launch the Gradio UI)

```

Pretrained base models are hosted on Hugging Face (lj1995/VoiceConversionWebUI).