This project is a simplified implementation of the evaluation part of the Retrieval-based Voice Conversion (RVC) system. It aims to make it easy to use in any Python code by removing unnecessary components. As a sample, a command-line interface (CLI) is provided for real-time voice conversion.
- Simplified RVC evaluation
- Easy integration into any Python code
- Real-time voice conversion via CLI
RVC Eval Simplified (rvc-eval) is an appropriate name for this project as it highlights the simplified evaluation aspect of the original RVC system.
- Clone this repository:
git clone https://github.com/esnya/rvc-eval.git- Initialize and update the RVC submodule in the
rvcdirectory:
cd rvc-eval
git submodule update --init --recursive- Install dependencies using Pipenv:
pipenv install-
Download the Hubert model (
hubert_base.pt) from Hugging Face and place it in themodelsdirectory: -
Activate the Pipenv environment:
pipenv shell- To list available audio devices:
python -m rvc_eval.cli --list-audio-devices- To run the voice conversion system (with the default
hubert_base.ptmodel or specify a custom path with the--hubertoption):
python -m rvc_eval.cli --model path/to/your/model.pth --input-device-index 0 --output-device-index 1
- Python: >=3.8.x (3.10 recommended)
- PyTorch: 2.0.0+cu118 recommended
Note: Pipenv is used for managing dependencies.
- This project is based on the Retrieval-based Voice Conversion system by liujing04.
- This project also refers to Voice Changer by w-okada, and some parts of the code are based on it.
This project is licensed under the MIT License, following the licenses of the original RVC repository and the Voice Changer repository. See the LICENSE file for details.