svc-toolkit 1.1.0

A self-contained singing voice conversion application using the so-vits-svc architecture, with Deep U-Net model for vocal separation feature and easy to use GUI.

Singing Voice Conversion Toolkit

A self-contained singing voice conversion application using the so-vits-svc architecture, with Deep U-Net model for vocal separation feature and easy to use GUI.

Getting Started

Installation

1. Install Python (3.10 is recommended, but 3.10 - 3.11 should work)

2. Install pipx

3. Install the package by running this following terminal command if you only have one Python version installed:

pipx install svc-toolkit

To install with a specific Python version, use the --python flag. For example, to install with Python 3.10:

pipx install svc-toolkit --python 3.10

Using NVIDIA GPU

To use the package with NVIDIA GPU, you need to upgrade the following dependencies:

pipx inject svc-toolkit torch==2.1.1 torchaudio==2.1.1 --pip-args="-U" --index-url https://download.pytorch.org/whl/cu121

For CUDA version 11.*, you can change the cu121 to cu118. So the command will be:

pipx inject svc-toolkit torch==2.1.1 torchaudio==2.1.1 --pip-args="-U" --index-url https://download.pytorch.org/whl/cu118

Note that AMD GPUs are not actively supported, but you can try using the package with the CPU version of PyTorch.

For other installation options, see Installation.

Usage

Windows

svct.exe

macOS/Linux

svct

For the detailed usage guide, see Usage.

Development

For the detailed development guide, see Development.

About

This project is the implementation of the final year project for the Bachelor of Science in Computer Science, Department of Computer Science, City University of Hong Kong, named "Singing Voice Conversion from Fully Mixed Track with GUI", with project code 23CS062.