/voice-generator-webui

A multi-speaker, multilingual speech generation tool

Primary LanguageJupyter NotebookMIT LicenseMIT

Voice Generator web UI

A Multi-speaker, multilingual speech generation tool.

日本語はこちら

Open In Colab

Features

  • Speech synthesis for 209 speakers (109 English / 100 Japanese)
  • Script generation using LLM
  • Accent and phoneme editing functions
  • Voice conversion by RVC
  • Batch voice conversion by RVC

Installation and Running

  • Python (tested on 3.10.6)
  • PyTorch for CUDA
    Check if the CUDA version of PyTorch is already installed with pip show torch.
    It is OK if cu is included in the Version as follows.
Version: 2.0.0+cu118

If not, install the CUDA version of PyTorch.

pip uninstall torch
pip install torch --index-url https://download.pytorch.org/whl/cu118

Windows

  1. download the zip or git clone https://github.com/log1stics/voice-generator-webui

  2. run setup.bat or setup.bat en
    setup.bat en avoids installation of pyopenjtalk used for Japanese generation

Run the webui.py file in PowerShell or at the command prompt

python webui.py

Linux

Run the webui.py file at a PowerShell or command prompt

apt-get install espeak # For English Text To Speak

git clone https://github.com/log1stics/voice-generator-webui
chmod +x setup.sh
setup.sh

# if you do not use Japanese Text To Speak
# you can avoid install pyopenjtalk
setup.sh en

Add RVC model

For example, if you want to handle an RVC trained model named EXAMPLE_MODEL.pth, place it as follows

vc/
└── models/
    └── EXAMPLE_MODEL/
        ├── EXAMPLE_MODEL.pth # file name should be the same as the directory name
        ├── added.index # not necessary (name is fixed to added.index)
        └── total_fea.npy # not necessary (name is fixed to added.index)

Contributing

Here's how to add code to this repo: Contributing

Credits

Dataset