XTTSv2 for text-generation-webui

This is a simple extension for text-generation-webui that enables multilingual TTS, with voice cloning using XTTSv2 from coqui-ai/TTS.

Update

XTTSv2 is now built into text-generation-webui (coqui_tts). I might still update this extension for the Narrator feature, but if you don't care about that, use the official extension.

Disclaimer

This is a very crude extension i threw together quickly based on the barktts extension. It may or may not work. Feel free to improve the code and submit a PR.

Installation

Activate your environment by running cmd_windows.bat/cmd_linux.sh/cmd_macos.sh/cmd_wsl.bat depending on your platform, or activate your conda environment if you installed it manually. Clone this repo:

cd extensions
git clone https://github.com/kanttouchthis/text_generation_webui_xtts

Install dependencies for TTS.

cd text_generation_webui_xtts
pip install -r requirements.txt

Install TTS. Their version requirements cause issues so we install the dependencies above, without version requirements.

pip install TTS --no-dependencies

Usage

Once you finished the steps above, you can add some voices to the voices folder. This can be any short (3-6 seconds) wav clip of someone talking. Make sure it's high quality audio with no long gaps. Then, run the webui with --extensions text_generation_webui_xtts and select your voice/language and other settings at the bottom. You might have to accept the terms and conditions via the console when you first run it.