Twitch Text-To-Speech Tool

This tool is added !tts command, and play audio from Python TTS.

It's Free (twitch-tts under MIT License).

The Goal :

1. Help that some people don't want to use their real voice.
2. Accessible feature shouldn't charge.
3. Making the datasets/models for public domain.

Use it carefully, it's an early version. Upstream, Coqui.ai is under unknown status.

Development On Ubuntu Linux.

A. Classical (support CUDA)

Twitch.py client - !tts [something] -> Coqui.ai TTS Feature - Curl -> ALSA:aplay

B. New (CPU only)

Browser (twitch-js) <-> Docker with Coqui.ai TTS

docker run -it --rm -p 5003:8000 -p 5002:5002 -e TWITCHCHANNEL=<ChannelName> dreamerwolf/twitch-tts-server:latest /root/twitch-tts/run-cpu-docker.sh

• Open http://localhost:5003
  • Allow auto play media
  • Allow JavaScript
• Wait the Model Download

0. Python Venv and Git

sudo apt install python3-venv git
mkdir -p ~/python3-venv-root/twitch-tts-venv
python3 -m venv ~/python3-venv-root/twitch-tts-venv
cd ~/python3-venv-root/twitch-tts-venv
source ~/python3-venv-root/twitch-tts-venv
git clone https://github.com/dreamerc/twitch-tts

1. Install https://github.com/coqui-ai/TTS and others

pip install -r requirements.txt

2. Start TTS WSGI server , and it will take minutes for the model based on your network.

CPU :

./run-cpu.sh

nVidia CUDA :

./run-cuda.sh

3. Edit config.txt and Start the twitch chatbot.

• Channel & nickname : your twitch account name in URL, but channel name start with #
• OAuth : https://twitchapps.com/tmi/
• client_id & client_secret : https://dev.twitch.tv/console/apps/create

Start :

python twitch-tts.py

Usage :

use it in chatroom

!tts <something> 

• Q: If you don't want to use browser as a client, you still can use python client.

  A: Use Python venv feature to install.

• Q: Where is autoplay and JavaScript setting?

  A: https://blog.mozilla.org/en/uncategorized/block-autoplay/ (But set allow for Firefox), Chrome is default enabled.