Wyoming GLaDOS

Wyoming protocol server for the GLaDOS text to speech system from R2D2FISH. It uses CUDA acceleration if supported.

The server part is an heavily stripped down version of wyoming-piper and the gladostts folder is a Git submodule of R2D2FISH's repo

TODOS:

• Docker image ✅
• Automatic download of model files ✅ (please use download.py)
• Download from homeassistant directly (for homassistantos users)
• Readme steps for connecting this to homeassistant
• Add options to the docker container
• Code cleanup
• Speedup the tts engine for rtx gpus ? See here
• PyPI package ?

How to run locally

git clone --recurse-submodules https://github.com/nalf3in/wyoming-glados # You will probably get a git lfs error, this is fine
cd wyoming-glados
python3 download.py # Can be done in the venv if you prefer
python3 -m venv .venv
source .venv/bin/activate
pip install -R requirements.txt
python __main__.py --uri tcp://0.0.0.0:10201

Docker Image

docker-compose (recommended)

---
version: "3"
services:
  wyoming-glados:
    image: docker.io/nalf3in/wyoming-glados:latest
    container_name: wyoming-glados
    ports:
      - 10201:10201
    restart: unless-stopped
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

docker cli

docker run --restart unless-stopped -p 10201:10201 --runtime=nvidia docker.io/nalf3in/wyoming-glados:latest