Mary Piper is a MaryTTS compatible web api for PiperTTS. It automatically resolves and downloads voices if they are not present yet.

Create a virtual environment.

python -m venv .venv
source .venv
pip install -r app/requirements

Run server.py

source .venv/bin/activate
python app/server.py

This will spawn the server at 0.0.0.0:59125.

Install docker and docker-compose.

Build the docker container locally.

docker-compose build

Start the container

docker-compose up -d

This will spawn the server at 0.0.0.0:59125.

The main goal for this project is to use PiperTTS in Rhasspy 2.5. Therefore the project's aims is to create a drop in replacement for MaryTTS compatible TTS systems that allows to use PiperTTS voice generation and can be easily configured in Rhasspy.

• The server does not have any error handling and is still pretty rough
• There is no caching of already used voices in memory
• There is no implementation for excessive multiple parallel requests