This dockerfile packages up the StyleTTS2 notebook into a container thats makes it quick and easy to run StyleTTS2 inference using a pretrained single speaker model.
Per upstream liscense you must notify users that the output is AI generated.
docker build -t stylett2 .
mkdir ./output
echo "This is awesome text to speech" | docker run --rm -v ./output:/output -i stylett2
will produce a file in the output directory called output.wav
Follow the setup instructions from nvidia
And then add the --gpus all argument
echo "This is awesome text to speech using a gpu" | docker run --gpus all --rm -v ./output:/output -i stylett2
This coontaier uses phonemizer, which is GPL licensed. Check out the original repository for a MIT-licensed version!
All of my modifications are reworkings of the orginal MIT Licensed code.