Contenerized text-to-speech engine based on Facebook's Massively Multilingual Speech model with web API
edit | blame | history | raw

Text-to-speech API based on Facebook's MMS

Simple Python-based container with everything that is needed
to have a self-hosted web-based text-to-speech API.

Using the container

Just run:

docker run -d -p 8000:8000 ghcr.io/jacekkow/docker-text-to-speech-api-mms:master

and then visit http://localhost:8000/docs

There is a simple /sythesize endpoint that expects a JSON and returns a wave file:

curl -o result.wav -X 'POST' \
  'http://localhost:8000/synthesize' \
  -H 'Content-Type: application/json' \
  -d '{"language": "en", "text": "Sample text."}'

Adding languages

Currently only English and Polish models are included in the image.

To add additional languages you can simply add extra entries
in src/config.py file and rebuild the container.

Supported language model codes can be found on Hugging Face:
https://huggingface.co/models?sort=downloads&search=facebook%2Fmms-tts-