| commit | author | age | ||
| 4e7cca | 1 | # Text-to-speech API based on Facebook's MMS |
| JK | 2 | |
| 3 | Simple Python-based container with everything that is needed | |
| 4 | to have a self-hosted web-based text-to-speech API. | |
| 5 | ||
| 6 | ## Using the container | |
| 7 | ||
| 8 | Just run: | |
| 9 | ||
| 10 | ``` | |
| 11 | docker run -d -p 8000:8000 ghcr.io/jacekkow/docker-text-to-speech-api-mms:master | |
| 12 | ``` | |
| 13 | ||
| 14 | and then visit http://localhost:8000/docs | |
| 15 | ||
| 16 | There is a simple `/sythesize` endpoint that expects a JSON and returns a wave file: | |
| 17 | ||
| 18 | ``` | |
| 19 | curl -o result.wav -X 'POST' \ | |
| 20 | 'http://localhost:8000/synthesize' \ | |
| 21 | -H 'Content-Type: application/json' \ | |
| 22 | -d '{"language": "en", "text": "Sample text."}' | |
| 23 | ``` | |
| 24 | ||
| 25 | ## Adding languages | |
| 26 | ||
| 27 | Currently only English and Polish models are included in the image. | |
| 28 | ||
| 29 | To add additional languages you can simply add extra entries | |
| 30 | in `src/config.py` file and rebuild the container. | |
| 31 | ||
| 32 | Supported language model codes can be found on Hugging Face: | |
| 33 | https://huggingface.co/models?sort=downloads&search=facebook%2Fmms-tts- | |
