5

If you follow this article https://blog.rasa.com/how-to-build-a-voice-assistant-with-open-source-rasa-and-mozilla-tools/

it will have you download an 80MB file from:

https://drive.google.com/drive/folders/1GU8WGix98WrR3ayjoiirmmbLUZzwg4n0

called best_model.pth.tar

and using https://github.com/mozilla/TTS

you get the "sara" voice (which is very impressive)

Are there other best_model.pth.tar files out there with other pre-trained voices? Similar to CSS themes where you can download some or free, for pay or a really good one, do we have a pre-trained TTS model directory somewhere?

Andrew Arrow
  • 4,248
  • 9
  • 53
  • 80

2 Answers2

5

You can choose from any of the available pretrained models on the TTS repo. Some models provide a better audio quality than the one currently used in the blogpost. The process of using other models should be the same as the one covered in the tutorial.

Juste
  • 246
  • 1
  • 3
1

You can also use the tts-server and tts endpoints on console after pip install TTS.

$ tts --list_models will give you all the released models with some meta info. You can choose the one you like.

erogol
  • 13,156
  • 33
  • 101
  • 155