Use local speech model

[ilesinge]

Currently the speech are generated through Google Cloud TTS free tier.
Ideally there would be no dependency on external service for that.
Try using a local model such as Supertonic: https://github.com/supertone-inc/supertonic