More naturalistic responses in the voice output

[JohnnySn0w]

Piper is currently very rudimentary in how it's being used. Either using piper, or using other methods, the voice can have pauses and other types of interjections that will make it sound more naturalistic.

[JohnnySn0w]

StyleTTS2 would probably be a better bet, if generation times aren't too bad, or scale well with compute resources.