Tortoise TTS

99% natural sounding text to speech

Another awesome Gitrepo find lately is neonbjb/tortoise-tts which has the most natural sounding voices of all TTS platforms I’ve tried. But as the name suggests the processing time is very very slow even on a GPU (RTX 3090). It does detect dialog though and will generate two separate voices. This can be over ridden sometimes with:

--voice <name> --voice <name>

and using the same name for both. Will try the voice training feature next. Maybe cloning one's own voice has some interesting applications after all? Does this make voice based security obsolete?