How does this compare to say piper-tts?

I ask because their models are pretty small. Some sound awesome and there is no depdendency hell like I'm seeing here.

Example: https://rhasspy.github.io/piper-samples/#en_US-ryan-high