Non-robotic realistic sounding TTS for Linux FOSS

I’ve tried eSpeak, RHVoice, and Festival even with HTS voices, and it sounds very robotic. Are there some recommendations for a TTS that has a more human-like voice?

Looking at this site I guess gTTS might be worth a go - python-gtts 2.5.1-1 is in the AUR although Arch install instructions aren’t mentioned on the site…

https://rhasspy.github.io/piper-samples/

What’s the use case?
It’s harder to find ones that turn text to speech on the fly, accessibility use cases and so on.

For entering text and getting an audio file or similar there are things like speech note and some decent options through Python (at varying speeds depending on your hardware).