Ask HN: Offline Text-To-Speech/Speech-To-Text Frameworks for Raspberry-Pi

Hi, Based on your experience, what's the best way to enable a natural way to read text using RPI, and transcribe speech commands on RPI. I tried most of some of the open source frameworks (Jasper, eSpeak, .. etc) but they don't feel natural. I tried also Watson, but the relying on the cloud is not really an optimal solution.

What are your recommendations ?