Show/Ask HN: Latex to audiobook – improving a hacked together proof of concept

I quickly hacked this together as a PoC to see if Google Cloud TTS's WaveNet voice was "good enough" for reviewing my thesis writing on the go in audio form.

It works quite well, I noticed a bunch of writing errors that I hadn't picked up visually. And it's just about listenable enough that I can see this being a good way to reduce some screen time while the weather is good.

The code is obviously pretty terrible as a starting point for anything more complete or complex though :)

Does this idea resonate with you? How would you suggest rewriting it? What features would you add next? Have you already seen this done better (I couldn't find much prior art)?