I made my own little service that converts any webpage to hopefully the parsed content then uses Google TTS and then published it to a bucket and s3 feed and I listen to them on my phone before bed.
Do you see anything that we could add to the tool to make it more useful for you?
One thing we played around with, which works quite well is directly interfacing it with GPT-realtime. This then allows one to talk to it about the paper. It also solves the problem of the language since any person can talk to it in their own language, which could increase accessibility in science. I have shown it to some Japanese colleagues the other day and they could interact with it in Japanese which was quite amazing.