I used to use the text-to-speech feature of pocket for this use case.