My friends that use synthetic voices prefer cleanliness of the older and familiar voices. One friend listens at about 900 WPM in skim mode and none of the more realistic voices work well at those rates.
Every once in a while I'll hear a blind person's phone audio while I'm out and about and it sounds like an unintelligible stream of noises, but they're interacting with it and using it and getting information from it. It's amazing, a whole other world of interaction from what I experience with my phone. I kind of want to learn how they're interacting with it.
Back in the 90s we reverse engineered / patched the classic MacInTalk to speed up playback. Our testers had it cranked so fast as they navigated the UI that to me it sounded more like musical notes than text.