I have not been tempted to narrate my own work, and I don't listen to audio books (one day I should give it a try), so I hadn't said anything on this thread. But I do use text-to-speech software as a form of editing, so I would add...
Good text-to-speech software allows you to select the voice - very few of which sound anything like Stephen Hawking's machine. I use some software called
TextAloud, and they offer a wide selection of voices. For most of my editing work I use one of the two Australian voices that I purchased, one male, one female. (The American voices didn't do it for me, I spent all my time listening to the accents. Some of the UK voices would work for me, but I found the Australian voices had fewer distractions.)
I have found it surprising how well these voices work, reading the text with surprisingly accurate inflections taken just from the context and punctuation. With more work on the text to tweak things a little more, I think it would be possible to produce an acceptable final result ... but since I don't have experience with professionally created audio books I can't offer a real comparison. I'm not about to suggest these should be used instead of true professional production, but the idea that computer generated voices should sound like Stephen Hawking's machine seem to be long gone.