Baidu Research have published a paper about their work using neural networks to build a better system for converting written text to audible speech. This is really neat. Listen to the samples in the summary. It is trained on audio files of recorded human speech correlated with written text.
Today, one has two options for listening to a book: One can listen to an audio book expensively narrated by a human or have your computer convert the text to speech automatically, which is currently of very poor quality. Tomorrow, we should have high quality computer narration which will be pleasant to listen to.
A step after that, one should be able to take a descendant of this technique and train it on historical figures for whom we have audio recordings (e.g., John F. Kennedy, Winston Churchill, Ernest Hemingway) to have it speak like them. One could have Churchill narrate his books or those of others. I look forward to hearing a simulation of Ernest Hemingway narrate “The Old Man and the Sea” and JFK reading aloud the works of Doctor Seuss.