Text to speech technology has come along way from when I first heard a computer speak. The following video is a TED speech by Roger Ebert. Roger lost the ability to speak, and for that matter eat, when he lost his lower jaw to cancer a few years ago. For part of his speech he is using the Alex voice on his Macbook. Take a listen. It's not bad, but still far from perfect.
I've learned that one of difficulties is in the fact that a computer doesn't breath. In addition to naturally pauses indicated in sentences by commas, we also pause elsewhere in our speech to simply account for our need as humans to breath. I have been experimenting with this in Adobe Captivate's voice narration capabilities. the North American voices included with the product are from a company called Neospeech. Their voices use a text to speech programming language known as VTML. In addition to the text you want spoken you can include VTML tags which will indicate items like speed, pitch, pauses, and a few other items. As I experiement with this, I find my narrative is sounding less and less like a robot and more and more human. It's not perfect, but it is getting better.