IMO, most audio read by humans (esp. voice actors) are far preferable to machine readings. Also, I found no demos on that page.