EBSCO recently announced that they have "added text-to-speech (read aloud) support to EBSCOhost databases, including its major school and public library databases, by embedding Texthelp Systems’ SpeechStream toolbar—a valuable benefit provided at no additional cost to the user. Users will be able to take advantage of this new feature with any full-text articles available in HTML." It's actually pretty good, IMHO. The speech is of course somewhat robotic, but not nearly as bad as many I've heard. You can choose between American, British, and Australian English, read at slow, medium, or fast speeds. These days I listen to podcasts at 2x speed, so even the fastest option here seemed a little slow to my ears. Have a listen, what do you think?
Oh, and the article I was looking at for this demo is worth a read too:
Web Scale Discovery: The Future's So Bright, I Gotta Wear Shades. Jeff Wisniewski. Online; Jul/Aug2010, Vol. 34 Issue 4, p55-57.