It's still crap. The french lab tcts[1] a few years ago had a speech engine which added tongue flicks, breathing and lip noises (Don't know the technical name for these) as well as using multiple forms of a sound to generate words. IE: match up the correct version of each letter to say "match" rather than force one form to work.<p>This is still short trousers for text to speech. But then text to speech isn't a big selling feature for computers - yet. Think mini, data specific watsons on your iphone. Pretty soon we'll have HAL-9000's in our pockets.<p>1. <a href="http://tcts.fpms.ac.be/" rel="nofollow">http://tcts.fpms.ac.be/</a>