Here is a thread with some of my progress on text-to-speech. In summary, I've amassed a collection of utilities which, when combined, can perform servicable text-to-speech on the Empeg. It doesn't sound like the passenger next to you is talking, but it's understandable speech, certainly good enough for many applications.
However, these apps need to be all glued together. I have an "action item" to email the author of the speech synthesis package and see if he can do something about adding some code to handle the Empeg's quirky soundcard requirements natively, this would be the best solution. If he can't do that, then we need to glue together the sample rate conversion and muck with the sound driver code ourselves. I don't think I'm capable of that kind of coding right now, so I posted the aforementioned thread, and nobody seemed really interested in making it happen.
Anyway, I will try to get back in touch with the Flite author and see if he would be nice enough to add Empeg support, or see what it would take to twist his arm into doing it.