Hehe, I don't think it is that simple. Speech to text relies on complex algorhytms analyzing hyphens (am I correct?).
At least according to the seminar I got from dragon naturally speaking. If I can dig up the powerpoint, I could share it, but I'm not certain that it would help.
Text to speech seems a little bit more probable, but you'll still need an extra sdk.
Have you researched the technical limitations of your project?
If you are still going on with this, I'd suggest you research a bit about speedlib. I think that is a library you might need.
If other people with more expertise can help, please do so. I'm not a master in speech recognition myself...