Microsoft showed a demo of new software at TechFest 2012 that can basically create an avatar of yourself and speak on the fly, using your own voice. It does so by only using a few minutes worth of actual spoken text from the user. These are called real-time multilingual vocal translations. Monolingual TTS (text to speech) can currently handle 26 languages but isn't instant.
This software demoed by Microsoft can even add foreign language pronunciation of place names with directions using your native speech. There is a 3D image of your head to go with it, along with animation of your lips to foreign languages that you may not even know how to properly speak in real life.
This video, eh hem, speaks volumes to how the software works:
Use this QR code in a QR reader application on mobile to open quickly on a mobile device