@Sigg3: I had plans for that before, but it turned out to be much much more complex, and if you look at the lipsync in the videos the original poster posted - that is not even close to any real speaking even if it's done by a professional company.
Once the model is done it's not that much of effort to add new animations because all you need is new spoken text from the TTS and a camshot from my lips moving for tracking. Also, we don't have that much cases which need that much of flexibility I think. For example the Install Wizard, that is not a thing which will change every day...
@thom: I DO work on it and still I'm interested in ideas, so it's alright if you don't care but why don't you just ignore this threat instead of spaming it all the time?