I've just been shown a demo of Veepers from a company called Pulse.  The best way of describing a Veeper is that it is like a streaming video.  In this case it was of a person speaking, but they also have text-to-speech capability as well.  So far, nothing new.  However the Veeper uses a 3D model generated from a set of still photographs to achieve very low bandwidth use.  The effect is remarkable and I recommend seeing the demo.

Source Louise Kehoe of the FT