Mix this with AI creating video from audio (http://spectrum.ieee.org/tech-talk/robotics/artificial-intel...) and you can make anyone say anything.