Gesture modeling and animation based on a probabilistic re-creation of speaker style