Human motion is incredibly complex, subtle and diverse. People do an amazing number of things in an amazing number of ways. Example-based animation is an attractive technique for reproducing these movements because it avoids the challenges of trying to create algorithmic models all of the diversity and complexity. Rather than try to understand each possible movement or style, we simply need to capture the performance of an actor doing the movement. To date, Synthesis-By-Example (SBE) approaches, methods that combine recorded movements to form the ones needed at runtime, have become standard practice in the creating animated characters in interactive systems such as games.

In this talk, I will give an overview of how and why Synthesis-By-Example is used to create character motion for games and other interactive systems. The basic idea of SBE approaches is to use generic algorithms that assemble desired motions from collections of captured examples. The specifics of the movement come from the data used. I will review the basic methods, particularly concatenation and blending. I will show how these methods are combined to synthesize novel motions at runtime for animated characters, and how automated tools not only simplify the creation of such characters, but also make possible more complex synthesis.