Players and viewers of three-dimensional computer generated games and worlds view such renderings from the viewpoint of a virtual camera. As such, determining a good view of the scene is important to present a good game or three-dimensional world. Previous research has developed technologies to find good positions for the virtual camera, but little work has been done to automatically select between multiple virtual cameras, similar to a human director at a sporting event.

This talk will present a method of selecting camera feeds from multiple virtual cameras in a virtual environment using semiring-based constraint satisfaction techniques, a soft constraint approach. The system encodes a designer's preferences, and selects the best camera feed even in over-constrained or under-constrained environments. The system functions in real time for dynamic scenes, using only current information (i.e. no prediction). To reduce computation costs for a release implementation the SCSP evaluation can be cached and converted to native code. This SCSP approach is implemented in two virtual environments: a virtual hockey game using a spectator viewpoint, and a virtual 3D maze game using a third person perspective.