CataRT is a concatenative real-time sound synthesis system, plays grains from a large corpus of segmented and descriptor-analyzed sounds according to proximity to a target position in the descriptor space, controlled by a mouse or by external controllers.