It synthesizes pure tones to harmonize with the notes it hears by playing their harmonics -- overtones or undertones -- or other harmonically related notes. For each detected note, it plays up to 3 notes chosen at random from a set of harmonics. The set can change by setting a parameter, by user interaction (e.g. by pushing an unlabeled button somewhere in the room), or in response to analysis of the sounds.

The strength and duration of each harmonic is determined by the amplitude of the detected note. There are three voices each playing one to three notes, so the system can play up to 9 notes at once. Each voice will only respond to detected notes when it is finished playing its 3 notes. This is why the system often doesn't seem to be responding directly to "prompting." It listens to you when it feels like it, and comments upon what it hears. Even though what it's playing is directly determined by what you're doing, you can't really control it like a musical instrument. It is talking about you, after all.