TransPose, by Feedtank, is a computer vision based instrument that addresses the absence of physical human expression found in most electronic music performances, as well as helps define a new movement based language for creating sound.

Computer vision technology captures the performer’s physical actions captured and translates them to audio in real time. The performer sits in front of a camera, and his or her silhouette is projected in front of them in relation to a number of predefined trigger areas called “noteboxes.” Using his/her silhouette to overlap the noteboxes, the performer triggers various tones.

In rhythm mode, six different drum sounds are mapped to noteboxes which the performer can strike to play a beat. In the melody mode, two instruments are mapped to two elongated noteboxes on either side of the performer. The hand’s vertical positioning inside these noteboxes determines the tone’s pitch. Its horizontal positioning determines the tone’s volume. The performer can change between instruments and record loops that can be turned on or off to create layered compositions.

Window Media Player + Quicktime videos.
See also Feedtank’s Dance floor moves and Full body games.