I describe here a real-time vision-based gesture recognition system used in interactive computer music performances. The performer moves his hands in a video-camera capture area, the camera sends the signal to a video digitizer card plugged into a computer. By processing the reconstructed images of the performer’s hands in movement the computer detects x-y positions, shape (posture) and angle of rotation of both the hands. Data extracted from image analysis every frame is used for controlling real-time interactive computer music performances. Two approaches, one more formal the other really operative, are presented.