While there have been several robots designed to play musical instruments, most are specialized (non-humanoid) designs that do not attempt to model a human performer's specific gestures and movements. Musical instruments are, of course, designed to be played by humans, and the process of developing humanoid performers offers the potential of revealing new insights into the control of instruments in a musically expressive manner. This paper describes our system that enables a small humanoid to play a music keyboard and also respond to acoustic input. The system uses a combination of pre-programmed positions, audio feedback, and self-calibration methods in order to perform simple two-finger piano pieces. As with human performers, control adjustments can be made in real-time when musical errors are detected during the course of a performance. This system is intended as a proof-of-concept and prototyping platform for humanoid music performance methods, which are ultimately destined for the Hubo adult-sized humanoid. The current system allows us to develop algorithms in a low-risk environment before moving to the much more capable, but costly, Hubo platform. In developing and analyzing humanoid performance gestures, this effort represents a step towards a better understanding of human musicianship, musical expression, and creativity, which may ultimately facilitate improved human-robot interaction.