We develop a system that enables a beginner of musical performance to easily participate elaborated ensemble performances with an experienced musician. This system is composed of two key elements: a robo-instrument, which enables the beginner player to take a part in the ensemble by automatic mechanical controls of fingering motions for musical instruments such as saxophone, and a realtime music alignment (score following) method that takes the performance signal of the experienced musician as input and estimates the current position in the performed musical score. To realize a seamless and flexible ensemble, we study a MIDI-based score following method that can treat performance errors, repeats, and large skips, and a method for predictive transmission of MIDI events to the robo-instrument. We demonstrate this system with pieces for piano and saxophone, where the latter part is played with a robo-saxophone, and discuss new problems of music information processing required for extending the research on such realtime interactive systems.