Interactive music systems coordinate in real-time an artificial perception of dynamics of human musicians with timely execution of machine reactions. As every human performance will differ from another, it is a challenging task to be able to predict the behavior or such systems in response to any possible performance, and prevent unwanted outcomes. We present here the application of formal models and methods from the real-time systems verification literature to the static analysis of interactive music systems. We consider in particular the good parameters problem, which consists in synthesizing a set of timing parameter valuations (representing performances here) guarantying a good behavior of the system analyzed. The methods presented have been applied to the system Antescofo, and are general enough to apply to other interactive music systems.