Programming a Music Synthesizer through Data Mining

Loviscach, Jörn

Proceedings of the International Conference on New Interfaces for Musical Expression

Sound libraries for music synthesizers easily comprise one thousand or more programs (”patches”). Thus, there are enough raw data to apply data mining to reveal typical settings and to extract dependencies. Intelligent user interfaces for music synthesizers can be based on such statistics. This paper proposes two approaches: First, the user sets any number of parameters and then lets the system find the nearest sounds in the database, a kind of patch autocompletion. Second, all parameters are "live" as usual, but turning one knob or setting a switch will also change the settings of other, statistically related controls. Both approaches canbe used with the standard interface of the synthesizer. On top of that, this paper introduces alternative or additional interfaces based on data visualization.