New Languages For Physical Modeling Synthesis

The first great computer music language was Music V [2]. It set the stage for the next forty years: nearly all of today’s computer music languages are philosophical descendants of the Music N family. This family of languages is organized around the paradigm of a unit generator network. A unit generator provides a stream of audio samples; simple examples are a sinusoid generator and a generator which adds two signals together. Individual generators are kept simple and fast; waveform generators were originally implemented as flexible table lookups. This supports a method of producing sound by means of operations on other sounds: filtering, splitting, recombining. The unit generator paradigm has been very successful, partly because it is flexible and partly because synthesis using unit generators can be (relatively) computationally inexpensive. However, it can be difficult to reproduce a wide variety of realistic playing techniques and instrument responses using this kind of synthesis.