A Jar, a Terracotta Pot, and Plenty of Room for Improvement is the multichannel live version of a minimalist, live-coded percussion piece that I've been developing since 2023.
First recording:
The focus for this piece is on exploring Mégra's Markov chain learning algorithm. Besides the input data (in this piece, small strings, often representing drum-rudiment like structures), the learning algorithm has a variety of parameters, such as the memory length or the maximum model size.
During the performance, I explore the spatial distribution of the sounds as a fundamental parameter, learned in near real-time from patterns and modified through the manipulation of the patterns and other parameters of the algorithm (i.e. the memory length). Other musical parameters such as pitch, rhythm and dynmamics are controlled in a similar manner, in a paradigm related to serialism.
To emphasize the structure, the sound material is reduced, to only two percussive, found sounds - the titular terracotta pot and glass jar (plus some pitch shifting, and some accidentally recorded birds in the background).
In contrast to current developments in machine learning, everything is learned live on stage, on the visible hardware, not on other people's computers.
The exploration of parameters performed through as code gestures. The dense, polyrhythmic textures are widened into an open cloud of percussive tones. The Markov chains (a kind of pattern in themselves) are visualized to make the performance more transparent.
Three instances of the project have been performed so far: