CRISPRave is a live-coded, small data algorave performance where all musical material is learned live on stage from genome sequences of a cat and a dog, juxtaposing the two to create unorthodox grooves.
It is created with the Mégra live coding language, a lisp-like, free and open-source live coding language that I've been developing over the last eight-or-so years. In its current implementation, it's a standalone language implemented on top of the Rust audio universe.
Mégra is designed around Probabilistic Finite Automata (a formulation of variable-order Markov chains) as its core data structure, which allow for wide range of musical structures to be expressed and learned from sequential data, with a concise syntax.
Rather than using the Markov chain learning algorithm to supplement a performance, in this performance it's the center of attention. All models are learned, visualized and turned into sound live on stage, only with the hardware present on stage.
Editing the gene sequence is explored as a performative gesture to control the complexity of the resulting sound (hence the name). Both model parameters and input data are modified to control density and intensity of the performance.
Two instances of the project have been performed so far: