℗ Generative Generation

Musical Comprovisation / Performance / Generative Art

Generative music is an art form whose disciplinary field has expanded as a result of multiple artistic, technological and aesthetic mutations. Since the 1940s, generations of composers and performers have employed various generative models, in line with the mutations, discoveries and revolutions in the fields of science, computer science and machine learning. From probability theory to chaos theory, from expert systems to complex systems, from machine learning to generative artificial intelligence, generations of “generators” have succeeded one another like so many markers of the dominant currents of thought of their eras.

In generative generation, a transgenerational approach combines two generative models in an interaction that could be evoked by the following metaphor: Let’s imagine a skull is opened up and we’re able to activate each neuron separately. The complexity of the brain is so great that any manual, punctual action would lead to a disappointing result (perhaps a nervous twitch of the eyebrow at most). The use of a generative probabilistic model, with a capacity for exploration whose complexity is controllable, offers us the perfect tool for creating generative artificial patterns of musical loops of neuronal activation.

To realize this hybrid machine, mixing different eras and theoretical approaches, Georg Hadju’s probabilistic model (DJster), heir to Clarence Barlow’s model (Autobusk), controls the navigation of several “activation heads”, each operating in a latent space of a Realtime Audio Variational autoEncoder (RAVE). The sound of the musicians’ instruments acts as a disrupter of these models, moving these playback heads within the latent spaces.