Noise and Numbers

Audion: Granular Synthesis with an Agent-based system

Audion: Granular Synthesis with an Agent-based system

This spring, I’ve been working on a system for generative music that uses a multi-agent simulation and granular synthesis to create unique soundscapes. I call this work-in-progress Audion, and I created it as part of my graduation project for my master’s.

You can read the entire paper here, but a summary and a video of some of its capabilities is below.

Audion

Composers and scientists have used algorithms to create new pieces of music for centuries, and they have often been inspired by natural phenomena for modeling their work. The advancement of recording technology and the development of the digital computer in the mid-to-late 20th century allowed musicians, artists and researchers to apply statistical models derived from natural sciences to musical composition and manipulate sounds on timescales that were previously impossibly small.

Audion is a multi-agent software system for granular synthesis and real-time musical exploration. Its design draws on the historical, artistic and scientific contexts for current approaches in generative music, paying special attention to methods inspired by cellular automata, multi-agent systems and artificial ecosystems. When I was considering ways to evaluate the development of Audion, I drew heavily from the work of Gordon Pask.

In Audion, “birds” (seen in color) interact with user-placed objects (in white) to generate grains of sound.

Audion allows a user to interact with a field of “birds” by placing digital objects in an on-screen environment. These objects modify internal states of the birds: hunger, fear and attraction. Individual birds and their internal states trigger and modify individual grains of audio, and ecosystem-wide variables control parameters of the resultant granular sound. The audio output references the ecologically inspired granular synthesis approach of Barry Truax and Damian Keller.

But there’s still a lot of work to be done, such as methods for applying the simulation’s emergent structures to audio output and creating environmental pressures in the simulation that are independent of the user’s manipulations.