Pseudo-Genetic Algorithmic Composition
Author: Harald Schmidl

Welcome to the project home page
This work presents an approach for algorithmic music composition using augmented genetic algorithms. We focus on generation of new material that lies within the context initially intended by the user, minimizing user input, avoiding randomness, and ensuring realtime execution. We achieve the latter by devising straightforward operations for generating new musical material and a fast filtering algorithm for selecting "good" samples.
The algorithm lends itself well to a theme and variation approach as used in traditional composition styles, such as canon, fugue, or counterpoint.

Resources
Access the paper and some mp3 samples through the links below.

The paper in pdf format as presented at GEM'08, The International Conference on Genetic and Evolutionary Methods, Las Vegas.

Sample 1: canon-oriented piece for recorder, oboe, and baritone sax.
Sample 2: cocktail lounge piano music for three hands.
Sample 3: less conventional composition for violin, cello, contrabass.
Sample 4: 80s disco for two synthesizers and percussion.
Sample 5: piece peace for sitar, organ, trumpet, and hand clap.

© 2008 Home