• A real-time multimedia exploration using genetic algorithms to evolve words and sentences from random letters for silent film intertitles
• Developed over 6 years using Java, and Mbrola voice synthesis

Exploring the uses of evolutionary algorithms, particularly genetic algorithms, for abstract musical use is not necessarily revolutionary, but it’s an idea that’s interested me for a long time.  So much music is based on rules, and the breaking of those rules, that the structure behind genetic algorithms seems like a natural fit.

An overly simplified look at the structure is as follows: Start with a population of something.  Encode it. Rate each individual versus some fitness function to determine it’s overall worth.  Take the most fit individuals.  These are now our parents for the next generation.  Mate the parents to create children.  This is accomplished by slicing the encoded information for each parent and joining the parts to make a child.  Make enough children for the next generation.  Decide whether or not to stop, and then if not, then start the process again. This is a simplified explanation, but it gets the basic process across.

In these pieces, I’ve written a program to evolve words from random combinations of letters.   The words are evaluated on the very simple notion that vowels are better next to consonants.  This, of course, neglects the useful qualities of dipthongs and consonant clusters, but still serves as a mechanism to produce interesting and perhaps even viable results.  What you hear in these pieces is the most fit word in subsequent generations as attempted by simple text to speech converters. These are meant to be able to handle the basic fundamental parts of speech, and fair relatively well here.

Looking forward

• Evolve digital music instruments from random unit generators 
• Develop interchangebale language maps to facilitate evolution of words in other languages
• Create more sophisticated fitness functions to better rate dipthongs and consonant clusters
• Use the volume envelopes of synthesized speech to control motors attached to physical mouth on sculpture or model face

Thanks to Holly Faurot, Sarah Paulson, and Nick Didkovsky

~ Available for museum installation/performance ~
~ Please contact me for more information ~