Damien Sprockeels- Melodizer : A Constraint Programming Tool for Computer-aided Composition

CONFERENCE

Abstract :

 

 My master's thesis presents the design of a tool aiming to assist composers in their creation process. It uses constraint programming to enforce musical rules through two constraint satisfaction problems (CSP). This thesis is built on an already existing thesis focused on rhythm, and is as a result more pitch oriented.

 

    Gecode, a powerful constraint solver, is used to model the musical problems in terms of constraints. OpenMusic is used to represent the musical aspect of this thesis, and as a support for Melodizer. GiL, an interface between Gecode and Lisp, is used to connect Gecode and OpenMusic and allow for the creation of constraint problems in a musical environment.

   

    A tool, Melodizer, is designed to help composers increase their expression and composition abilities by using constraints to generate melodies. It provides an intuitive and interactive interface making it easy for composers to customize their problem by giving input and setting parameters such as solution variety or optional constraints, and allowing him or her to edit any proposed solution.

    

    Two CSPs allow Melodizer to model a variety of problems using a set of mandatory and optional constraints. It does not aim to generate ready-to-use music but rather to provide interesting out-of-the-box ideas for composers. It provides a solid base towards a bigger picture allowing composers to create symphonies from a similar tool.

    

   During my demonstration, I will show how Melodizer can be used to help composers write musical pieces from their original ideas.

 

Bio :

I'm a 25 years old student that recently graduated from College as a civil engineer, and my master's thesis was about computer-aided composition. I designed a constraint-based tool allowing composers to generate melodies from ideas they had, and also allowing them to store and combine the generated melodies easily.

@Damin Sprockeels

 Back to the abstracts collection