RiffShuffle: A Flexible Interface for Interactive Automatic Harmonization

We design, implement, and evaluate a system that allows users to interactively generate a chord progression to accompany a given melody, with the philosophical goal of bringing the experience of composing to people with some musical experience. We use a simple Hidden Markov Model to drive the chord recommendation engine, but add multiple adjustable parameters to the model. We expose both established and novel interface elements that allow users to control these parameters and interact with the chord recommendations in other ways. We specifically investigate the interaction of randomly sampling chords from the Hidden Markov Model, and find that it is well-received by users. User studies suggest that users enjoy using the system, and that they are able to use it to create chord progressions efficiently in ways that reflect their own preferences. Thesis Supervisor: Eran Egozy Title: Professor of the Practice