| Description |
This thesis describes a novel technique to algorithmic composition which combines two data structures-an N-gram language model and an association network-with a user friendly interface meant for music producers. The MelodySmith is a music composition tool which incorporates this logic into a product which music producers can use alongside their usual production workflow to generate ideas, samples, or even complete melodies for their music. Given a corpus of MIDI files, this algorithm will compose an original melody which aims to be stylistically similar to the corpus as a whole. An important part of modern music production is having fine grain control over each tool in the producer's work chain, so additional effort has been made to allow producers the ability to choose exactly which musical elements will be utilized by the algorithm, and to what extent each of those elements should contribute to the final composition. My work on the MelodySmith comprises all the logic necessary to replicate the style of a user-provided corpus into new and creative melodies, and to allow users to influence those melodies using simple, musically relevant controls. |
