AlgoRemix: Algorithmically Remixing Songs Using Neural Networks and My Own Voice

Abstract

Because music is one of the last forms of non-physical art that the general public cannot easily generate using AI yet, my motivation for this paper is to make AI-generated music in a way that people will want to realistically use. Although AI generated music does exist, most current forms of AI generated music focus on creating every element of 2 completely new songs. However, I strongly believe that AI generated music will find its niche in altering existing music or as a tool to generate specific elements of music instead of entire songs at once. For example, a model could be used to generate a new instrumental for an existing song’s vocals or be used to alter the drums on a pop song to make them more jazzy. The reason behind this motivation is that, similar to other AI generation tools, creating a tool that allows more people to create is more valuable than a model that does all the creating on its own. We do not need more music, we need to make it easier for people to channel their creativity and create music.