Project Abstract

I came up with this project because I am a musician, and market available musical composition interfaces weren’t implemented to my liking. My tool, implemented as a web app, would allow easy input from the microphone to allow the quickest start to composition possible. It would also be integrated with AI so that users could get suggestions on how to continue their piece if they ever became stuck on what to write. The tool would also output a finished composition to sheet music so that the user can play at home. The nature of this project is highly experimental, as a product combining all of these elements does not exist, so the dissertation was primarily focused on whether or not it could be done.In the end all project aims were achieved, save the sheet music output. The result is a sleek web application that is fully capable of composing a piece of music. Development was done in React.js, used Magenta.js for AI music generation, and used React-Grid-Layout to hold the notes in the workspace. Magenta.js takes the user’s current input into its generation algorithm, and will generate a continuation to fit their piece, which can be any length and can be placed anywhere.The main workload was translating between different forms of data, from audio waveforms from the mic, to the note data structure required by magenta, to the visual form of the music you see in the workspace. This project has shown me just how difficult creating a multimedia application can be, but that the challenge of doing so was worth it, because I believe the application created shows itself as a powerful proof of concept.

Keywords: Machine Learning, Music, Web Application

 Conference Details

Session: Poster Session B at Poster Stand 39

Location: Sir Stanley Clarke Auditorium at Wednesday 8th 09:00 – 12:30

Markers: Neal Harman, Matt Roach

Course: BSc Software Engineering, 3rd Year

Future Plans: I’m looking for work