Data
This project was developed primarily using instrumental audio downloaded from Youtube to .WAV files, starting with synthesized piano tracks for the simplest tests. Recordings of real-world instruments (still primarily piano) were then used to test how well our program handles noise, reverberation, and more complex articulation. As a final real-world demonstration, the team collaborated with University of Michigan School of Music students to demonstrate the program's function as music was being played in real-time.
The University of Michigan Student musicians we collaborated with are Max Crandell and Spencer VanDellen.
Large datasets of royalty-free music were identified. These were not used for this project, but could be useful should in a future adaptation of this project, perhaps involving machine-learning techniques which require large datasets.
Royalty free music and 'free to download' music - Free Music Archive
Youtube Links
These files were downloaded as WAV files using YouTube to wav - Motionbox and Y2Down.CC. We do not claim any legal rights over these recordings, and they are used strictly for educational purposes.