MusicNotes

Bianca Wilson, Franklin College of Arts and Sciences, Hugh Hodgson School of Music, AB Music and AB Cognitive Science, 3rd year undergraduate


We’ve all had songs stuck in our head at some point—that catchy commercial jingle, a favorite movie
soundtrack, or a random childhood tune. MusicNotes takes advantage of this and helps students study for exams in a unique way: through song! This Python-based program allows students to upload their notes or flashcards, which are then transformed into short, catchy songs that can be downloaded as mp3 files. This is made possible by two types of generative AI: Google Gemini, which converts the notes into song lyrics, and YuE, a MusicGen software that sets those lyrics to an AI-generated melody. Users can even choose the style of music they prefer for their song.

The inspiration for MusicNotes came from research in music psychology, which shows that the brain processes musical information differently from standard linguistic information. This explains why people with conditions like Aphasia, who struggle with regular speech, are still able to sing. Similarly, it’s also why children find it easier to learn things like the alphabet through song, rather than through rote memorization. Music has a unique way of helping the brain encode and retain information—but while singing songs might seem simple for humans, it’s incredibly challenging for a computer. YuE, the MusicGen model used in this project and the first open-source AI model capable of setting lyrics to music, must account for over 7 billion parameters when creating a 30-second song.

MusicNotes merges the generative text features of Gemini and the music generation capabilities of YuE using Python and the Google Colab platform. Colab’s cloud-based environment allows users to bypass the massive memory constraints of YuE, making it possible for anyone with a Google account to use the program on most devices. By adjusting the prompts fed into the Gemini engine, MusicNotes converts notes on any subject into song lyrics. These lyrics are then passed to YuE, which creates a melody and “sings” the lyrics over it based on user-selected musical preferences. All the user needs to do is upload their notes, choose a music genre, and follow the outlined steps to generate their song. Even with the same lyrics and style prompts, each song is unique—offering students a fun, efficient, and engaging way to study for their next exam.

OSZAR »