University of Toronto ECE Students use AI and deep learning to create music
If you’re a regular on TikTok, you may have stumbled across people on your For You Page lip-syncing or dancing to bizarre audios with strange lyrics that seem completely out of place. After a quick deep dive into the audio uses, you realize it’s AI-generated music. And while the subject matter or lyrics might not make total sense, you have to admit – it's catchy.
Yiming Chen and Yanhao Li are two graduate students from the Department of Electrical and Computer Engineering at the University of Toronto. They developed an innovative Artificial Intelligence (AI) project using deep learning to generate music.
Using an extended, short-term memory neural network, the team trained the model to generate accurate melodies by feeding diverse music data sets. This means that the program can remember the order of things, helping it make predictions based on a sequence, like guessing the next word in a sentence. Thus, making sense for an AI program fed a series of music notes to determine and output the rest of the song.
They believe that their program will be able to support musicians and producers in coming up with music ideas and patterns or enhancing and improving their current music. However, this program is not just for professionals, it also gives non-musicians or aspiring musicians an opportunity to create music in an accessible way.
By simply humming a short tune, their program will help turn this into a complete song. Chen and Li say that in the future, they want to create a large language model, or in other words, a program that lets people generate music just by inputting a text description of it. This potential is not just exciting, but also gives us hope for the future of music creation.