Artificial Intelligence in Music

Rudresh Veerkhare
4 min readMar 24, 2021

--

“Music is the art of arranging sounds in time to produce a composition through the elements of melody, harmony, rhythm, and timbre”. Well, Wikipedia might say this but music is not just arranged sounds, Music is pure bliss, it’s a relief from pain, it’s creativity in the purest form, and a universal language.

It has been proved that music increases creativity, reduces stress and anxiety also makes learning more fun and memorable. While it’s the Ultimate form an art, it’s also truly complex in nature as music is directly linked with human emotion. and while keeping this in mind we’ll discuss the use of AI in Music with minimum technical details.

credits — liveforlivemusic.com

What is Artificial Intelligence?

Artificial Intelligence is often described as machines that mimic cognitive functions that humans associate with the human mind. AI has evolved tremendously over the last decade after the rise of Deep Learning. Deep Learning is a machine learning method inspired by the human brain neural network.

AI has made substantial advancements in Computer Vision and Natural Language Processing, from generating real-looking human faces to writing Shakespeare’s plays.

Deep Learning (DL) utilizes a ton of data to learn the underlying patterns and learn them without any need of explicitly stating rules, just like how a human brain would learn things by observation. Music is really a complex art, but deep learning is exactly for things like this.

Here are 3 main aspects of using AI in music which we’ll discuss :

  • Recommendation
  • Identification
  • Generation

Music Recommendation

You suddenly remembered a song and now you want to listen to it, you open your music player and start playing that song and after half an hour later you realize that you couldn’t stop cause the next song was perfect for your mood and you kept on listening.

That’s exactly what AI recommendation is, this field of AI is called Recommendation Systems. In the area of music, the algorithm first tries to learn the relation between different songs and based on that it finds out the similarity between songs. This is called Content-Based Filtering (CBF).

An example of a song recommendation system is the autoplay feature in Spotify.

Credits — djinit-ai.github.io

Music Identification

If you are a music junkie then you know how annoying an earworm can be if you don’t remember the song. Well, back in 5 years from now, you’d be in misery until you accidentally come across the song or someone points it out for you.

But not now!, all thanks to the AI, earworms are not annoying anymore cause you can just Hum or sing or whistle or play the song using instruments and AI will tell you the name of that song.

Demo of Hum To Search

This Machine Learning(ML) algorithm developed by Google is called hum to search. This ML model converts input into the special fingerprint of the song and then matches it with existing fingerprints of thousands of songs, based on the similarity between fingerprints it gives the list of similar songs.

Music Generation

Although Music is very complex, requires years of practice and creative thinking, Artificial Intelligence also has come a long way. There is really promising research going on in the field of Music generation, one of which is MuseNet. It’s a deep learning model which can generate 4-minute musical compositions with 10 different instruments.

Screenshot of Music Generation Demo By MuseNet

AI can generate music on its own but it’s still in the primitive stage, rather than generating whole music, AI can also be used to expedite the process of generation, an example of it would be TensorFlow’s Magenta. Magenta is an “open-source research project exploring the role of machine learning as a tool in the creative process”.

Future of AI in Music

AI is already doing things that we’d have thought impossible 2–3 years ago, and considering this progress I think, in the field of Music Generation, days are not far when AI will recommend songs not just based on the song history and previously played songs, but also on other factors like heart rate, brain waves, stress level, etc. There might be a time in the future, not far from now, where everyone would have their personalized music generated by an AI. Possibilities are endless, but whatever it’ll be, will be amazing!

--

--