Using Artificial Intelligence in music production is an exciting and strange idea. Different AI tools have come and gone since 2016, but this is the first one that caught a deeper interest. When it comes down to it the output still seems random. Today that has changed. Magenta is a suite of Max for Live Devices that gives musicians easy access to experimenting with AI for generating new ideas.
Walkthrough of Magenta Max For Live Devices
Magenta released a free Max For Live Patch to give you access to the power of Magenta Studio. It has four devices you can use to make new musical ideas. This video is a walkthrough of Magenta by Google:
Download Magenta At https://magenta.tensorflow.org/studio/ableton-live
Magenta Studio of Machine Learning Tools
Magenta is an open source research project by Google. It was made to exploring the role of machine learning as a tool in the creative process. Magenta has taken massive amounts of data and created a database that can then be used with machine learning. A great example of this is Generate.
The Generate module using AI and Machine Learning to make brand new ideas. Here is Google’s explanation of how it works:
Generate uses a Variational Autoencoder (VAE) that has been trained on millions of melodies and rhythms to learn a summarized representation of musical qualities. Generate chooses a random combination of these summarized qualities, and decodes it back to MIDI to produce a new musical clip.
So, in other words, they have given Magenta a massive collection of MIDI they have collected from live performances and music. The AI then sees similarities and tries to make a new melody that would be similar yet different from the originals.
Artificial Intelligence as a New Toolset
Machine Learning is really good at average. Right now we are pumping a massive amount of info into programs to have them spit out things that are similar. That means it is taking the average of data and trying to make something. It also means these tools have yet to make something spectacularly new and unexpected. It can’t come up with the next best idea, right now it can just average out past ideas.
What this means as a music producer is we shouldn’t rely on these tools. But they are a great way to come up with unexpected ideas. After you get a bunch of ideas you can use your own tastes and decision making to narrow it down or edit them into new ideas.
Interpolate is particularly useful. I was able to take 2 melodies I wrote and use it to move between the two and come up with some new ideas. Taking the generated MIDI from Magenta, I edited it and made a whole new track. It’s a really unique way to augment your writing process.
I don’t know what the future holds for Machine Learning and Music Production, but this is the first step I have seen that really excites me as a musician.
Check out Magenta Studio and download to start playing: https://magenta.tensorflow.org/studio/ableton-live