AI in the Music Industry
Data is a massive problem in the music industry.
Issues surrounding data are causing half a billion dollars in royalties to go unallocated. Streaming has really strengthened its position as the centerpiece of revenue in the industry, and that happened faster than the development of tools needed to really understand all the data surrounding it. Artists need to be present on many different social platforms, all of them have different data that needs to be kept track of. Marketing budgets are increasing but capabilities around optimizing those budgets and actually attributing consumption/revenue growth, and therefore ROI, to specific campaigns are not increasing at the same rate.
I believe the best way to solve some of the problems that have stemmed from these large and disparate datasets is the same way most industries are solving problems surrounding large and disparate datasets - with machine learning and big data analytics.
This is starting to happen - there are some cool companies getting a handle on these pain points with AI and it’s even making an impact in some really cool ways completely outside of these pain points. As a fan, I want to highlight some of the most significant ways I think AI/ML is changing the game in music and some amazing tools out there*
Music recommendation engines
I think this is where we may have seen the first true and thorough implementations of deep learning in the music industry, and this technology has gotten pretty amazing. Discovering music has never been easier. What platforms like Spotify and Pandora have done when it comes to tagging music, analyzing listener behavior and then recommending songs is incredible. I check my Discover Weekly every Monday.
This is probably the most rapidly expanding and saturated (I don’t mean that negatively) uses of AI in the music industry. The amount of tools creators now have that utilize AI to help them make music, as well as systems that create music all by themselves, is pretty insane. These are some of my favorites:
AIVA - beautiful pieces of music, composed entirely by AI
LANDR - music mastering powered by AI
AI Duet - A Google AI/Magenta experiment where you can trade riffs with a trained computer
Amper, Loudly, Boomy, beatoven.ai - compose music by deciding on things like tempo, key, and certain layers, let AI do the rest
Latent Loops - a Google AI experiment, use AI to blend different musical attributes
Not specific to music creation but definitely worth mentioning in this category, Respeecher - make your voice sound entirely like someone else's.
This might be the area that has advanced most rapidly in recent years. There is certainly still room for growth and standardization across the industry, but by building on the sound recognition technology that created the foundation for music recommendation engines, leaders like Musiio, Cyanite, and Musicube have made headway in this area. We can now have machines that can listen to music and tag it in seconds, labeling everything from tempo, to key, to genre and mood.
This area seems to have captured the interest of the majors the most, which is really no surprise, and one where we might see some of the most new advancements soon. Warner acquired Sodatone and the others have been making similar moves. It’s worth mentioning Musiio again as they have also made some really impressive moves in this space. These tools can look at social and streaming trends across millions of artists and help predict which ones will be the next big thing.
I believe there is still a lack of intelligent tools addressing social media, streaming, and marketing, which is why this is the exact dataset that we're tackling at Immensity. Check out more here.
Before parting ways, a few predictions on how AI will continue to reshape the music industry -
Music recommendations will continue to get more targeted
In the health/fitness industry AI is already being used to measure heart rates and stress levels, it’s not far-fetched that pretty soon we might be getting music recommendations based on our exact mood.
Music making will become more democratized
I want to go on record and say that I don’t think AI will ever replace musicians and songwriters entirely, and that humans will already play an essential role in hit-making. Humans are fueled by creative drive and curiosity and there will always be someone out there who wants to pick up an instrument and write. Not to mention the music instrument market is projected to hit $15.2B by 2028. However, as we’re seeing in other industries where AI is taking over repetitive sales or customer support tasks, it will takeover the creation of more repetitive music like background music. We will also see tools that allow even non-musicians to create very high quality music, and potentially even perform that music in front of huge crowds, leveraging technology we’re seeing with acts like Hatsune Miku.
Referencing back to Google AI Experiments I brought up earlier, there are two other experiments on there: “FREDDIEMETER” which is code that will record you singing a Queen song and give you an AI-generated score on how close you were to sounding like Freddie Mercury, and “SEMI-CONDUCTOR” which uses AI to track your body movements and conduct a virtual orchestra. Music schools and programs will likely start to use technology of this sort to augment teachings in their classrooms.
Automated marketing campaigns
The market for AI-generated creative is growing fast. Between this technology, and technology like Immensity’s, we’re not too far away from the most optimized marketing campaigns being created entirely with machine learning.