Artificial Intelligence and the Future of Music Search for DJs

DJs being replaced by robots is a tale I remember being told to me when I first started DJing… and that was 20 years ago. As a 14 year old, I’d stay up late to watch clips of DJs like Carl Cox on the Space Terrace on Ibiza Uncovered, and try to emulate them rather unsuccessfully in my bedroom, fumbling mix after mix from one vinyl to another on my Gemini XL-400 belt drive turntables. Learning to DJ was hard, but worn belt drives made it even harder. I often wondered if this was time I was squandering… not least because technology was improving much faster than my abilities.

At the turn of the millennium, music was going digital. iPods were promising 1,000 tracks in your pocket and the CDJ-1000 was setting the new industry benchmark for a digital turntable; while many had tried in the past, Pioneer had created a deck with the same touch, feel and reaction as vinyl, but in a digital package which allowed more creative manipulation of the music. Most importantly though, the cost of buying music was no longer an issue. The CDJ was a giant leap for DJs and what many had been waiting for. Most vinyl bags could carry around 100 vinyls which were not only big and bulky to carry around, but it was also an expensive hobby that took up a lot of space. By switching to “the dark side”, you could carry a few hundred CDs with upto 15 tracks on each so instantly this offered DJs many more options when playing out.

Fast forward to 2020 and it’s DJ controllers that have flooded the market. Hardware devices from entry level to professional grade that allow users to plug and play and manipulate their digital catalogues or pay monthly subscriptions without having to purchase any music, but have instant access to millions of tracks. I often think about how difficult it was to learn without all this technology; how I’d spend most of my weekend travelling miles to vinyl shops which had less than 50 vinyls in the soulful house section. For the first few months I had 6 records, all of which are still burnt into the back of my mind. At £5 each plus delivery, I couldn’t afford any more than that. But today, I have access to millions of tracks at my fingertips, no longer bound by what’s in my local record store; I can define my sound from my bedroom.

Today’s technology means that I can reorder my catalogue of 10,000+ tracks instantly in order of BPM or genre. I can keyword search tracks or type the artist name. I can set up cue points ahead of time. It seems like technology that has been here forever but these have all been developments in the last 20 years. This technology has shaped the way we think about music and the way that we DJ. Being able to take 10 thousand tracks into a club would have sounded so out of this world 20 years ago, yet here we are.

So where does AI come in? Well, I’m here first to tell you that AI will not replace DJs. And here are 3 simple reasons why;

1. DJing is fun -  Most automation technology that exists and becomes popular, is technology which replaces boring and mundane tasks that people don’t want to do, or it’s humanly impossible to do. I haven’t done a count recently, but there is no shortage of DJs. If anything, DJing is more fun and more accessible than ever. 
2. Watching DJs is fun - To assume that just because AI could put together a mix of tracks that were of the standard of Seth Troxler that people would want to see it, is a misconception. People want to pay to see Seth Troxler because they love him, his brand and who else he attracts. They love the music he plays but also the surprises that he brings with him to every performance. To have him replaced by an all-seeing, all-hearing, all-sensing AI, is just too big brother and I’m sure there aren’t many people that would want to be a part of it, beyond a one-off concept in a pod on the London Eye. 
3. Time and investment - You could eventually be able to train an AI to react to the amount of people on the dance floor - to use facial recognition to see how they are reacting to the music, to make tweaks to the music that draws them closer to the DJ booth, to adapt according to the sound system and the size of the club, to play different music if it’s a basement, beach bar or warehouse, to play differently if the walls or concrete or have drapes, and eventually a hive of clubs around the world that are interconnected and understand how to get the crowd to stay all night. But for the time it would take to train an AI to do this and the investment it would take, is currently insurmountable… though technically not impossible. 

So where is AI and how will it change DJing?

Well, with the massive influx of music in the last 20 years, it has completely changed the way in which DJs dig for music. Gone are the days when you would listen to all 50 soulful house records in an afternoon and feel satisfied that you’d listened to everything. We are now talking about millions of tracks and it’s now impossible to listen to all of it. You can set yourself up for success by following your favourite artist and labels on your favourite online music stores, but there’s nothing that really can replace just sitting and listening through thousands of tracks. But what if I told you that you no longer have to listen to music that you don’t want to? And that AI will help you get to the good stuff faster than ever before? Let me show you how...

Music Tagging

At Musiio, we’ve trained an AI to “listen” to music. Similar to Shazam, it creates a fingerprint of each audio file, but instead of finding the name of the track you are looking for, we create tags. Here’s a few examples…

We simply dropped YouTube links of these tracks into the live demo on musiio.com and we were able to generate all of this information from the audio file itself. Not only are we able to tag its genre, but we are also able to tag the mood, energy, and instrumentation of the music, and not only its vocal presence, but whether it’s male or female. To have this information while you are DJing is very useful, but even better information to have when you are searching for music for your next set.

Music Search

So cut back to 18 months ago (I’m sure we’d all be on board with that!) when I was DJing most weekends: I would get home from work and spend anywhere between 20 and 40 hours per week listening and preparing music for my sets at the weekend. I’d usually play a 1 to 2 hour set, but sometimes had to prepare for more if I was playing multiple sets in a weekend. I love playing music. And one of my favourite things to do is to find new music. The feeling you get when you know everyone is hearing the track you are playing for the first time is hard to match. But if there is one frustration I have, it’s that I can’t seem to listen to enough music. I found a formula that worked for me while searching through catalogues online. If I listened to roughly 1000 tracks, I would drop around 100 to a cart, and then edit down to around 15-20 for 1 hour of music. I’d always end up with a good quality hour this way, but the thing that struck me, was that 98.5% of the music I listened to, was music I didn’t like or would never play. 

https://soundcloud.com/its_justmack/live-from-the-fort-iii

An hour set like this would usually take me 20 to 40 hours per week to listen and prepare.

Many would consider this just part of the job and I did too, but as Henry Ford famously said “If I had asked people what they wanted, they would have said faster horses”. The reason I joined Musiio, was when my CEO, Hazel Savage told me what it can do; I understood right away what this would do for DJs. If a music store is powered by Musiio’s AI technology, the user is able to paste in a music link from YouTube and our AI scans the audio file for 1500 datapoints per track, compares it to every track in the database and find the closest audio matches in seconds. The user is then able to filter results based on genre, mood, bpm, vocal presence, instrumentation and more. 

So, you hear a track being played in a bar on a night out and you love the vibe. It’s a little too mainstream to play in your sets, but you like all the elements. Currently, there’s not a great deal you can do with this information. With our tech, you are able to take the reference track, paste it into a music database and return the closest audio matches. Perhaps you like the results, but you didn’t want vocals - you can simply remove all results with vocals. Perhaps you wanted a similar track, but with a different energy or mood - you can filter the search results to suit.

Next to every track, there will be a Find Similar button, so you can reorder the catalogue in order of audio similarity in seconds. But doesn’t this just make DJing too easy? Well, if you are just creating playlists maybe. But if you think about the possibilities and options this gives you to actively search, build sets your way and define your sound, then being able to search for the precise music you are looking for using AI music search gives you more creative freedom than ever before.

The Future of Music Search for DJs

The problem with music search for DJs currently, is that we are all shopping from the same stores and being guided down the same isles. We start on the same release page, go to the same genre pages and move to the same DJ Charts each week. Unfortunately this causes a little bit of an echo chamber effect. AI will change all of this and put the power of music search back in the hands of DJs. You’ll be able to enter down the rabbit hole of any catalogue from your first interaction. You’ll be able to start with a concept; Maybe it’s an old Plastikman track, or a track you heard the Martinez Brothers play in the summer. You’ll be able to paste it in and find the closest audio matches. But then you’ll be able to change the mood, increase the energy and add or remove vocals; You’ll be able to start with one mood and move towards another. You’ll be able to design and build your set actively, adding tracks with certain instruments, or searching for tracks with different shapes; Maybe you need a track with several peaks and troughs, or you need a consistent rolling baseline. If the human ear can hear it, we can teach our AI to tag it. And when you consider some of the largest catalogues have close to 10 million tracks in their catalogues, the rabbit hole is deep and the possibilities are vast. 

Our AI is getting smarter, but don’t worry, there’s no plan for it to take your next gig. But expect to see some of the laborious processes you're used to disappearing and who knows, maybe music search will be fun again. 

Try our live demo live at musiio.com