“For the third time, stop tapping!” my primary school teacher screams at me from across the room. I must not have heard her the first two times. I’d been drumming on the desk again, using my fingers for sticks and the floor beneath for a kick drum. While my body was in maths class, my mind was elsewhere.
It was 1970. I was John Bonham, drummer of legendary rock band Led Zeppelin, on stage at the Royal Albert Hall, performing “Moby Dick” — one of the most iconic drum solos of all time. The lights are low, the atmosphere electric, and I’m thundering along, each beat pulling the crowd deeper into my rhythmic spell.
These kinds of daydreams happened a lot. More than my teacher, and my parents, would have liked. But that didn’t stop me. Drumming was my creative outlet, an escape from the whirlwind of adolescence — and maths, of course.
Back then, the ultimate form of musical immersion was playing drums to my favourite tunes. For that, you had to get your hands on drumless tracks. This way you wouldn’t just play with your favourite drummer — you could become your favourite drummer.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!
But in the early 2000s, removing drums from a song was almost impossible. The only option was to get your hands on an original recording of the band playing the song without drums. There were a few of these tracks scattered across the web or recorded on CDs, but only for the most popular songs. This technological impasse forced me, and millions of others, into the role of backup drummer. If only there were a simple way to remove the drums from any song, I mused…
Fast forward to the present day and my musical dreams have become reality. There are now several apps that use AI to separate and remove “stems” — like bass, drums or vocals — from any song. One of them is Moises, founded by Brazilian web developer Geraldo Ramos.
Like me, Ramos is a drummer. Unlike me, he’s also a tech whizz.
“I’ve been involved with computers since very young, but I also play the drums,” Ramos tells TNW. “I always had these two tracks in my life: music as a hobby, and then tech as a career. With Moises, I bought the two together.”
Ramos first launched Moises using Spleeter, an open-source AI model created by the research team at French music streaming company Deezer. Spleeter was revolutionary for the time, but it was built for researchers, not musicians. Ramos took the model and used it to create an alpha version of the Moises app. Over 50,000 people signed up within the first week.
“I realised that this was just the tip of the iceberg — this new generation of tools will be able to change everything, how people create, consume, produce music,” says Ramos.
Moises says it now has 50 million registered users on its platform. The app is used by amateurs looking to practise their craft. It’s also endorsed by an ensemble of rising stars.
YouTube drummer Jorge Garrido, aka “El Estepario Siberiano”, says the tool is “a total game changer.”
“Now not only can I play any drum part over the songs that I cover but also I can learn any song by extracting the drums out of the original mix,” he tells TNW.
El Estepario, from Valencia, Spain, rose to fame through viral Instagram videos. The drummer, who has over 4.5 million subscribers on YouTube, is one of a cohort of young musicians using technology to perfect their art and reach wider audiences. Increasingly, that includes using artificial intelligence.
“Tools like AI are just making things easier,” he says. “You no longer require a PhD in mastering to be able to master nor do you need a PhD in audio engineering to separate the instruments on a song. Technology is the new democracy for artists.”
You judge the results in this clip of El Estepario in action:
How does AI separate drums from a song?
Moises’ developers train their machine learning algorithms on thousands of stems so that the AI can learn to recognise the unique frequencies and rhythms of each instrument. Over time, it gets better at identifying and separating these sounds from mixed audio, even when they overlap.
Once the AI isolates and removes an instrument, it fills in the space by reconstructing the remaining audio, smoothing over any gaps to make it sound seamless.
While Moises got its break with song separation, it has since developed a whole suite of AI tools aimed at helping musicians practise. One of these tools picks up the beat of any song and then adds a metronome to it. Another for guitarists can automatically detect the chords of any track.
Moises is also working on a generative AI toolset to launch later this year that can create an entirely original stem for you.
While Moises designed the first version of its app using Deezer’s Spleeter, it now has a team of data scientists building AI models in-house.
According to the company, all the algorithms are trained on licensed music from studio houses and compositions created by producers in Moises’ studios.
Ramos says the company is committed to “ethical AI.”
“Ninety percent of our team are musicians,” he says. “We’re not trying to replace real music but enhance it.”
The good and bad of AI for music
In recent years, AI has faced significant scrutiny in creative industries over concerns ranging from copyright infringement to job losses.
Last year, a band of US record labels sued Suno and Udio, two of the most prominent AI music generators, alleging copyright infringement on a “massive scale.”
Udio’s and Suno’s tools allow users to produce entire songs by typing in written descriptions. The companies claim their use of copyrighted material falls under “fair use,” a common defence from AI companies.
Aside from allegations that AI companies are ripping off original works, some worry that using algorithms to generate music risks replacing the vital human element that makes every piece of art unique.
“I’m fascinated and horrified in equal measure,” British new wave artist Gary Numan told Blitzed Magazine in an interview last month. “I fully expect Al to write great songs. There will be Al pop stars and actors who will become as popular, if not more so, than any human. We will go to shows where the stars are Al but appear on stage just the same. Everything is about to change.”
But Numan does believe that human creativity will endure. “I think for quite some time the world will be amazed and entertained by all the wonders Al will create in the arts. But, ultimately, if we survive long enough, I hope and suspect that people will slowly return to human-created art,” he said.
Others are less doomsday-ish.
“The phonograph, synthesizer, cassette tape, computer, and internet didn’t manage to kill the music industry as many feared, so there is no reason to start clutching our pearls now,” Austin Milne, a lecturer at the London College of Contemporary Music (LCCM), tells TNW.
LCCM is one of many music schools that have integrated AI into their teaching approach. However, Milne stresses that AI in music isn’t a monolith.
“There are some types which take the authorship and human touch out of the equation, and there are others that merely speed up processes musicians already undertake manually,” he says.
It’s an important distinction — like any powerful tool, it’s how AI is wielded that makes all the difference.
Whether AI popstars will usurp their human counterparts or not, I’m more excited about the potential of the technology to up my drumming game. So for now, thank you, machines, for allowing me to relive my musical fantasies.
Get the TNW newsletter
Get the most important tech news in your inbox each week.