Music education is deeply rooted in tradition: students often learn from teachers who, in turn, teach much as they were taught, sometimes using poor or outdated methods. UK startup MuseCool has found a way to break that cycle.
Since launching publicly in April, the startup has processed around 2,000 music lessons, using audio AI not just to transcribe what teachers and students say, but to understand the music being played — while building a dataset that could ultimately reveal something the sector has historically struggled to quantify: what actually happens inside a music lesson.
Check out our earlier interview with MuseCool CEO and co-founder Petru Cotarcea.
Bringing new technology to a traditional sector
According to CEO and co-founder Petru Cotarcea, the last six months have been an interesting case study in how the music education world adopts technology.
“Music education can be quite conservative. People go to conservatoires to study, and naturally a lot of what you learn relates to how music has historically been taught and played,” he shared.
Of the roughly 100 tutors approached, about 40 didn’t even try MuseCool. Some essentially said, “I have my methods.” Of the 60 who did try it, 28 became users. They brought their entire studios with them.
Cotarcea has learnt from the public beta that adoption in music education can be difficult, but it’s sticky.
“Once people become accustomed to a system, they’re likely to continue using it for quite a long time. That’s where many other apps in our field struggle.
You can get amateur musicians to try apps, but it’s much harder to get professional musicians and teachers to incorporate apps deeply into education.”
Teaching AI to understand a music lesson
The platform has now processed around 2,000 lessons. With MuseCool, a music lesson is essentially an open canvas: countless things can happen. Its “magic sauce” is understanding the lesson itself — almost like Shazam 2.0 — identifying the piece, where the student is in it, what’s being played, and what is happening around it.
At the core of MuseCool is its work in Music Information Retrieval (MIR), the technology that enables the platform to analyse and understand what is happening musically during a lesson.
The company works closely with Queen Mary University of London on MIR research and is preparing to begin an Innovate UK project with the university to further develop its engine. According to Cotarcea, this ability to analyse the music itself — rather than simply transcribing what is said during a lesson — is MuseCool’s key differentiator.
“We listen to the lesson itself and understand what is happening musically. Everything else comes from that.”
Cotarcea points to a well-funded US competitor that launched this summer. Initially, its VC backing made him nervous, but trying the product changed his view.
“They can record the speech, generate summaries, and isolate the music so you can play it back. But they don’t seem to analyse the music itself. For us, understanding music is fundamental.”
From lesson analysis to an all-in-one teaching platform
And this is just the beginning. According to Cotarcea, sheet music is a major issue for many people. Thankfully, much of the music used in classical music education is in the public domain, so MuseCool hired a team over the summer, mostly in Ukraine, to expand its collection of public-domain music and convert it into XML.
The app now has around 4,000 piano pieces, and the collection is growing rapidly. At the moment, it’s almost doubling every month because of software and progress improvements. This means a parent doesn’t have to buy a Mozart book, for example, because the app offers public-domain sheet music.
By the end of the year, the company also hopes to offer licensed music alongside the public-domain catalogue.
Further, as MuseCool has refined its lesson analysis, its summaries have become more detailed. Heat maps let users click individual bars of music to see what the teacher and student discussed and what was happening musically. Winning over the next generation of music teachers, MuseCool aims to solve multiple problems for tutors: “We discovered that tutors can find it surprisingly awkward to talk about money. Obviously, it’s incredibly important that people actually get paid, so we wanted to make that process easier.
“It takes around five minutes to sign up, and then you can manage your lesson payments through the platform,“ shared Cotarcea. It also provides lesson summaries, helps get children practising, and provides access to sheet music. It can also potentially bring new students to tutors.
MuseCool operates a B2B2C model, with tutors providing the route to paying parents and students. According to Cotarcea, tutors on the platform have an average of 14.8 private students, meaning a relatively small number of tutors can reach hundreds of potential subscribers.
The startup is currently working with institutions and student communities around the Royal College of Music, the Royal Academy of Music, Trinity Laban, the Royal Conservatoire of Scotland, the Royal Welsh College of Music & Drama, Leeds, and others.
“We’re particularly interested in reaching conservatoire students because they’re generally between 18 and 25, and many of them already teach. They’re also younger and more open to technology because they grew up with it.
We have the potential to expose MuseCool to around 6,000 musicians across these conservatoires.”
The company has also automated a local marketplace model. Once MuseCool reaches a certain number of participating tutors in a location, it can generate a local website where opted-in tutors can be discovered by parents looking for music lessons. The longer-term aim is to replicate these marketplaces across cities and countries, including in local languages.
Turning thousands of music lessons into data
Through its work, MuseCool is accumulating one potentially valuable commodity: data about what actually happens inside music lessons. Already, it is producing some surprising insights. MuseCool’s data suggests that only around 40 per cent of the average lesson is spent playing the instrument. The rest is devoted to explanations, discussing the music, and general conversation, particularly in lessons with children.
Cotarcea says this challenges the assumption that most of a paid music lesson is spent playing:
“There is much more happening around the playing.”
But the potential value of that data goes far beyond understanding how lesson time is spent.
“Imagine tutors using MuseCool inside a conservatoire, music school, or another formal educational institution,” said Cotarcea.
Management, administrators, or programme heads could have a real-time understanding of how students are performing, how they’re progressing, and whether lessons are happening.
“You don’t need somebody to listen to every lesson. You have data points showing what is happening across 600 or 800 students.
If something starts going wrong, the institution can potentially intervene much earlier — before a student gives up or reaches a crisis point. That could help educational institutions keep students learning for longer.”
This means that instead of waiting until end-of-year examinations to identify problems, schools could have a real-time picture of what is happening across hundreds of students.
Then there’s academic research into the science of teaching.
“I dream of doing so many papers with this data,” said Cotarcea.
“I go to a lot of teaching events, and so much of what I hear is anecdotal. People tell stories about how they taught something, what they tried, and whether they thought it was successful.
There isn’t much genuine quantitative data showing what happens inside lessons and what actually works. With enough data, that becomes possible."
But the potential value of that data goes far beyond understanding how lesson time is spent. Another potential customer is instrument manufacturers. If you can understand how an instrument is actually being played, that gives manufacturers knowledge they currently don’t necessarily have.
Then there are music companies.
“If you have a genuine understanding of what musicians, students, or children in a particular area are learning, you can understand what kinds of music, composers, or artists resonate with them,” shared Cotarcea.
From lesson assistant to a talking piano
For Cotarcea, this is just the beginning. He imagines a talking piano à la Beauty and the Beast:
“Now imagine that applied to a piano — a talking piano. Maybe that’s five or ten years away, or perhaps much sooner.”
Technology already exists to understand how someone physically presses piano keys—how you touch and interact with the instrument — as well as technology that can understand speech.
“What we bring is the layer that understands the music in the room and analyses the audio.
Imagine rehearsing with your friends or your band while the piano listens to the room. It understands what is happening, who is playing what, who did what, and who said what.
We could collaborate with instrument manufacturers and bring the software directly into the piano. Then the piano effectively comes alive. It understands what is happening in the room, alongside technologies that understand how you physically interact with the instrument and what you say. That’s the vision.”
From a musical bus to the next funding round
MuseCool’s ambitions extend beyond its technology. Its documentary, The Most Beautiful Noise, will be released on PBS on the US West Coast in two weeks.
The film follows the transformation of a retired double-decker bus into a mobile music school, born from the ambition to make music education free and accessible to every child in London.
The company is also working to close its next funding round as it looks to expand the platform.
“There are still plenty of hurdles. Music education isn't the sexiest area for investors, so it can be an uphill battle. But we’re managing.”
From putting a music school on a London bus to building AI that could one day make a piano understand the room around it, Cotarcea’s underlying ambition remains the same: to rethink how music is taught and make music education more accessible.
Would you like to write the first comment?
Login to post comments