In a recent piece I wrote about practical, everyday ways arts organisations can use AI to save time, things like meeting notes, biographies, funding applications, that sort of thing. This is the companion piece, and it looks at something rather different.
Because while a lot of the AI conversation in our industry is about admin and marketing, a quieter and arguably more interesting shift is happening at the heart of the art form itself: in how concerts are filmed and shared, in how musicians read their music, and in how a performance reaches someone who will never set foot in the hall.
I have been looking closely at three companies: OnstageAI, Notation AI and Enote. Alongside them, I read a thoughtful medici.tv article by Charlotte Gardner on AI’s impact on performance, recording and streaming. What emerges is a clear and rather hopeful theme: AI is starting to remove some of the cost and infrastructure barriers that have long kept high-quality classical music behind closed doors.
It is not all rosy, and I will come to the genuine concerns later. But let’s start with what is actually exciting.
Bringing the concert hall to a far wider audience
For decades, there has been a simple, frustrating truth about filming classical music: doing it well is expensive.
A properly produced concert film traditionally can often need multiple cameras, a team of operators, a director calling shots in real time, and hours of editing afterwards. That is fine for the Berlin Philharmonic. It can be completely out of reach for many regional orchestras, conservatoires, a chamber series in a parish church, or a youth ensemble. These are exactly the organisations whose work most deserves a wider audience, and who most need the reach to build one.
This is the gap that OnstageAI and Notation AI are both trying to close, and it is striking that two separate companies have arrived at such a similar idea.
OnstageAI describes itself as a technology that automates the video production of live classical music. The clever part is its “Sound-Understanding Technology”: rather than relying on a human director, the AI assistant actually listens to the performance in real time and cues the cameras based on where the musicians are in the music, what they call a “dynamic time-code”. It drives PTZ (pan-tilt-zoom) cameras automatically and is designed to plug into whatever video setup an organisation already has, “whether you have a fully equipped backstage video recording studio or a simple setup”. Crucially for smaller venues, it runs on an unlimited subscription model (you can film as many concerts as you like), which turns a series of large one-off production bills into a single predictable cost.
Notation AI approaches the same challenge from a similar angle: a combined hardware-and-software solution that automates concert video production for orchestras and concert halls. Its system reads, analyses and digitises the musical score, then directs cameras towards the right musicians at the right musical moments, automatically tracking and zooming in on individual players. The pitch is unambiguous: “what once took a full crew and hours of editing is now an efficient, streamlined process”, delivering something close to a director’s cut immediately and making professional-quality concert video “accessible and affordable for orchestras of all sizes”.
Now, I would add a note of realism here. An AI camera director will not, yet, match the instinct of a great human director who knows the score intimately and anticipates the oboe solo before it arrives. For flagship films and broadcast releases, the human craft still matters enormously.
But that is not really the point. The point is the concert that currently gets filmed by nobody at all. The student recital, the new-music premiere in a 200-seat room, the regional orchestra’s community programme. For those, the honest comparison is not “AI versus a world-class production team”. It is “a watchable, professional-looking film versus a single fixed phone on a tripod, or nothing”.
When you frame it that way, the accessibility argument becomes obvious. Automated video production lets organisations that never had the budget or the infrastructure:
- livestream concerts to audiences who can’t travel, can’t afford tickets, or have access needs that make attendance difficult
- build an archive of their work to support funding bids, marketing and education
- reach diaspora audiences, alumni and supporters anywhere in the world
- give artists usable footage without a five-figure production invoice
That is a real democratisation of something that used to be a luxury good.
Sheet music for the 21st century
The second barrier AI is chipping away at is one most audiences never think about, but every musician knows intimately: the state of the sheet music itself.
An enormous amount of the classical repertoire still circulates as ageing, poorly scanned PDFs and dog-eared photocopies: cramped page turns, faint engraving, fingerings and cuts pencilled in by a stranger decades ago. It works, but it is hardly “21st century”.
Enote is tackling this directly, describing itself as “sheet music for the 21st century” and rebuilding scores in a native digital format rather than just photographing the old paper. The difference that makes is significant. Because the music is genuinely digital rather than a flat image, Enote can offer:
- Instant transposition to a different key, a transformation for singers and accompanists
- Intelligent navigation that actually understands repeats, Dal Segno and Coda markings, so you can jump straight to a bar rather than hunting for it
- Automated highlighting and colouring rules that, as they put it, “save hours of repetitive work” marking up a part by hand
- Cross-referencing between individual instrumental parts and the full score
- Customisation of size, layout and style, plus annotation and on-demand printing
Their library already runs to thousands of works from the baroque to the late romantic era, with plans to expand into vocal, orchestral, stage and contemporary repertoire, and the platform has been featured by The Guardian, ClassicFM and ARTE.
Enote is rebuilding scores natively, but there is a parallel movement using AI to rescue the music that already exists in poor condition. A whole category of optical music recognition (OMR) tools, such as Newzik’s LiveScores, Soundslice and klang.io’s Scan2Notes, now use AI to turn a photo or a flat PDF into an interactive, playable, editable score: guided reading, playback, the ability to solo or mute individual parts, and export to notation software.
Put the two trends together and you get something genuinely useful for the whole ecosystem. The accessibility angle here is quieter than concert video, but it is real: interactive scores help students who can slow playback down and isolate a line; they help musicians with visual impairments who need to enlarge or re-colour a part; and they help under-resourced ensembles and education programmes that simply cannot afford pristine professionally engraved editions of everything they perform.
Behind the scenes: programming, librarians and logistics
The medici.tv article is a useful reality check, because it widens the lens beyond any single product and talks to working musicians and producers about where AI is genuinely helping, and where it worries them.
Several of its examples are about the unglamorous machinery that keeps an orchestra running. Vasily Petrenko, Music Director of the Royal Philharmonic Orchestra, points to AI search tools helping to discover repertoire and identify pieces that complement one another when building a programme. The same article notes AI helping orchestral librarians catalogue parts, and helping managers find stand-in musicians and rearrange logistics at speed, something the RPO leaned on when its January 2026 tour was disrupted.
On the marketing side, AI-driven audience data lets orchestras understand who their ticket-buyers actually are and recommend concerts of less familiar repertoire to the people most likely to enjoy them. For an organisation trying to broaden its audience beyond the same core programme, that targeting is exactly the kind of capability that used to require a budget most simply don’t have.
Breaking the language barrier
The single most striking accessibility statistic I came across is from medici.tv itself. The streaming platform expanded its subtitles from 3 languages to 28 using AI translation.
Their Chief Technical Officer, Charles Bourgeaux, is refreshingly honest about how that works in practice. It is not “press a button and publish”. AI generates an initial translation which native speakers then review across roughly “10 iterations”, what he calls “a magical ratio of AI to Human”. On documentaries, he reckons AI gets to around 95% accuracy, with humans refining the final 5%.
I think that “ratio of AI to Human” framing is the healthiest model in this whole article. The AI does the heavy lifting that makes a previously impossible task affordable; the human supplies the judgement and the polish that make it good. And the result is that a performance filmed in Paris can now be understood by an audience in dozens of languages, a scale of access that would have been financially unthinkable a few years ago. This framing is something I have found time and time again in my own work when using AI, it speeds up the initial phase considerably, but you should still take the wheel for the final leg of the journey.
In the recording studio
AI is creeping into recording and post-production too. The Grammy-winning engineer Mark Willsher describes using AI to scan a score and “come back with three specific things” in seconds rather than hours, catching missing details and speeding up error correction that used to be painstaking manual work.
But he also voices one of the article’s most important cautions: a warning against an “obsession with perfection” that strips out “things which are actually human and work”. Pre-recording individual notes and pitch-correction long predate AI; the risk is that making perfection cheap and easy tempts us to sand away the very imperfections that make a performance feel alive.
The honest concerns
It would be naïve, and frankly dishonest, to present all of this as an unalloyed good. The medici.tv contributors raise concerns that anyone in our industry should take seriously.
Copyright and monetisation. The cellist Anastasia Kobekina warns that AI-generated music mimicking real composers and performers is increasingly being created, uploaded and monetised on streaming platforms, while the copyright questions around it remain unresolved. This sits alongside the UK government’s own ongoing, and still contested, work on copyright and AI: a live issue, not a settled one.
The flattening of the art form. The pianist Lucas Debargue offers the sharpest warning. If classical music is reduced to pure technical perfection, he suggests, you could end up with “10,000 versions of the same Beethoven sonata”, and an art form that has made itself irrelevant. He frames it, rather beautifully, as “an ontological crisis to which the solutions can only be creative,” insisting that the magic of music is “about human beings communicating… about what it is to be alive”.
The attention economy. There is a concern that AI-curated playlists feed listeners “bite-size chunks” of the familiar, gradually eroding the curiosity and sustained attention that a Mahler symphony actually requires. This is not just a concern that started with AI however, it goes back to the mp3 and the start of the digital music revolution, but is certainly being heightened with the involvement of AI.
And a revealing paradox. In an RPO survey, 78% of respondents believed live performance was the area least affected by AI, yet many also admitted that, if AI cost them their own livelihoods, attending the performing arts is one of the first things they would cut. Audiences and artists are economically intertwined in ways that are easy to forget.
The reassuring counterpoint runs through the whole piece, though, and it is worth holding onto. As Kobekina puts it, “performing in front of people is something that can’t be faked.” Live performance, the irreplaceable thing that happens in a room between musicians and an audience, remains the part of our art form that AI cannot touch. The same can be said of live theatre. So despite what ABBA might tell you, real, live, spontaneous performance - happening only in that moment, with that specific group of people - can’t be electronically reproduced.
What this means for organisations
If there is a single thread connecting OnstageAI, Notation AI, Enote and the medici.tv experiment, it is this: AI is most valuable in classical music not when it tries to replace the artistry, but when it removes the cost and infrastructure barriers around it.
The camera crew you could never afford. The 28 languages you could never translate. The pristine score you could never license. The audience data you never had the tools to read. These are the things AI is making accessible to organisations that were previously priced out. And in doing so, it is helping a genuinely wider audience reach the music.
The right posture, I think, is neither breathless enthusiasm nor reflexive suspicion. It is medici.tv’s “ratio of AI to Human”: let the technology do the expensive, repetitive, infrastructure-heavy work, and keep human judgement, taste and live performance firmly at the centre. Used that way, AI is not a threat to classical music’s future. It is one of the better tools we have for widening its reach.
Need more hands-on help?
If you would like to talk through any of this for your own organisation, whether that is exploring automated concert video and livestreaming, building a digital archive of your performances, reaching new audiences online, or simply working out which of these tools is worth your time, then I would be very happy to help. At Knight Classical, we work across websites, film and video, digital media, social media, audience engagement and strategic consultancy, always with a focus on helping artists and arts organisations communicate their work with clarity and impact. If you would like to discuss a project or just get some advice, you can contact me directly at martin@knightclassical.com.
