FM PRO TECH 5
AI-assisted performance is moving beyond generated tracks and into the full architecture of the live show. The second Reply AI Music Contest drew more than 1,400 applications from 45 countries — a 55% increase on 2025 — with submissions spanning music production, reactive visuals, interactive systems and real-time performance.
For FM PRO TECH 5, Reply Chief Technology Officer Filippo Rizzante discusses what that growth reveals about artists’ changing relationship with AI, how live audiovisual workflows are evolving, and why questions around consent, licensing and creative control must remain part of the conversation.
He also considers AI as an “Alternative Intelligence”: not a replacement for the performer, but a system capable of extending what individual artists and small teams can create, control and improvise onstage.
Q1. Reply AI Music Contest received more than 1,400 applications from 45 countries this year. What does that increase tell you about how artists are now approaching AI as a creative tool?
It’s well-documented that AI music-making tools have made their way into the music industry zeitgeist – AI is now capable of helping more people make music, whether fully generative or more assistive.
Entries for this year’s edition were up 55% on 2025, and when you look at where these entries came from – 24% Italy, 18% Spain, 13% the UK, then France, Germany and the US not far behind – you get a palpable sense of how spread out this interest has become rather than it being concentrated in one or two scenes. The uptick in submissions aside, what I find interesting is the makeup of who applied, alongside the DJs and producers we’d tend to expect: we had visual artists and a number of people who described themselves as AI researchers submitting live performances, not to mention the range of uses of AI. That’s not something we saw much of as little as a year ago – and it suggests the boundary between someone who researches this technology and someone who performs with it is starting to blur, at least among the people drawn to a contest like this.
Q2. This year’s finalists work across sound, visuals, interaction and live performance. From your perspective, where is AI changing the live music experience most clearly?
I would point to how visuals are generated during the set rather than beforehand. A good number of entrants this year were working with tools like TouchDesigner and Resolume Arena, which are built specifically for making visuals react to audio as it’s happening on stage. Even without coding knowledge, you can start creating engaging audio-driven visuals, so what you see isn’t a video that was rendered last month and gets triggered at any given moment; it’s responding to the performance as it happens and reacting to specific frequencies. And when you put that next to Suno, which was far and away the tool most people used for the music itself, what you end up with is a set where both halves, the sound and the image, are being shaped live rather than prepared and played back.
Put together, what this year’s finalists demonstrate is a move toward performances where both audio and visuals are not fixed in advance, but continuously generated and shaped in response to what is happening on stage. That changes not just the aesthetics of live music, but the very nature of what it means to “perform”.
Q3. The wider music industry is still debating consent, licensing and control around AI. How do you separate responsible creative experimentation from the more problematic uses of the technology?
The ethics of AI should be front of mind for everyone building and using these technologies, and they’re not unique to music, these considerations are by-and-large industry agnostic. A competition like this isn’t in a position to resolve debates around consent, licensing or training data, and I’d rather acknowledge that than pretend otherwise. Responsibility here sits in two places at once: with the artist’s choices, and with the wider industry conversation about how these systems came to exist in the first place, both deserve to be named.
I don’t think there’s a clean separation to be made. The question of how an artist uses AI and the question of how the underlying systems were trained are connected, not independent. A competition like this can showcase creative practice, but it can’t adjudicate questions of consent or licensing upstream. Those are industry-wide issues that deserve continued scrutiny.
Q4. Agoria describes AI as “Alternative Intelligence” rather than artificial intelligence. Does that idea reflect how you see artists working with AI inside performance?
I think it does, and what’s interesting is how closely it lines up with how the artists sum up their own process, not just how Agoria describes his in isolation. A lot of what came through in submissions was this idea of AI as something that suggests a rhythm or a visual direction, and then the artist takes that and shapes it into something that actually carries emotional weight – nobody described handing the performance over to the tool. If anything, what people kept describing was reaching a scale they couldn’t have gotten to on their own: whole narrative worlds, characters and visuals that stay consistent across an entire set, the sort of thing that would normally need a much bigger and experienced team behind it.
In that sense, Agoria’s idea of “Alternative Intelligence” feels feels borne out in practice, extending what an individual or small group can hold and orchestrate in real time during performance.
Q5. Looking at this year’s submissions, what do you think AI-assisted performance could look like over the next few years. More automation, more collaboration, or something completely different?
AI and music have a much longer shared history than most people realise. It certainly didn’t start with us or even with this decade. Long before Auto-Tune and deepfake compositions, Alan Turing had a computer playing music in 1951, the same machine the BBC recorded for what’s thought to be the first captured computer performance. The pairing of machines and music has been around for as long as we’ve had machines capable of either. It’s more of a case that in the digital age, no such instances have been afforded as much attention.
The speed of throughput is really what struck me most in this year’s submissions. Artists are testing musical ideas and visual directions in hours rather than months, and that compression changes what becomes realistic to attempt for a single live set. If that trend continues, I’d expect the next few years to bring more work that simply wasn’t achievable by one or two people before: full audiovisual environments, sustained in real time, without the production team that scale used to require.
©fm