Why music’s AI opportunity is more Splice-shaped than it is Spotify-shaped Spotify and UMG’s AI ‘superfan’ announcement has grabbed the headlines, and rightly so. While the AI debate has focused on full track creation, the Goldilocks zone is audio-modification. Much of the opportunity for rightsholders may sit in DSPs but the format will be more Splice-like than Spotify-like 🎹 Most consumers will want to spend most of their time listening, but a growing share of consumers want to spend some of their time leaning in and doing more 🚀 This isn’t rocket science. It’s what Instagram and smartphones did for photography. 🎸 ‘Hang on, everyone takes photos while only a minority make music’ I hear you say. Well, in actual fact half of all consumers do something creative with music, whether that be playing an instrument, singing or making music with software. Music creativity is already mainstream, even before the advent of mass-market tools 📈 Full track creation is a big opportunity (gen AI MAUs tripled between 2022 and 2025) but fan creation is an even bigger one. Expressing yourself via your own take on your favourite artists’ music is the ultimate expression of fandom 💪 The grey zone between artist and fan is going to become the attention battleground between listening and creation. This is why Spotify is going all in – they need to act before the likes of Suno and Udio steal the attention, time and spend of significant portions of its user base 🎧 This is why the largest scale AI opportunity for rightsholders is more about seeding fan-modification tools with artist content in a Splice-like way rather than a pure Spotify-like licensing one 🧑🎤Expect this to trigger a whole new economy of artist sonic-likeness products. It won’t be long before we see artists in the studio with prompt engineers building prompt packs for consumer AI tools 🦢 AI is a black swan event for music, intertwining the futures of the creator economy and the music industry 🔋AI has quickly become the music creator economy’s growth driver and is set to become a fandom driver. Over time, it could become the defining factor of both For those who want to dive deeper (a lot deeper) on this, look out for the forthcoming MIDiA report The future of the music creator economy: Tipping point. The report features market sizing and forecasts for gen AI, audio modification, traditional creator tools and much, much more.
The moment this is described as an economy of “artist sonic-likeness products,” the central question is no longer whether fan modification will happen but what rights framework governs CREATIVE PATTERNS being monetised? Masters and compositions protect specific works. Voice and likeness protections address identifiable replication. But prompt packs, artist-seeded modification tools and sonic-likeness products are commercialising something broader: the recognisable creative pattern of an artist. Who consents to that use? The label controlling masters? The publisher? The artist? Who approves the outputs, the contexts, the derivative identities and the posthumous uses? How is value attributed back to the actual creative source? Fan participation can be legitimate and exciting. But without rights architecture around source, consent, boundaries, attribution and revenue participation, this becomes a new derivative market built from artists’ identities while calling itself fandom...
We’re running into a dire supply/demand problem though. When everyone is an artist there are no fans left and the ecosystem dries up. This has already become an issue in DJing, with the bar of entry becoming so affordable and technical barriers reduced in the last 10-15 years. AI generated music is going down the same path in production, with everyone listening to their own creations and spending significantly less time listening to music made by musicians.
First time that I fundamentally disagree with Mark‘s prediction. I don’t see generative Music AI in the B2C/fan context becoming a sustainable massmarket usecase. It is somehow funny in the beginning, you show your „work“ to some friends, then you lose interest. However, some users/fans might stick to the usecase and lean in as predicted, but the vast majority won‘t.
Spot on about the opportunity being more Splice-shaped than Spotify-shaped. Consumers want to touch the actual audio assets, not just let a streaming algorithm passive-aggressively remix a track for them. That being said, if a project ever requires me to sit in a room with a "prompt engineer," I think I'll spend all day on a Pure Data patch instead.
What in the hell is a prompt pack!?
Cherie Hu's recent Udio / Starstruck was interesting in this space too. There's certainly something here that's reminiscent of the 'everyday creativity' vibe of the late 00s (i.e., the first wave of garageband / youtube homebrew activity, with no real focus on monetisation) - where lean-in has that gamification / do it for fun / share it with your friends appeal. I guess at the structural layer, the challenge is around interoperability (which is Cherie's point about walled gardens) - i.e., if the homebrew AI-powered deepfake remixes are only available within Starstruck / Spotify etc., then beyond the licensing deals with labels and already-existing platform subscription, then routes to broader monetisation of user-generated content isn't obvious. Maybe prompt packs will be the new sample packs...? But outside of the realm of specialists, sample packs have never been a huge market in themselves.I think you're right though, Mark - rightsholders finding a way to monetise this emerging and evolving zone of licensed modification of copyrighted content does feel like it's going to be a focus for the near future.