Who's Really Remixing Who? The AI Remix Debate Hitting Underground Electronic Music Hard

Not long ago, a remix meant something specific: a human producer, a set of stems, a creative decision to take someone else's raw material somewhere new. The original artist usually got paid, or at least credited. The remixer brought their own identity to the work. There was a transaction — creative, financial, cultural — that made sense to everyone involved.

Now, an algorithm can ingest a track, deconstruct it, and generate something new in minutes. No stems required. No licensing conversation. No credit line on the release. And depending on how you feel about that, it's either the most exciting development in electronic music production since the DAW, or a slow-moving disaster for the people who actually make the music.

Probably some of both.

What These Tools Are Actually Doing

It's worth getting specific, because "AI remix" means different things in different contexts, and the distinctions matter.

On one end, you have tools like Suno, Udio, and various stems-separation platforms that can analyze a piece of music and reconstruct or reinterpret it based on user prompts. Feed in a moody ambient track and ask for a four-on-the-floor club version — some of these systems will give you something surprisingly listenable. They're not just pitch-shifting or time-stretching; they're actually generating new audio informed by the source material's characteristics.

On the other end, you have production assistants — tools embedded in DAWs or available as plugins — that use machine learning to suggest variations, generate complementary patterns, or even build out full arrangements from a seed idea. These feel less controversial because the human is still clearly driving. The AI is more like an unusually fast collaborator than a replacement.

In between those poles is where things get genuinely murky. There are systems that can train on a specific artist's catalog and generate new material in that artist's style without their knowledge or consent. That's the version that's causing the most friction, and for good reason.

The Experimental Community's Complicated Relationship With This

Here's something interesting: the experimental electronic world has historically been more open to algorithmic and generative music than almost any other genre. Artists like Aphex Twin, Autechre, and Holly Herndon have all engaged seriously with machine-generated sound. Generative composition — using code, randomness, and systems to produce music — has been a legitimate creative practice in this space for decades.

So it's not that experimental producers are anti-algorithm. Many of them have been building with algorithms for years.

The tension is more specific than that. It's about agency and compensation. When an artist chooses to use generative tools as part of their creative process, they're making an intentional decision about their own work. When an external system trains on their recordings without permission and produces derivative material — potentially competing with them commercially — that's a fundamentally different situation.

Independent labels are navigating this in real time, and their approaches vary wildly. Some small electronic imprints have started adding explicit language to their artist contracts about AI training, prohibiting the use of their releases as training data. Others have gone the opposite direction, experimenting with releasing AI-generated companion pieces alongside traditional albums, treating the technology as a creative extension of their roster's work.

Neither approach is obviously right. Both are responses to genuine uncertainty.

The Compensation Problem Nobody Has Solved

Let's be direct about the economics here, because they're pretty rough for working artists.

Traditional remixes — even in the underground — typically involve some form of compensation, whether that's a flat fee, a royalty split, or at minimum a formal licensing agreement. That structure exists because both parties bring something to the exchange and both deserve recognition for it.

AI-generated remixes, in their current form, mostly skip that entirely. The models were trained on music that was scraped from the internet, often without the knowledge of the artists whose work fed the training data. The outputs don't trigger royalty payments. The original creators see nothing.

For major-label artists with lawyers and leverage, this is a fight being waged in courtrooms. Several high-profile lawsuits are working through the system right now, challenging the legality of training AI on copyrighted material without licensing. The outcomes of those cases will matter enormously.

For independent artists on small electronic labels — people releasing 500-copy vinyl runs and making rent from a combination of streaming pennies and live gigs — the courtroom fight is largely inaccessible. They don't have the resources to pursue litigation, and by the time any legal precedent trickles down to their level, the technology will have moved on several more times.

That asymmetry is real and it's worth naming plainly.

Where the Creative Upside Actually Lives

None of this means AI production tools are without genuine creative value. They're not, and pretending otherwise would be intellectually dishonest.

For experimental producers specifically, some of these tools open up genuinely interesting territory. The ability to rapidly generate textural variations, explore unfamiliar harmonic spaces, or push a source recording into completely unexpected sonic territory — that's useful. It can break creative blocks. It can surface ideas a human working alone might never reach.

Artists like Holly Herndon have been thoughtful about this, building tools that specifically train on their own voice and creative output — maintaining authorship over the model itself. That's a model (no pun intended) worth paying attention to: using the technology as an extension of personal creative identity rather than a bypass around it.

There's also a real argument that AI remix tools democratize access to production techniques that previously required expensive equipment or deep technical knowledge. For a producer working with limited resources, that access can be genuinely liberating.

The question isn't whether these tools have creative value. They clearly do. The question is whether the current ecosystem distributes that value fairly — and right now, the honest answer is no.

Where Independent Labels Go From Here

The underground electronic community has always been good at building its own infrastructure when the mainstream fails to serve it. That instinct is going to matter here.

Labels that get ahead of this — establishing clear policies, advocating for their artists in conversations about AI licensing frameworks, and experimenting with the technology on their own terms — will be in a better position than those who ignore it until it becomes unavoidable.

And for producers: learn the tools, understand what they're doing, decide deliberately how you want to engage with them. The worst outcome is having this technology happen to you rather than making conscious choices about whether and how it fits into your practice.

The remix has always been about transformation and dialogue. The question now is whether AI changes who gets to be part of that conversation — and who still gets heard when it's over.