An AI music remixer can turn a demo into a different genre in minutes. Sheffield’s electronic history offers a better question than whether the machine is creative: what will people do with it?
The first record Warp ever released was made in about four hours, one evening, round at a bloke’s house.
It was 1989. Robert Gordon, Winston Hazel and Sean Maher recorded a track with no vocal, no chorus and, fittingly, no name. They pressed 500 white labels, drove them round the country to record shops and sold the lot inside a week.
They called themselves Forgemasters, after the steelworks.
That may be the most literal steel-to-techno connection in British music. It is a band name, not a sound. Decades of music writing have been trying to turn it into a theory about the whole city ever since.
The argument Sheffield never settled
You’ll know the story. Sheffield made electronic music because Sheffield made steel. Drop forges gave the city its rhythm, factories gave it noise, and the musicians who grew up hearing both rebuilt those sounds with synthesisers.
The people who were there do not entirely agree.
Martyn Ware, who co-founded The Human League and then Heaven 17, has described the muffled booms of drop forges drifting through summer nights like a heartbeat. For him, the industrial inheritance was real and audible.
Richard H. Kirk of Cabaret Voltaire resisted the neat version. The works could be heard drifting up the valley at night, but the band did not sit down and decide to sound like a factory. Much of that story was noticed afterwards.
There is a third explanation, less romantic and probably more useful. Steel shaped Sheffield’s music economically. When industry contracted, workshops emptied. Empty buildings were cheap enough to become studios, rehearsal rooms and venues. Cabaret Voltaire’s Western Works occupied part of a former cutlery works. Yellow Arch in Neepsend continues the pattern in an old industrial building.
The steel did not hand Sheffield a sound. It handed musicians space.
That history matters because the same mistake keeps returning whenever a new piece of music technology appears. We give the machine too much credit and ignore the conditions around it: who can afford it, where they can use it, who teaches them and what kind of scene decides the result is worth hearing.
Cheap gear, used the wrong way
Sheffield’s electronic tradition was never about owning the most advanced equipment. It was about making limited equipment do something it had not been designed to do.
The Human League’s “Being Boiled” was recorded in mono on a domestic tape machine and reportedly cost £2.50 to make. Cabaret Voltaire treated tape loops, primitive electronics and film equipment as instruments. Forgemasters turned a short home session into Warp catalogue number one.
As Exposed’s own interview with Winston Hazel makes clear, the point was never technology for its own sake. It was a city building nights, labels and communities around adventurous records.
Bleep was a Yorkshire sound before it was only a Sheffield story. LFO came from Leeds; Warp released them. Bassline, meanwhile, became inseparable from Sheffield club culture. Niche was shut in 2005 after a police operation involving hundreds of officers, yet two years later T2 and Jodie Aysha’s “Heartbroken” spent three weeks at number two.
The scene outlived the attempt to contain it. Exposed’s recent conversation with the Queens of Bassline found the same music still crossing generations and filling festival stages.
Tools change. Scenes decide what lasts.
What an AI music remixer actually does
The current sales pitch can make AI remixing sound like a magic DJ: upload any song, name a genre and receive a finished club weapon.
The useful reality is narrower.
Some tools begin with source separation, pulling a stereo recording into vocals, drums, bass and other parts. That is helpful for detailed editing, but it is not the same as generative remixing. A generative system analyses the source and creates a new performance or arrangement guided by a description.
With AI music remixer, for example, a musician can upload a track, describe a new genre, mood or instrumental direction, and compare two new versions. It is closer to asking a very fast session band for alternate takes than opening a traditional mixer.
That distinction sets the right expectations. This kind of tool is useful for hearing a demo as lo-fi, drum and bass or something cinematic before rebuilding it properly. It can suggest a direction, produce an alternate version for a video, or shake a writer out of an arrangement they have heard too many times.
It is less useful for surgical work. If the job is to nudge one hi-hat, automate a filter or control the exact transition into a drop, a DAW still offers far more precision. An AI remix can be a sketch, a surprise or sometimes a usable result. It is not a replacement for knowing what every element of a mix is doing.
The most productive workflow is therefore simple:
- Start with a track you made or have permission to use.
- Ask for a clearly different direction rather than “make it better”.
- Compare more than one result and listen for an idea, not just polish.
- Take the useful version back into a production workflow if it needs detailed control.
- Keep notes about the source, permissions and tools used.
The last point is less exciting than a genre flip. It is also the one most likely to prevent trouble.
The bit the adverts skip
Uploading a commercial release to a remix tool does not become lawful because a model, rather than a person, changes it.
There are normally at least two sets of rights in a released track: the composition and the sound recording. A remix may require permission covering both. Platform terms also place responsibility on users to own or obtain the rights to what they upload.
That makes original demos, self-released tracks and properly cleared recordings the sensible material for generative remixing. Anything else carries the same basic problem that has followed bootlegs, mashups and uncleared samples for decades, only with faster production and a larger digital trail.
Sheffield dance music knows that history. White labels and CD-Rs could move quickly through a local scene because they were physical, temporary and often below the industry’s attention. Online tools change the scale, not the underlying rights.
There is a creative reason to begin with your own work too. Turning a record everybody already knows into drum and bass is a novelty. Hearing your own unfinished chorus survive a completely different arrangement can teach you something about the song.
Britain has not finished the argument
In February 2025, more than a thousand musicians backed Is This What We Want?, an album of recordings from empty studios and performance spaces. It protested a proposed copyright exception that would have allowed AI companies to train on protected work unless rights holders opted out.
The government received 11,520 responses to its consultation. Its March 2026 report said the opt-out exception was no longer its preferred approach and that it would not change copyright law until it had stronger evidence and a workable consensus. It also proposed more work on training-data transparency and labelling.
So the UK has not solved the issue. It has paused one answer.
Meanwhile, the volume of generated music is no longer theoretical. Deezer said in July 2026 that it was receiving about 90,000 fully AI-generated tracks each day, more than half of daily deliveries at peak. Those tracks accounted for only 1 to 3 per cent of listening, and Deezer identified up to 85 per cent of their 2025 streams as fraudulent.
Enormous supply. Very little genuine demand.
That is the part of the AI debate Sheffield may understand best. Making sound is not the same as making culture.
The Sheffield answer
The city has never answered the machine question by banning the machine or worshipping it. It hands the thing to somebody who does not respect the instructions.
Rian Treanor runs an electronic music club for young people in Rotherham, bringing visiting artists into a space where production and DJing are learned by doing. His work with Mark Fell has also explored collaborative algorithmic music in the browser. Machines that answer back are not a sudden novelty here. They are part of a longer local interest in systems, participation and sound.
That tradition is visible in this year’s No Bounds programme, which connects Fell, Treanor and other established experimental artists with youth projects, grassroots collectives and venues across Sheffield and Rotherham. The important word is not “technology”. It is “connects”.
An AI model can produce a competent drum-and-bass version of a demo. It cannot programme a night, run a door, teach a teenager to beatmatch or persuade a room full of people to leave the house in the rain.
Exposed recently reported how Gut Level is pursuing community ownership. That work — protecting space, sharing skills and giving new artists somewhere to fail in public — will shape Sheffield’s next sound far more than any model release.
Daniel Dylan Wray’s Groovy, Laidback & Nasty, built from more than 150 interviews about seven decades of independent Sheffield music, returns to the same lesson. The city’s breakthroughs did not come from machines acting alone. They came from people finding cheap tools, using them incorrectly and building something social around the result.
AI remixing will matter here if it joins that tradition: not as a shortcut to unlimited content, but as one more cheap instrument that somebody can push past its intended use.
The machine can supply versions. Sheffield’s contribution has always been deciding which strange one deserves a room.
