DJ Fresh doesn’t want AI to fall in the wrong hands — so he started Voice-SwapDaniel Stein, better known to the world as DJ Fresh, has a unique vantage point from which to see the future of music. After rising to fame in the early 2010s with a string of dance hits, including the ultra-catchy Hot Right Now, Stein has spent the last few years not on the decks but digging deep into machine learning algorithms.
The fruits of that effort is Voice-Swap, an AI voice modelling service aimed squarely at music producers. However, Stein is not simply jumping on the ‘vocal deepfake’ bandwagon – he’s laying out a framework for how AI technology can be used ethically in the music industry.

READ MORE: How will young people learn music and production in a post-AI music industry?

“There is a feeling, within the company, that we are fighting for the cause of artists,” Stein says. “And not just artists; we respect the importance of labels, rights holders, distribution companies. Basically, we don’t want the music industry to end up in the hands of technologists who don’t care about music.”
Unbound creativity is exactly where AI will have an exciting impact on music creation
The platform partners with singers to create artist-authorised voice models that can be used for demos or legally licensed music releases. Users can choose from three monthly subscription tiers, depending on how many minutes of audio they need to generate. Moreover, 50 per cent of the platform’s gross subscription income goes directly to the artists making their voices available on the platform.
Credit: Voice-Swap
Since launching in mid-2023, Voice-Swap has racked up a host of high-profile users like Diplo, Skream, and Pendulum’s Rob Swire. “It was a surreal moment showing Farley “Jackmaster” Funk a video of Skream blowing up a club with Farley’s voice-swap model,” Stein enthuses. “Farley, one of the fathers, if not the father of Chicago House, and Skream, the father of dubstep — these guys are both pioneers. Seeing Skream embracing the weird artefacts that the models sometimes create — we think that kind of unbound creativity is exactly where AI will have an exciting impact on music creation.”
Voice-Swap has also steadily expanded its roster of artists, which now includes vocalists like Ruth Royall, Robert Owens, and legendary house vocalist, Angie Brown. Recently, the team began working with electro-pop’s premiere innovator, Imogen Heap, on a voice model, and, while there’s no word yet on whether Heap’s voice will become available for the Voice-Swap platform, Stein says the company will be announcing “new projects in collaboration with Imogen and other well-known artists in the near future.”

The service was initially positioned as a tool for creating high-quality demo tracks, with users free to share their creations with collaborators, management teams, record labels, or post on social media. However, Stein says the use cases have expanded to include translating songs into other languages, communication assistance for those with speech difficulties, and even as a gimmick for influencers and podcasters looking to make interesting content.
Then there are users who want to take their demo and turn it into an official release. Voice-Swap offers three AI ‘session singers’ whose voices can be commercially released at no extra charge – however, if a producer wants to release a track using a model from one of Voice-Swap’s featured artists, then Stein says they need to first attain the artist’s “blessing” in the form of a commercial licence.
Luckily, the platform offers a straightforward system to facilitate this process, and artists retain an 80 per cent cut of any licence fee. Crucially, vocalists signed up to the platform have the final say over these licence requests and, if they have any artistic or moral objections, are free to deny permission for the song’s release.
Stein says users who monetise a track without attaining a proper licence can expect takedown requests to be issued, along with a promise of potential legal action: “We’re making sure that people are aware of the consequences of doing things illegally, and trying to make sure users value the artists who we’re working with.”

View this post on Instagram

A post shared by Voice-Swap.ai (@voiceswapai)

“At the moment,” he continues. “There’s a number of platforms out there selling models of people’s voices without any permission — that is wrong on so many levels. We’re competing against those platforms and, in some sense, we’re competing at a disadvantage because we’re trying to do things the right way.”
Putting artists at the centre of Voice-Swap’s founding ethos is not just the right thing to do, says Stein, it’s also a practical consideration for the businesses’ long-term success: for artists to have their voices modelled and released on Voice-Swap, there needs to be a sense of trust in the platform.
“Some of the less ethical people just view this as a gold rush,” he says. “They don’t really care if they mess things up because they’ll just jump on the next tech fad in a few years time. However, a lot of the people involved in Voice-Swap have had long careers in the music industry and we care a lot about our reputations.” Such professionals at Voice-Swap include Benn Jordan, Declan McGlynn, Ausrine Skarnulyte and Nico Pellerin, among others.
Much of the discussion around AI voice models has centred on the potential negatives for singers, and for good reason. Stein himself says that, over the past couple of years, “artists and creators have been abused by the AI industry”. However, he makes clear that Voice-Swap is more than an attempt to redress the balance and claw back some remuneration – he believes the technology opens up whole new ways of working and even new career paths for singers.
DJ Fresh. Image: Press
“Big artists don’t have time to go into the studio and demo every song idea being sent to them by producers,” Stein says. “There is this whole layer of the industry that determines which songs actually get made into demos and played to labels, but the question is – why do these artists have to demo anything? In theory, they don’t have to. Producers can use a model of their voice to create demos and the artist only goes into the studio when they’re sure the song is going to be released.”
Such a system would be a massive shift in how the music industry works at a functional level, but productivity is just one area where Stein sees positive outcomes. The technology could also offer new revenue streams for artists who find themselves unable to tour or record in-studio.
He points to Mylène Schuler, one of the artists whose vocal model is available on the platform. “She was an up-and-coming singer in Switzerland, part of an electro-pop group that was doing really well,” Stein says. “She had to bow out because she had a kid and she needed to focus on her family, but she still wanted to keep a foot in the music industry.”
After uploading her vocal model to the platform, Stein says that within two weeks he got an excited call from Pendulum’s Rob Swire, who had written a song using Schuler’s model.
“It’s such a great example of this technology having benefits for people that, for one reason or another, can’t sing anymore,” Stein exclaims. “Maybe it’s a physical reason, maybe their lives have changed, but they can still be involved in the industry and know that a producer might find their voice and make a massive hit record with it.”

In an industry whose ‘move fast and break things’ maxim has left many creatives disillusioned, Voice-Swap aims to show that tech start-ups can put fair remuneration and artist rights at the centre of the business and still succeed. Part of that is creating a culture of transparency around the datasets being used, and ensuring that the correct people are credited. Recently, the company partnered with one of the world’s largest music rights attribution services, BMAT, to certify their voice models.
If we don’t get the policy around this under control soon, there could be a few companies that make a fortune and a lot of people that lose out.
“Our partnership with BMAT helps us make sure that models are not trained with copyright data,” Stein points out. “We very much believe that AI is to benefit humans, not the other way around, and the roots that we put down at the beginning of 2023 are now sprouting into partnerships that cement our ethos.”
What makes that stance even more notable is that, right now, platforms like Voice-Swap are operating in a profound legal and political vacuum. “When it comes to policy and high-level governance, there’s no clarity,” Stein says frankly. “We’re talking to lots of people in the music industry, to lawyers and government bodies, but it feels like nobody wants to set the agenda. Instead, everybody’s waiting to see what happens. We’re trying to get it right, and to set a precedent that artists and labels can look to – but we’re certainly not getting any guidance.”
The need to set the right precedent is a topic that comes up repeatedly in conversation with Stein. Over the coming years, he sees personalised audio content for games, media, and marketing kicking off an explosion of demand for AI voice models. If artists, platforms, and governments can coalesce around a shared set of standards for this new economy rather than engaging in a cut-throat race to the bottom, then generative AI could ultimately be a net positive for music creators.
“The market is growing at this enormous rate,” Stein says. “There’s a lot more money to be made. As long as artists are being paid fairly it’’ a bright future – but, if we don’t get the policy around this under control soon, then, just like so many other things that have happened over the past 20 years, there could be a few companies that make a fortune and a lot of people that lose out.”
Learn more about Voice-Swap. 
The post DJ Fresh doesn’t want AI to fall in the wrong hands — so he started Voice-Swap appeared first on MusicTech.