Articles · Ahead of the Curve
AI Will Clone Your Voice Whether You Like It or Not. Some Artists Are Already Getting Paid for It.
The Set Newsroom · Thu Sep 24 2026
Grimes licensed her voice in 2023 and still gets paid every time someone uses it. Universal just spent a year in court to build the same deal for its artists. Independent musicians can skip the lawsuit.
AI is going to clone your voice. That part is not up for debate anymore. The only real question left is whether you get paid when it happens, or whether you find out about it in a TikTok comment section. A handful of artists already answered that question for themselves, and independent musicians can copy the exact same move today, no label required.
Grimes Did This First, And It Still Works
In 2023, Grimes did something almost nobody else in music was willing to try. She released her own voice as an open-source AI model called GrimesAI-1 through a free platform called Elf.Tech, and committed to a 50% master-royalty split with anyone who used it to make a commercial song. Instead of fighting every AI cover of her voice one takedown at a time, she made licensing the clone easier than stealing it.
https://www.youtube.com/watch?v=WPr6feJcsl0
The model is not a novelty that quietly died. As of June 2026, the program has produced 168 official commercial releases on the GrimesAI Spotify profile, distributed through TuneCore with her approval. In October 2025 she also released "Artificial Angels", a song built around the same idea, sung from the perspective of an AI. Three years in, Elf.Tech is still running, and it is still paying her.
Grimes: Instagram | YouTube | Spotify
Independent Artists Don't Need A Major's Permission
You do not need Grimes' fame to do this. Voice-Swap, founded by drum and bass producer DJ Fresh and Nico Pellerin, lets any artist upload 25 minutes of singing or eight a cappella tracks and turns it into a licensed voice model fans can use to make their own tracks. Featured artists on the platform, including Robert Owens, Ayah Marar and Liam Bailey, get a 50% share of the revenue their voice model generates, paid out at the point of inference, which means the moment someone actually uses your voice, not months later on a royalty statement.
DJ Fresh: Instagram
That is a meaningfully different deal than streaming. On Spotify you wait for someone to hear your song and hope they stream it enough times to matter. On Voice-Swap, you get paid the instant your voice does the work, whether or not the resulting song ever gets released anywhere.
Voice-Swap is not the only door in. Kits AI runs a similar model, giving artists a way to license a vocal model with defined usage terms instead of just hoping nobody rips their stems off a livestream. Neither platform is asking you to hand over your masters or sign away your catalog. You are licensing one specific thing, your vocal timbre, the same way a sample clearance licenses one specific loop.
The Majors Just Caught Up To This
Universal Music Group spent most of 2025 suing Udio over unlicensed AI training. By late October 2025 they had settled, and the two companies are now building a "walled garden" platform launching in 2026 where UMG artists can opt in, catalog by catalog, to license their voice and music for AI generation. Participating artists get paid both when their catalog trains the model and again every time a fan generates a new song from it. Nothing generated on the platform can be downloaded or posted anywhere else, which is the label's way of keeping the whole thing inside a system it can actually track and pay out from.
The point is not that the UMG deal is generous. It is that a major label just spent a year of legal fees to arrive at the same basic structure independent artists have been able to set up for themselves since 2023.
The Part Nobody Advertises
None of this is a guaranteed payday. Voice-Swap's subscription tiers run $5.99 to $29.99 a month for the fans and producers using the models, which means your 50% cut is a share of a small number, not a windfall, unless your voice model actually gets used a lot. Licensing your voice is also not reversible in the way a bad contract clause can sometimes be renegotiated. Once a model exists and people have used it, you cannot fully claw it back, even if you shut the platform access down later. Read the terms on usage scope and duration before you upload anything, the same way you would before signing a publishing deal.
How To Actually Do This If You're Independent
- Decide what you are licensing before anyone asks. A voice model, a specific song, a vocal style. Vague consent gets you a vague check.
- Use a platform built for royalty tracking at the source, like Voice-Swap or Kits AI, instead of handing raw vocal stems to whoever asks nicely.
- Put your terms in writing. Where the clone can be used, for how long, and what you get paid, whether that is a flat fee, a percentage, or both.
- Disclose when a track uses a licensed AI voice, yours or someone else's. Spotify is now actively de-ranking undisclosed AI identities, so transparency protects you twice over.
If you want help building the actual songs and content around this instead of just the licensing side, The Set's beat and mixing and mastering services exist for exactly that, and every artist on the platform keeps their masters regardless of what gets licensed out.
The Takeaway
Somebody is going to generate a song in your voice this year whether you sign anything or not. The only choice you actually get is whether that happens for free or for a royalty check. Grimes made that choice in 2023. DJ Fresh built a business around it. Universal just spent a year in court to catch up. You can set the terms today, for free, before anyone asks.
Questions
Can independent artists license their voice for AI music without a record label?
Yes. Platforms like Voice-Swap and Kits AI let any artist upload vocal recordings and turn them into a licensed AI voice model, with royalties paid directly to the artist and no label involvement required.
How much do artists get paid when their AI voice model is used?
On Voice-Swap, featured artists receive a 50% revenue share from their voice model, paid at the point of inference, meaning the moment someone actually generates audio with it, not only through eventual streaming royalties.
What did Grimes do with her voice and AI?
In 2023, Grimes released an open-source AI voice model called GrimesAI-1 through Elf.Tech, taking a 50% master-royalty split on any commercial song made with it. As of June 2026, the model has been used on 168 official commercial releases.