AI can now produce singing voices and instrumental performances that sound real, and that leaves the music industry with a money question: when musicians’ recordings are used to train these systems, who gets paid?
Most AI-generated vocals and virtual instruments are built on recordings of real musicians. Their skill, technique and artistic expression go into the model. The terms under which those recordings are acquired, and what the musicians get for them, vary a lot from one company to the next.
Some developers are taking a different route. They train on performances they have licensed, they get informed consent from the artists, and they sign compensation agreements that give musicians a share of the money the models make. In this setup, musicians are treated as partners in building the product rather than as a source of training data.
Consent and licensing in AI music training
Teaching an AI to reproduce an expressive performance takes more than a folder of audio files. The model picks up traits that singers and instrumentalists spent years developing as professionals.
For a vocalist, the voice is a large part of their identity as an artist. Timbre, articulation, phrasing and expression are what make one singer sound different from another. When those qualities end up in an AI voice model, two questions follow: who authorized the use of the recordings, and how will the model be sold?
A responsible licensing process settles four things before training starts. First, performers give informed consent, meaning they know how their recordings will be used and what the finished model will be able to do. Second, the developer obtains the rights it needs to use the recordings and performances for the purposes it has in mind. Third, the agreement sets out the commercial rights: how the model can be distributed, who can access it and how it can be used commercially. Fourth, musicians are told clearly how they will be paid for their performances.
With these terms in place, developers and the performers whose work trains the models know where they stand. Producers and companies that use AI vocals or instruments also get more certainty, which matters most when a track is going to be released commercially.
Licensing doesn’t settle every question about fair pay, though.
Royalties based on how often a model is used
Musicians have traditionally been paid through session fees, licensing payments, royalties, or some mix of these.
AI voice and instrument models make another arrangement possible. A musician might record for a fairly short period, while the model built from those recordings stays on sale for years and turns up in many productions.
That means payment can follow the model’s commercial use over time instead of stopping at the original recording session. Under usage-based royalty sharing, participating musicians earn according to how often their licensed AI models get used.
For a professional vocalist, this could be an extra source of income on top of session work, recording contracts and live shows. Instrumentalists could earn the same way when producers use AI instruments built from their licensed performances.
It also changes the relationship between musicians and technology companies. An artist who would otherwise get a single payment for recordings that feed a successful product can keep earning from that product for as long as it brings in money.
Not every royalty agreement is equally good for artists, of course. Payment rates, how usage is calculated, how it is reported and the contract conditions all decide whether an arrangement is worth much to the people taking part. A transparent agreement explains these terms in a way artists can understand and gives a clear basis for working out what they are owed.
The underlying point is that musicians whose work goes into AI development should have a chance to earn from it, instead of being shut out of the value the technology creates.
How ACE Studio pays its artists
Some AI music companies already build artist licensing and ongoing payment into the way they develop their products.
ACE Studio is one of them. It is an AI music production platform that offers singing voice synthesis and virtual instruments based on performances by professional musicians.
The company says its AI voices and instruments are made from recorded, licensed performances by participating vocalists and instrumentalists. Instead of relying only on one-time payments, ACE Studio says its professional artist partners get royalty shares based on how much the relevant AI models are actually used.
So the musician’s connection to the technology doesn’t have to end once the recordings are delivered. As producers use an artist’s AI voice or instrument model, the platform’s royalty-sharing arrangement keeps paying that artist.
The company describes its approach as working with real artists as partners and making sure they are paid for what they contribute.
One example is the professional vocalist Clara Sorace, whose voice has been turned into AI singing models available on the platform. Her collaborator Stefano Maccarelli has talked publicly about how the recordings were made and the models developed, and about the payments tied to the use of her AI voice.
Her case shows a professional performer taking part in AI development while keeping a financial stake in the result.
Royalties paid to participating artists and royalties paid by music producers are two separate things. Depending on the license attached to a given AI voice, a producer may be allowed to release music made with that model without paying the original vocalist any additional per-release royalty. The platform pays its artist partners separately, under their own agreements.
In other words, a producer using a voice royalty-free doesn’t mean the musician whose recordings went into the model goes unpaid.
Licensing terms differ from case to case, but this model shows that artist pay can be built into the commercial structure of AI music technology.
A fairer deal for musicians
The questions raised by AI music go beyond what the software can do.
Musicians need to know how their performances are obtained, what rights they are granting, and whether they will earn anything from the commercial products those performances help create.
Licensing agreements, informed consent and usage-based pay are practical ways to answer these questions.
Royalty sharing isn’t the only way to pay artists fairly. It does acknowledge that a performance can keep earning money long after the recording session is over.
As AI becomes a bigger part of music production, developers that deal openly with professional performers can show that new technology and paying artists don’t have to pull in opposite directions.
Whether AI music turns out to be ethical may depend on what these systems can generate, and also on whether the musicians who helped build them are properly credited and paid.
