AI Voice Licensing: How Performers Can Protect Their Voice

What does it mean to own and protect your voice when a recording can be used to create a digital replica?

For performers, the question is becoming increasingly practical. AI voice technology is improving quickly, but the contracts, platforms and business models surrounding it can determine whether an actor retains meaningful control over how their voice is used.

In Episode 6 of HIGH NOTES, Melissa Thom speaks with BAFTA award-winning actress Cissy Jones, known for performances including Firewatch, The Walking Dead, Baldur's Gate 3, Fallout 4, Starfield, Life is Strange, Halo 5 and The Owl House.

Cissy is also co-founder and CEO of Ethovox, a voice AI company created by people from within the voice industry. Their conversation explores what performers should ask before licensing their voice, how consent can work in practice, what happens when contracts were written before today's technology existed and why Cissy still believes human performance offers something AI cannot reproduce.

From having her voice cloned to building an alternative

Cissy's involvement with AI did not begin as a business opportunity.

Around 2020 and 2021, as The Owl House attracted a passionate online following, fans began creating their own scenes and dubbing characters themselves. At first, Cissy found it fun to see people trying to imitate her character's voice.

Then she began hearing her actual voice saying things she had never recorded.

Some of the material quickly became explicit. As a parent, she was particularly disturbed by the possibility that her children would encounter recordings that sounded like her but contained things she had never said.

Cissy contacted AI voice platforms asking for her voice to be removed. Some responded positively. She recalls another taking a very different position and telling her, in effect, that she did not own her voice and that the platform was not responsible for what users uploaded.

The experience made the implications of synthetic voice technology suddenly very real.

Rather than treating AI as something the industry could simply reject, Cissy began meeting regularly with others working in voiceover to consider a different question: if the technology was going to exist, what would responsible participation look like?

Those conversations eventually connected with the National Association of Voice Actors, NAVA, and its work around performer protections and responsible AI use.

The principle, as Cissy describes it, was not to be anti-AI or anti-technology. It was to be pro-actor.

That thinking ultimately contributed to the creation of Ethovox.

Consent has to come before the replica

One of the fundamental distinctions Cissy draws is between creating a replica with a performer's agreement and creating one simply because enough recordings of their voice exist.

At Ethovox, she says a voice does not enter the system without a corresponding agreement with the actor.

Once that agreement exists, the performer records material specifically for the replica. Scripts can be designed around the kind of work the voice may eventually be used for, including particular vocabulary or specialist terminology.

But the bigger difference is what happens afterwards.

Cissy argues that an open voice marketplace can create serious problems for professional performers. An actor may, for example, have an exclusive relationship with one automotive brand. If another automotive company can independently select that performer's synthetic voice from a marketplace, the actor could find their digital replica doing work that conflicts with an agreement governing their human career.

In Cissy's model, the performer therefore retains a role in approving individual uses.

If a client wants access to a replica, the actor can say yes or no. And if the voice is used, the performer is compensated.

The principle is relatively simple: the existence of a digital replica should not mean the performer disappears from the decision-making process.

What should you ask before licensing your voice to AI?

When Melissa asks what every performer should check before agreeing to AI voice work, Cissy starts with safety.

Who can access the voice? What happens to the recordings used to create it? Can that material also be used to improve other AI models? Can it be licensed onwards? And does the performer share in any value created from that use?

For Cissy, understanding the terms of service is essential.

A contract may appear to concern one particular use of a synthetic voice while simultaneously granting much broader rights over the audio supplied to create it.

She also raises the issue of exclusivity. If anyone can access a performer's replica through an open marketplace, the actor may no longer be able to guarantee a future client that their voice will remain exclusive within a particular sector.

That could affect the kind of conventional voiceover work they can accept.

Another important question is what happens if the performer changes their mind. Can they withdraw? Can their data be deleted? Or has the platform acquired rights that continue indefinitely?

Cissy points performers to NAVA's resources for considering these questions before joining an AI platform. Her broader message is to understand not only what you might gain from an agreement, but what you may be giving up in return.

Digital replica - A synthetic version of a performer’s voice created from recordings of their real voice. AI rider - An additional clause or document attached to a contract that sets out specific protections around AI use. In perpetuity - Contract wording meaning the rights being granted do not expire. Synthetic voice - A computer-generated voice, which may be entirely artificial or based on a real performer’s voice.

Why old contract language becomes complicated in the age of AI

Voiceover contracts have long contained broad usage clauses.

Cissy gives the familiar example of wording that grants rights “in perpetuity”, across the known universe and through technologies that currently exist or may be developed in future.

Historically, that language existed in a very different technological environment.

Generative AI changes what “technology to be developed” can potentially mean.

A recording originally created for one game, animation or commercial may now also contain the raw material from which a synthetic version of a performer could potentially be produced.

Cissy therefore urges performers to pay close attention to broad clauses that might once have seemed routine.

She also points to NAVA's Synthetic Voice/AI Rider, a contractual addendum designed to provide specific protections around AI and synthetic voice use. NAVA makes the rider available publicly rather than restricting it to members.

The wider lesson is that a clause written for yesterday's media environment may carry very different consequences when applied to today's technology.

What is fair compensation for an AI voice?

There is no single straightforward answer to what performers should currently be paid for AI licensing.

Cissy describes rates as being “all over the board”.

Her own rule of thumb is more personal: when the payment arrives, will she regret having agreed to the job?

That may sound informal, but it reflects something important about AI licensing. Payment cannot be considered separately from the rights being granted.

A relatively attractive session fee may look very different if the resulting model can be used indefinitely, licensed elsewhere or applied to work the actor never gets to approve.

Cissy also describes Ethovox's longer-term ambition to build its own foundational model using fully contracted and traceable recordings.

Under the SAG-AFTRA agreement discussed in the episode, performers contributing recordings to the foundational model would receive payment for the original recording work and a share when the resulting model generates licensing revenue.

The objective is to create a system in which the people whose performances contribute to the technology continue to participate economically in the value it creates.

Even someone running an AI voice company sees the limits of AI performance

One of the most striking moments in the conversation comes when Cissy turns from contracts to performance.

She is clear that AI voices can sound very good.

Sometimes, she says, they can even sound great.

But as somebody running an AI voice company, she also sees the difference between manipulating synthetic speech and directing another performer.

An AI system can offer controls over how something sounds. But compare that with simply saying to an actor: punch that word differently, finish the line slightly higher, or make this moment feel more wistful.

A human performer can understand the intention behind that direction and respond creatively.

Cissy's conclusion is emphatic: however capable the technology becomes, it is still not as simple or responsive as turning to another human performer and explaining what you need emotionally.

“AI can't,” she says. “AI could never.”

That distinction matters because acting is not simply producing a technically correct sound.

Direction is a conversation.

Control means knowing who is using your replica and why

Melissa asks what additional protection a managed system can offer compared with a performer negotiating directly with an AI platform.

Cissy reduces the answer to two fundamental points.

First, a replica should not exist without the performer's agreement.

Second, each client wanting to use that replica should require the performer's approval.

“You get to grant the key,” she explains.

That means the performer knows who wants to use the synthetic voice and for what purpose.

The technology can still provide scale and flexibility, but the actor does not surrender all involvement simply because a replica has been created.

It is an important distinction between consenting once to the creation of a digital voice and consenting to every possible future use of it.

AI could expand what performers offer rather than simply replace them

When Melissa asks what the future might look like if performers retain control over licensing, Cissy is optimistic about some of the possibilities.

She does not believe AI can reproduce the human condition behind a performance.

Characters such as Delilah in Firewatch, she argues, depend on levels of emotional communication that come from a performer rather than merely from the sound of their voice.

But she also sees uses for synthetic voices that extend what an actor can offer.

One example is pre-production or “scratch” audio for video games. A synthetic voice could allow developers to test and iterate scenes before the final performance is recorded.

Another is work that would be practically impossible for a human performer to deliver at the required scale or frequency. Cissy gives the example of producing constantly updated material in many languages at a particular time every day.

A licensed replica could potentially do that work while leaving the performer available to record human pickups or other material when required.

For Cissy, that represents expansion rather than straightforward replacement.

The technology is not the whole question

Across HIGH NOTES Season 4, the recurring issue has not simply been whether AI voices sound convincing.

The harder questions concern who gave permission, who retains control, how the resulting voice can be used and who benefits financially.

Cissy's experience brings an unusual perspective to that discussion because she sits on both sides.

She is a working performer who has experienced her voice being reproduced without permission.

She is also building a company that uses the same broad category of technology.

Her answer is therefore not to pretend AI voice technology does not exist, nor to assume that every use of it is inherently harmful.

Instead, she argues for a model in which the performer remains visible throughout the process.

They know when their replica is created.

They understand what they are licensing.

They can decide which uses they accept.

And they are paid when their voice creates value.

At the same time, the conversation makes another point equally clearly: a digital replica and a human performance are not necessarily interchangeable.

Synthetic speech may become increasingly sophisticated. It may take on work that is impractical for humans and become another useful tool in production.

But when a director needs an actor to understand a thought, interpret a feeling and respond creatively in the moment, Cissy still sees the human performer as something fundamentally different.

The future she describes is therefore not simply human voice versus AI voice.

It is a question of where each belongs, and whether performers are given enough information and control to make that decision for themselves.

Listen to Episode 6 of HIGH NOTES, Owning and Protecting Your Voice with Cissy Jones, for Melissa's full conversation with Cissy about Ethovox, AI voice licensing, contracts, performer consent and the future of human performance.

This article reflects the discussion in the episode and is for general information, not legal advice. Performers considering AI licensing agreements should seek advice appropriate to their own circumstances.