What Is Ai Likeness Licensing?

AI likeness licensing lets people authorise the use of their face, voice or digital replica in AI-generated media. Here's how it works.

AI Likeness Licensing: When A Real Person Authorises A Synthetic Version Of Themselves

AI likeness licensing is the authorised commercial use of a real person's face, voice, appearance or digital replica in AI-generated media under agreed terms. Instead of somebody creating a synthetic version of a person without permission, the individual deliberately allows particular uses of their likeness and may receive payment when that likeness appears in advertising, entertainment or other digital productions.

The phrase is useful shorthand rather than one universal legal category. Different industries and jurisdictions use terms such as digital replica, digital double, likeness rights, voice rights and personality rights. The common idea is that a recognisable part of a real person is being reproduced synthetically and used under some form of permission.

That distinction matters because the underlying AI technology can look similar whether consent exists or not. A generated face, cloned voice or synthetic performance may technically resemble what people casually call a deepfake. The commercial relationship with the human is completely different when the person has agreed to the use, understands what has been authorised and can negotiate compensation or restrictions.

A developing market in China now shows what that could look like beyond celebrities. Platforms are allowing ordinary people, influencers and professional performers to register their likenesses for potential use in AI-generated dramas and advertising. It is still an early business model, but it points towards a future where human identity can become something people selectively licence rather than something synthetic-media companies simply copy.

How Does AI Likeness Licensing Work?

The basic process begins with material that allows an AI system to reproduce recognisable characteristics of a person. Depending on the application, that might include photographs, video, voice recordings, facial scans, motion data or an existing performance.

The person then authorises specific uses under a licence or contract. A production company might be allowed to place their digital likeness into an AI-generated drama, use their face in advertising, generate a synthetic voice performance or create variations of an approved character based on them.

The important part is the scope of that permission. Agreeing to one advertisement should not automatically mean agreeing to every future advertisement. Giving permission to create a synthetic performance does not necessarily mean giving permission to use the same data to train a general-purpose AI model. Licensing a face for one production should not quietly become permanent permission for another company to inherit the replica years later.

That makes AI likeness licensing as much a contractual problem as a technical one. The AI may be able to generate thousands of possible performances, but somebody still needs to define which of those performances are actually authorised.

China Is Already Building A Market For Licensed Faces

One of the clearest current examples comes from China's rapidly growing AI microdrama industry.

Platforms have emerged where people can register photographs or have images captured specifically so their likeness can be offered to AI-production companies. Producers can browse available faces, choose people suited to particular roles and negotiate or accept licensing terms for synthetic productions.

Shenzhen-based ActID is one example. By July 2026, the platform had registered around 800 people, with about 300 agreeing to licence their image for AI production. Participants included professional extras and influencers, but also ordinary people with no acting experience.

At that point the actual commercial market remained much smaller than the registrations suggested. Two AI drama productions had licensed around 10 faces through ActID, with reported prices ranging from 99 to 500 yuan per episode and the platform taking a commission.

That scale is useful because it keeps the story in perspective. This is not yet a mature global industry where millions of people earn passive income from synthetic versions of themselves. It is an early experiment showing that a commercial mechanism now exists.

The wider Chinese market also gives that experiment a reason to exist. AI is being used heavily in short-form drama production, while unauthorised use of faces and voices has already produced disputes. A licensing marketplace attempts to solve part of that problem by replacing imitation without permission with a clearer commercial relationship.

Is AI Likeness Licensing The Same As A Deepfake?

No.

A deepfake generally describes synthetic or manipulated media that realistically imitates a real person. The term tells us something about the content or technique, but it does not automatically tell us whether the person agreed to appear in it.

AI likeness licensing is about permission.

A performer might knowingly authorise a digital replica for a film, an influencer might licence their appearance for an advertisement, or an ordinary person might allow their face to be used in an AI-generated drama. Those productions can still contain highly realistic synthetic media, but the individual has participated in the arrangement rather than discovering their identity was used afterwards.

That does not mean licensed synthetic media is automatically ethical simply because somebody signed something. Contracts can be vague, bargaining power can be uneven and people may not fully understand how reusable AI data can become. Consent therefore needs to be specific enough to mean something.

The distinction is still essential. A system where people can knowingly participate, negotiate terms and receive compensation is fundamentally different from one where synthetic replicas are created first and permission becomes somebody else's problem later.

What Exactly Can Someone Licence To AI?

There is no single package called "your likeness". Different agreements can cover different parts of a person's identity and performance.

A licence might permit the use of a person's face or general appearance in generated visual media. Another could cover a voice replica capable of producing speech the individual never personally recorded. A more sophisticated digital replica could combine image, voice, movement and performance characteristics to generate new appearances of that person.

The important distinction is between allowing a particular output and allowing a reusable synthetic model.

If somebody licenses their face for one finished advertisement, the producer may only need permission to create and distribute that work. If the company instead wants to build a reusable replica capable of generating future performances, the commercial relationship is much broader.

Training rights create another layer. Images or recordings supplied for the creation of a particular replica might also be technically useful for improving an AI model. That does not mean permission for the first purpose should automatically include the second.

For creators and performers, those differences will become increasingly important as digital replicas become more capable.

Why The Scope Of The Licence Matters

A useful AI likeness agreement needs to answer more than whether the word yes appears somewhere in the contract.

The first question is what the replica can be used for. Permission for a fashion advertisement does not necessarily imply permission for political messaging, gambling promotions, adult content or an unrelated entertainment production. People may be comfortable associating their identity with some products and completely opposed to others.

Duration matters too. A licence might cover one campaign, one production, a fixed number of episodes or a defined period. Permanent permission creates a very different relationship from allowing a replica to exist for six months.

There is also the question of modification. AI can change age, body shape, accent, clothing, expressions and behaviour. A person who licenses an ordinary digital replica may not expect to be made to appear nude, violent, intoxicated or involved in something that could damage their reputation.

Generated performances make the issue even more complicated because the replica can potentially say or do things the real person never performed. That capability is part of what makes synthetic performers useful, but it also makes detailed permission more valuable.

Sublicensing matters for similar reasons. A person may agree to work with one company without agreeing that their digital identity can later be transferred to another business. And if they withdraw consent or the licence ends, the agreement needs to explain what happens to the replica and to media that has already been distributed.

These are not minor legal details attached to an otherwise simple transaction. They define what the product actually is.

How Do Performers Get Paid For Digital Replicas?

There is no universal payment model.

China's emerging face-licensing platforms show one possible approach, where individuals set or accept prices based on the type of use and production. Other arrangements can operate more like conventional performer contracts, where payment is tied to the creation or use of a digital replica.

SAG-AFTRA's current US commercials agreement gives a much more developed example. Under its 2025 contract, performers must consent before a digital replica is created and receive a reasonably specific description of how it will be used. When a replica generates a performance for a commercial, the performer is entitled to compensation under the agreement.

The contract also places limits around retention. Producers cannot simply keep a digital replica indefinitely after the relevant use period without further consent.

That approach is important because it treats synthetic performance as work with economic value rather than assuming the ability to automate a performance removes the human from the commercial relationship.

The exact rates and rules will vary outside union agreements, but the principle is likely to remain important: if somebody's recognisable identity helps create commercial value, there is a reasonable argument that the person should participate in that value.

Can Ordinary People Licence Their Face To AI?

Yes, at least in some emerging marketplaces.

The Chinese examples are interesting partly because they are not limited to famous actors or influencers. ActID has registered ordinary people alongside professional extras and online personalities, and other platforms have marketed faces according to age, appearance and the kinds of roles or settings producers may need.

That creates a different creator-economy possibility.

Traditional commercial casting usually depends on a person being physically available for a production. Synthetic media can separate the appearance from the original performance. Someone could potentially authorise a digital likeness that appears in multiple pieces of media without personally acting in every scene.

That could lower some barriers to commercial work, particularly for localisation, background roles, virtual presenters or productions operating across different markets.

It could also create new inequalities if people are encouraged to surrender broad rights for very little money while companies gain much greater long-term value from the replica.

The existence of a marketplace does not automatically make the transaction fair.

The interesting part is that ordinary identity can now have direct digital production value. That used to be a concern associated mainly with celebrities whose image already carried obvious commercial worth.

AI is expanding the question to everybody else.

What Does The UK Say About AI Digital Replicas?

The UK does not currently have one general legal right that gives everybody complete control over the commercial use of their face or voice.

Instead, protection can come from a combination of areas such as data protection, defamation, passing off, intellectual-property rights and criminal law depending on what happened. The government has acknowledged that this patchwork does not cover every form of unauthorised digital replication.

Its March 2026 Copyright and Artificial Intelligence report specifically examined digital replicas, including AI-generated reproductions of people's voices and appearances. The government recognised legitimate uses such as dubbing, synthetic voices for disabled people, reduced reliance on risky stunt work and digital doubles used in professional settings.

It also acknowledged that non-consensual replicas can create serious harm and that existing UK protections may leave gaps, particularly for ordinary people who do not have the commercial reputation required to rely on some existing legal remedies.

The government therefore plans to explore possible additional protections, including whether a new digital-replica or personality right could be useful.

That discussion is still developing. The UK has not simply created a new ownership right over everybody's face or voice. The significance is that the government now accepts the underlying problem: AI is making realistic replication cheap and accessible enough that older assumptions about impersonation may no longer provide adequate protection.

Why Consent Alone Is Not Enough

Saying that AI likeness use should be consensual sounds simple until we ask what someone actually agreed to.

A person can technically consent to a contract containing terms they barely understand. A small creator might also have much less negotiating power than the company asking for extensive rights to their image. A one-off payment can look attractive without making the long-term value of the licence obvious.

Good consent therefore needs information.

The person should understand what is being created, where it can appear, how long it can be used, whether the replica can generate new performances, whether their appearance can be materially altered, whether the data can train AI and whether another company can later receive those rights.

Security matters as well. A valuable authorised replica can still be copied or stolen. Once high-quality facial or voice data exists in a reusable form, protecting it becomes part of protecting the individual.

The more capable digital replicas become, the less useful vague permission becomes.

AI Could Create A New Identity Economy

It is too early to describe human likeness as a fully developed new asset class. The commercial experiments are still small, legal frameworks are inconsistent and much of the synthetic-media industry is developing faster than the contracts around it.

But the direction is becoming visible.

Creators could maintain authorised digital versions of themselves that can appear in advertisements, games, entertainment, virtual environments, localisation work or other synthetic productions. Performers might approve specific generated performances without being physically present for every production. Brands could licence a recognised creator's synthetic presence with clearer terms than simply feeding their content into a model and hoping nobody complains.

That creates what we can reasonably describe as an emerging identity economy: commercial activity built around controlled access to a person's recognisable digital self.
The existence of unlimited synthetic faces does not necessarily make real human identity worthless. It may do the opposite in some contexts.

If anybody can generate a convincing fictional person, then a verified real person who has deliberately authorised a particular use offers something the fictional character cannot: a legitimate human relationship behind the representation.

For influencers, actors, musicians and other creators, that could eventually make identity-management another part of running a career. They may need to think not only about who can repost their work, but who can generate new work as them.

Does AI Likeness Licensing Make Synthetic Media Safer?

It can make one important part safer: consent.

Licensing creates a route for somebody to participate knowingly rather than discovering an imitation after publication. Clear contracts can define payment, permitted uses, retention periods and restrictions around how a replica is changed.

It does not remove every risk.

A licensed digital replica can still be stolen. A company can still exceed the agreed scope. Someone can still create an unauthorised copy somewhere else. People can also sign poor agreements that hand away more control than they intended.

That means licensing is not a technical solution to every synthetic-media problem. It is a commercial and legal framework for distinguishing authorised uses from unauthorised ones.

The technology will keep getting better either way. Giving people better mechanisms to control legitimate uses of their identity is therefore likely to become more useful, not less.

Tanizzle Says: Permission Has To Scale With The Technology

We do not think the interesting future of synthetic media is one where every realistic digital person is treated as inherently suspicious. AI can create new forms of entertainment, localisation, advertising and accessibility, while digital replicas can give performers and creators ways to participate in work that would otherwise require their physical presence every time.

But technical capability cannot be allowed to turn one permission into unlimited permission.

If somebody licenses their identity for a specific purpose, the agreement should be capable of respecting the same level of precision that the AI itself can achieve. Who can use the replica, where it can appear, what it can say, how it can be changed and how long it can survive should not be hidden behind one vague approval.

We are pro-AI because the technology creates possibilities. We are equally comfortable saying that the person whose face or voice creates the commercial value should retain meaningful control over those possibilities.

If human identity becomes programmable, permission needs to become programmable too.

From Tanizzle: For You

AI likeness licensing is easier to understand once the difference between authorised synthetic media and imitation without permission is clear. Our guide to deepfakes under the EU AI Act looks more closely at synthetic depictions and the growing transparency rules surrounding them.

Digital replicas also connect with a broader shift towards entertainment built around AI from the beginning rather than added as a final effect. We explore that creative model in What Is AI-Native Entertainment?.

And if the idea of maintaining a digital version of a person sounds familiar, our guide to digital twins explains the wider concept and how a digital counterpart can represent a real-world person, object or system.

Tanizzle FAQs: Understanding AI Likeness Licensing

What is AI likeness licensing?
AI likeness licensing is an arrangement where a real person gives permission for their face, voice, appearance or digital replica to be used in AI-generated media under agreed commercial terms. The licence can define where the likeness may appear, how long it can be used and what compensation the person receives.

Is AI likeness licensing the same as a deepfake?
No. A deepfake describes synthetic or manipulated media that realistically imitates a person, whether authorised or not. AI likeness licensing specifically concerns uses where the real person has granted permission under agreed terms.

Can ordinary people licence their face to AI?
Yes. Emerging platforms in China are already registering ordinary people alongside influencers and professional performers so their likenesses can potentially be licensed for AI-generated dramas and advertising. The market is still early and relatively small.

Can a company make an AI version of me forever once I agree?
Not necessarily. The duration and permitted use should depend on the licence or contract. A well-defined agreement may limit use to a particular campaign, production or time period rather than granting permanent access to the digital replica.

Does licensing my likeness mean a company can train AI on my data?
Not automatically. Permission to create or use a particular digital replica is not inherently the same as permission to use the underlying images, recordings or performance data to train an AI model. Training rights should be addressed separately in the agreement.

Can an AI replica say things I never said?
Potentially, yes. A sufficiently capable digital replica can generate performances that the real person never personally recorded. That is why licensing terms may need to define what kinds of generated speech, behaviour and modifications are permitted.

How do people get paid for AI likeness licensing?
Payment models vary. Some emerging platforms allow people to set prices for particular uses of their face or image, while professional performer agreements may tie compensation to the creation and use of a digital replica. There is no universal industry rate.

Does the UK have a digital likeness right?
Not currently as one general standalone right covering everybody. The UK relies on a mixture of existing legal protections, but the government has acknowledged gaps and is exploring whether additional protection, including a potential digital-replica or personality right, would be useful.

Can someone revoke permission to use an AI replica?
That depends on the agreement and applicable law. A licence should ideally state whether permission can be withdrawn, when the replica must be deleted and whether previously released material can continue to be distributed.

Why would someone licence their likeness to AI?
Potential uses include advertising, entertainment, localisation, games, virtual presenters and other synthetic productions. Licensing can allow a person to participate commercially without physically performing every generated appearance, provided the terms give them meaningful control over how their identity is used.