AI Actors, Synthetic Voices, and Digital Replicas: Who Controls a Performance?


Episode 6 | AI Filmmaking & Copyright
Who Controls a Synthetic Performance?
Imagine casting an actor who has never existed.
You decide what that actor looks like, how they move, how they speak, and how they respond emotionally. You direct the performance, revise the dialogue, and assemble the results into a finished scene. But the actor has never stepped onto a set, read a screenplay, or spoken a single word.
Welcome to the world of synthetic performers.
As generative AI becomes increasingly sophisticated, filmmakers are confronting questions that go well beyond whether AI-generated images can be copyrighted. Who controls a synthetic performance? What happens when an AI character resembles a real actor? And where do the rights of the filmmaker end and those of the performer begin?
Three recent examples illustrate just how complicated those questions have become.
Tilly Norwood: The Actor Who Never Existed
In 2025, a London-based production company introduced Tilly Norwood, an entirely synthetic actress created using generative AI.
Her creator, Eline van der Velden, described Tilly as a fictional character developed through extensive human creative oversight. In 2026, Particle6 announced Misaligned, a feature film starring the synthetic performer.
The reaction from Hollywood was immediate.
SAG-AFTRA strongly criticized the concept, arguing that AI performers threaten the livelihoods of working actors and that generative systems have been trained on performances without adequate permission or compensation.
But Tilly raises an interesting distinction.
Filmmakers have been creating fictional characters since the earliest days of motion pictures. We've all seen the familiar disclaimer at the end of a film: "Any resemblance to actual persons, living or dead, is purely coincidental."
Creating an original synthetic character is, in my view, an extension of that same creative tradition. The filmmaker establishes the character's appearance, personality, behavior, and dramatic purpose. Generative AI simply provides another means of bringing that character to the screen.
But suppose I take a photograph of a recognizable actor, use AI to reproduce that person's facial features, and generate a performance the actor never agreed to give.
That's an entirely different proposition.
Now we're no longer discussing the creation of an original fictional character. We're potentially reproducing someone's identity without permission.
The distinction isn't whether AI was used. It's whether we're creating an original character or appropriating the recognizable identity of a real person.
That distinction doesn't eliminate concerns about training data, employment, or an unintended resemblance to an actual person. And the familiar fictional-character disclaimer doesn't automatically protect a filmmaker from a legitimate likeness claim.
But those concerns shouldn't automatically make every original synthetic character equivalent to an unauthorized digital replica.
When a Voice Becomes Someone Else's Property
In 2024, professional voice actors Paul Skye Lehrman and Linnea Sage sued AI voice-generation company LOVO, alleging that recordings they provided for limited freelance assignments had been used to create and commercially market synthetic versions of their voices without appropriate authorization.
In July 2025, a federal judge allowed several claims to proceed, including claims involving New York publicity rights and breach of contract, while rejecting substantial portions of the plaintiffs' federal copyright and trademark theories.
The ruling wasn't a final determination that LOVO had violated the law. But it illustrated something important: the legal right to reproduce someone's recognizable voice is not necessarily the same as copyright ownership of an audio recording.
This distinction matters to filmmakers using platforms such as ElevenLabs.
A filmmaker may have a commercial license to use a generated voice without owning the underlying voice model. And a platform subscription doesn't automatically establish that every voice available through the service has been cleared for every conceivable use.
The responsible approach is to understand where a voice originated, what rights the platform grants, and whether any identifiable performer's consent is required.
Val Kilmer: A Digital Performance with Permission
A very different example emerged in 2026.
The filmmakers behind As Deep as the Grave announced that the late Val Kilmer would appear in the film through an AI-generated recreation.
Kilmer had originally agreed to play the role but was unable to participate because of his health. After his death, the filmmakers worked with his family and estate to recreate the performance.
The production reported obtaining permission, compensating the estate, and following applicable SAG-AFTRA requirements.
SAG-AFTRA publicly acknowledged the reported family consent while emphasizing that digital replicas of deceased performers must comply with applicable contractual and legal protections.
The important distinction here is authorization.
A filmmaker using a digital replica of a real actor isn't necessarily acting improperly. But the producer must address the performer's rights, consent, compensation, and any applicable contractual obligations.
In Kilmer's case, the technology offered a way to complete a performance the actor had originally intended to give.
Whether audiences accept that performance artistically is another question entirely.
Directing a Performance Versus Owning It
These examples bring us back to the central question of this series: where does human authorship enter AI filmmaking?
A director makes creative decisions about casting, characterization, blocking, emotional delivery, timing, and editing. In AI-assisted filmmaking, many of those decisions are communicated through generative tools rather than directly to human performers.
But directing software doesn't automatically make every resulting image, movement, or vocal inflection copyrightable.
Under the U.S. Copyright Office's current framework, copyright protection depends on identifiable human authorship. Human-written screenplays, original character artwork, and sufficiently creative editing or arrangement may qualify for protection, even when the finished film incorporates AI-generated material.
Copyright, however, is only part of the picture.
The right to use a performance may also depend on platform licenses, performer agreements, rights of publicity, and collective bargaining obligations.
For independent filmmakers hoping to distribute their work, these distinctions matter. A distributor may expect documentation establishing not only copyright ownership but also the legal authority to exploit the performances incorporated into the film.
Protecting Performers While Expanding Filmmaking
I believe synthetic actors will become an increasingly important part of independent filmmaking. I offer no judgment on this, it is inevitable.
They offer opportunities to create original characters and tell stories that might otherwise be economically impossible to produce.
But those opportunities shouldn't depend on appropriating the identities or performances of existing actors without permission.
The practical approach is straightforward: create original characters, understand your platform licenses, document human creative contributions, obtain appropriate consent when identifiable performers are involved, and preserve the records necessary to establish a clear chain of title.
We don't have to choose between protecting performers and embracing new filmmaking technology.
We should be able to do both.
AI may change how a performance is produced. It doesn't eliminate the filmmaker's responsibility for how that performance is created, authorized, and used.
This article is provided for general informational and educational purposes only and does not constitute legal, financial, or professional advice. Laws, licensing terms, and industry practices may change. Filmmakers should consult qualified legal and financial professionals regarding their specific projects.

COMING NEXT — EPISODE 7 - Friday, Novermber 20, 2026
The Stories We Couldn't Afford to Tell: The Economics of AI Filmmaking
For decades, ambitious stories have remained unproduced because the cost of bringing them to the screen was simply too high. But what happens when filmmakers can create extraordinary worlds, enormous sets, and complex productions without the traditional price tag?
In Episode 7, we'll explore how AI-assisted filmmaking is changing the economics of independent production—and opening the door to stories that might otherwise never be told.





Comments