In the second piece I ever wrote for the Prindle Post, I talked about the celebrity chef Anthony Bourdain. More specifically, I talked about the use of AI in the documentary film Roadrunner.
In the film, and prominently in its trailer, AI was used to recreate Bourdain’s voice. That synthetic voice was then used to bring to life emails that Bourdain had written. The filmmakers could not, of course, ask Bourdain to read the emails himself. On June 8th, 2018, three years before the film’s release, Bourdain died by suicide while in France filming his travel and food program Parts Unknown.
The problem, as I argued at the time, was that the use of AI was not disclosed to the audience. Indeed, the only reason we know about it is that the director revealed it during an interview. This opened an entire can of worms concerning whether we can trust what we see and hear on screen, even when what we are watching is a documentary and therefore, one assumes, the genuine article. You can read the piece here.
That was five years ago. Unsurprisingly, as AI has grown more powerful and pervasive, an ever-increasing number of ethical issues have emerged. Yet the problem of using AI to mimic a person, perhaps eventually with perfect fidelity, remains. Deepfakes are no longer curiosities. They have become tools that can be deployed at scale, whether for entertainment, marketing, fraud, or political manipulation.
This topic sits outside my usual intellectual haunts of medical ethics and law. Nevertheless, it is one I keep finding myself returning to. And it was the death of another counterculture icon that brought me back to it once again: Tim Curry.
Curry had a long, illustrious, and yet I think understated acting career. He played Nigel Thornberry in The Wild Thornberrys. He played the menacing clown Pennywise in the 1990 IT mini-series. He was Long John Silver in Muppet Treasure Island, a performance so beloved that he was eventually awarded the incredibly rare distinction of honorary Muppet. And while there are many other roles with which he is associated, for me, he will always be the mad scientist Frank-N-Furter from the Rocky Horror Show, and the subsequent film, the Rocky Horror Picture Show (in my opinion, the greatest film ever made – although not for the faint of heart).
In 2012, Curry suffered a severe stroke that left him paralyzed. In the immediate aftermath, he was unable to speak and underwent extensive therapy to regain his voice. He largely succeeded, but his speech changed. As Curry himself acknowledged, the paralysis affecting one side of his face meant he no longer sounded quite the same as he once had.
It was his death, then, that prompted me to listen to the audiobook of his memoir, Vagabond, which he recorded in 2025. Being the fool that I am, however, when I hit play, I was surprised, just for a moment, that Curry’s voice wasn’t that of Frank-N-Furter. And, obviously, it wouldn’t be. Not only because Curry played that role, in the film, in 1975, 50 years before recording the audiobook, but also because that was a role, and along with the costume and persona came the iconic voice. Even though that is what he sounds like, as a default, in my head because that’s the role I associate with him, that wasn’t who he was or what he sounded like (obviously when you write it down).
Yet the experience got me thinking. My momentary disappointment was irrational, but it revealed something interesting: I wanted Curry to sound less like himself and more like the version of him that existed in my memory. I wondered whether someone, somewhere, would eventually take Curry’s memoir and feed it through AI.
While this might sound like an odd thought experiment dreamt up during a walk in the park, which it was, the underlying idea is not new.
Earlier this year, NBC reported that actor Michael Caine licensed his voice so that the AI company ElevenLabs could create a synthetic version of it. The result was an audiobook of the Odyssey, narrated not by the older Michael Caine of today, but by a recreation of the younger Caine that many listeners remember from film.
This may seem harmless enough. After all, if Caine consents, which he did, and listeners enjoy the experience, where is the problem?
Well, I think there is a problem. In fact, I see two clear issues.
The first concerns posthumous consent.
Curry had died. Unlike Caine, he would therefore not be in a position to agree to the creation of a synthetic version of his voice. Nor could he specify how that voice should be used, which periods of his life might be reproduced, or which of his characters might be permitted to narrate his story. We would be recreating the likeness of someone who had died without knowing whether they would have approved.
Having celebrities feature in productions after their deaths is, of course, nothing new.
Peter Cushing appeared digitally in Rogue One. Harold Ramis appeared seven years after his death in Ghostbusters: Afterlife. A digital Audrey Hepburn appeared in a Galaxy chocolate advert way back in 2012. Heck, The Simpsons even made a joke about this way back in 1998, when Homer talks about celebrities being owned by the public and that, after their death, they’ll all star in commercials dancing with vacuum cleaners. Indeed, the basic idea of using technology to place dead celebrities on screen is well worn.
Despite how frequently posthumous recreations now occur, there remains something disrespectful about placing the image, voice, or persona of a deceased person into a context they never chose. That concern becomes more acute when the recreation is not merely being used to perform another fictional role.
A synthetic Curry narrating Vagabond would be used to tell Curry’s own story.
That compounds the problem. A digitally recreated actor appearing in a fictional film raises questions about consent, ownership, and commercialization. A digitally recreated actor narrating his own autobiography raises questions about authenticity itself. The technology would not merely be appropriating Curry’s voice. It would be appropriating his authority to speak about his own life.
The second problem concerns disability, ageing, and the symbolic meaning of replacing Curry’s actual performance.
We already possess a version of Vagabond narrated by Curry himself. It exists. The appeal of creating a synthetic alternative would rest largely on the belief that some listeners might prefer an earlier version of his voice or the stylized voice of one of his characters.
Yet that preference carries normative weight. Curry’s altered speech was a consequence of his life. His voice bore the marks of ageing, illness, survival, and extensive rehabilitation. Those marks were part of the person telling the story.
Replacing that performance with a reconstruction of the voice Curry once had would therefore not be a neutral act. It would suggest that the older Curry was somehow less authentic, less capable, or less worthy of being heard than the person many of us remember from our younger years.
The technology would not simply preserve a voice. It would select one version of a person over another and, in doing so, implicitly judge which version was worth hearing.
Now, you could argue that I’m coming on too strong here. That perhaps listeners who chose the synthetic version would not be making a judgment about Curry’s capacity to recount his own life. They might simply enjoy the voice of Frank-N-Furter and want to hear more of it. Their choice might be motivated by nostalgia rather than disrespect.
And yeah, sure. That might be true. Nevertheless, our actions express unintended meanings. Choosing the synthetic performance over Curry’s own narration would still involve turning away from the voice with which he chose to tell his story. Whatever an individual listener intended, the existence and commercial promotion of the alternative would communicate that the genuine article was subpar.
Perhaps the reproduction would sound smoother. Perhaps it would sound more familiar. Perhaps it would sound more like the Tim Curry that audiences wanted to remember.
But it would not be Tim Curry narrating his life.
If we can set the ethicists aside for a moment and adopt a business, dare I say capitalist, mindset, it is easy to see the commercial argument. Where actors or public figures possess iconic voices, audiences might reasonably want to hear autobiographies narrated in those voices. A company able to offer that experience would almost certainly find willing customers.
Suppose, for example, that an autobiography by James Earl Jones could be narrated in the voice of Darth Vader or Mufasa. Listeners could choose between the actor’s natural voice and the characters by whom they remember him. The appeal is obvious.
The model could be taken even further for performers associated with numerous voices. Consider Dan Castellaneta, who has voiced many characters in The Simpsons, including Homer Simpson, Krusty the Clown, Barney Gumble, Mayor Quimby, and others. An audiobook of his life could be sold with a selection screen allowing listeners to choose which character would narrate it. One person might want Homer. Another might prefer Krusty. A particularly adventurous listener could assign different characters to different chapters. Again, I can see the appeal.
There is, admittedly, a difference between this scenario and using AI to replace Curry’s narration of Vagabond. Castellaneta might consent to such a project, participate in its design, and receive payment for the use of his performances. If so, the ethical objections concerning posthumous consent are less forceful.
The examples, however, sit in the same sphere. Both alter the relationship between performers, characters, and audiences. Both invite us to treat a voice as something detachable from the person who produced it. Both suggest that a performance can be extracted, preserved, and made anew indefinitely.
Now, maybe my concerns will eventually be swept aside by consumer demand. Perhaps audiences will embrace the ability to hear autobiographies narrated by characters rather than people, choosing Homer Simpson over Dan Castellaneta, Darth Vader over James Earl Jones, or Frank-N-Furter over Tim Curry. If so, the technology may prove enormously profitable.
There remains, however, at least from where I sit, something unsettling about the prospect. Fictional characters can become important to us. They may shape our identities, provide comfort, or become entwined with significant moments in our lives. But the affection we feel for those characters should not cause us to overlook the people who brought them into existence.
Voices change. Bodies change. We age, become ill, and acquire histories that leave their mark. Those changes are not flaws to be corrected. They are part of what makes a life a life. And in using AI to erase those imperfections, we may gain a more familiar version of a celebrity. What we lose is the person.