Hollywood’s obsession with AI-enabled ‘perfection’ is making movies less human

The notion of authenticity in the movies has moved a step beyond the merely realistic. More and more, expensive and time-consuming fixes to minor issues of screen realism have become the work of statistical data renderings—the visual or aural products of generative artificial intelligence. Deployed for effects that actors used to have to create themselves, with their own faces, bodies, and voices, filmmakers now deem these fixes necessary because they are more authentic than what actors can do with just their imaginations, wardrobe, makeup, and lighting. The paradox is that in this scenario, “authentic” means inhuman: The further from actual humanity these efforts have moved, the more we see them described by filmmakers as “perfect.”

Is perfect the enemy of good? It doesn’t seem to matter to many filmmakers working today. These fixes are designed to be imperceptible to humans, anyway. Director Brady Corbet’s obsession with “perfect” Hungarian accents in his Oscar-nominated architecture epic, The Brutalist, is a case in point. Corbet hired the Ukraine-based software company Respeecher to enhance accents by using AI to smooth out vowel sounds when actors Adrien Brody and Felicity Jones (American and British, respectively) speak Hungarian in the film. Corbet said it was necessary to do that because, as he told the Los Angeles Times, “this was the only way for us to achieve something completely authentic.”

Authenticity here meant integrating the voice of the film’s editor, Dávid Jánsco, who accurately articulated the correct vowel sounds. Jánsco’s pronunciation was then combined with the audio track featuring Brody and Jones, merging them into a purportedly flawless rendition of Hungarian that would, in Corbet’s words in an interview with GQ, “honor the nation of Hungary by making all of their off-screen Hungarian dialogue absolutely perfect.”

The issue of accents in movies has come to the fore in recent years. Adam Driver and Shailene Woodley were, for instance, criticized for their uncertain Italian accents in 2023’s Ferrari. Corbet evidently wanted to make sure that would not happen if any native Hungarian speakers were watching The Brutalist (few others would notice the difference). At times, Brody and Jones speak in Hungarian in the film, but mostly they speak in Hungarian-accented English. According to Corbet, Respeecher was not used for that dialogue.

Let’s say that for Corbet this will to perfection, with the time and expense it entailed, was necessary to his process, and that having the voice-overs in translated Hungarian-accented English might have been insultingly inauthentic to the people of Hungary, making it essential that the movie sound, at all times, 100% correct when Hungarian was spoken. Still, whether the Hungarian we hear in The Brutalist is “absolutely perfect” is not the same as it being “completely authentic,” since it was never uttered as we hear it by any human being. And, as it turns out, it was partially created in reaction to something that doesn’t exist.

In his interview with the Los Angeles Times, Corbet said that he “would never have done it any other way,” recounting when he and his daughter “were watching North by Northwest and there’s a sequence at the U.N., and my daughter is half-Norwegian, and two characters are speaking to each other in [air quotes] Norwegian. My daughter said: ‘They’re speaking gibberish.’ And I think that’s how we used to paint people brown, right? And, I think that for me, that’s a lot more offensive than using innovative technology and really brilliant engineers to help us make something perfect.”

But there is no scene in Alfred Hitchcock’s 1959 film North by Northwest set at the United Nations or anywhere else in which two characters speak fake Norwegian or any other faked language. Furthermore, when Corbet brings in the racist practice of brownface makeup that marred movies like 1961’s West Side Story, he is doing a further disservice to Hitchcock’s film. The U.N. scene in North by Northwest features Cary Grant speaking with a South Asian receptionist played by Doris Singh, not an Anglo in brownface.

Corbet’s use of AI, then, is based on something that AI itself is prone to, and criticized for: a “hallucination” in which previously stored data is incorrectly combined to fabricate details and generate false information that tends toward gibberish. While the beginning of Hitchcock’s Torn Curtain (1966) is set on a ship in a Norwegian fjord and briefly shows two ship’s officers conversing in a faked, partial Norwegian, Corbet’s justification was based on a false memory. His argument against inauthenticity is inauthentic itself.

AI was used last year in other films besides The Brutalist. Respeecher also “corrected” the pitch of trans actress Karla Sofía Gascón’s singing voice in Emilia Pérez. It was used for blue eye color in Dune: Part Two. It was used to blend the face of Anya Taylor-Joy with the actress playing a younger version of her, Alyla Browne, in Furiosa: A Mad Max Saga. Robert Zemeckis’s Here, with Tom Hanks and Robin Wright playing a married couple over a many-decade span, deployed a complicated “youth mirror system” that used AI in the extensive de-agings of the two stars. Alien: Romulus brought the late actor Ian Holm back to on-screen life, reviving him from the original 1979 Alien in a move derided not only as ethically dubious but, in its execution, cheesy and inadequate.

It is when AI is used in documentaries to re-create the speech of people who have died that is especially susceptible to accusations of both cheesiness and moral irresponsibility. The 2021 documentary Roadrunner: A Film About Anthony Bourdain used an AI version of the late chef and author’s voice for certain lines spoken in the film, which “provoked a striking degree of anger and unease among Bourdain’s fans,” according to The New Yorker. These fans called resurrecting Bourdain that way “ghoulish” and “awful.”

Audience reactions like these, though frequent, do little to dissuade filmmakers from using complicated AI technology where it isn’t needed. In last year’s documentary Endurance, about explorer Ernest Shackleton’s ill-fated expedition to the South Pole from 1914 to 1916, filmmakers used Respeecher to exhume Shackleton from the only known recording of his voice, a noise-ridden four-minute Edison wax cylinder on which the explorer is yelling into a megaphone. Respeecher extracted from this something “authentic” which is said to have duplicated Shackleton’s voice for use in the documentary. This ghostly, not to say creepy, version of Shackleton became a selling point for the film, and answered the question, “What might Ernest Shackleton have sounded like if he were not shouting into a cone and recorded on wax that has deteriorated over a period of 110 years?” Surely an actor could have done as well as Respeecher with that question.

Similarly, a new three-part Netflix documentary series, American Murder: Gabby Petito, has elicited discomfort from viewers for using an AI-generated voice-over of Petito as its narration. The 22-year-old was murdered by her fiancé in 2021, and X users have called exploiting a homicide victim this way “unsettling,” “deeply uncomfortable,” and perhaps just as accurately, “wholly unnecessary.” The dead have no say in how their actual voices are used. It is hard to see resurrecting Petito that way as anything but a macabre selling point—carnival exploitation for the streaming era.

Beside the reanimation of Petito and the creation of other spectral voices from beyond the grave, there is a core belief that the proponents of AI enact but never state, one particularly apropos in a boomer gerontocracy in which the aged refuse to relinquish power. That belief is that older is actually younger. When an actor has to be de-aged for a role, such as Harrison Ford in 2023’s Indiana Jones and the Dial of Destiny, AI is enlisted to scan all of Ford’s old films to make him young in the present, dialing back time to overwrite reality with an image of the past. Making a present-day version of someone young involves resuscitating a record of a younger version of them, like in The Substance but without a syringe filled with yellow serum.

When it comes to voices, therefore, it is not just the dead who need to be revived. Ford’s Star Wars compatriot Mark Hamill had a similar process done, but only to his voice. For an episode of The Mandalorian, Hamill’s voice had to be resynthesized by Respeecher to sound like it did in 1977. Respeecher did the same with British singer Robbie Williams for his recent biopic, Better Man, using versions of Williams’s songs from his heyday and combining his voice with that of another singer to make him sound like he did in the 1990s.

While Zemeckis was shooting Here, the “youth mirror system” he and his AI team devised consisted of two monitors that showed scenes as they were shot, one the real footage of the actors un-aged, as they appear in real life, and the other using AI to show the actors to themselves at the age they were supposed to be playing. Zemeckis told The New York Times that this was “crucial.” Tom Hanks, the director explained, could see this and say to himself, “I’ve got to make sure I’m moving like I was when I was 17 years old.”

“No one had to imagine it,” Zemeckis said. “They got the chance to see it in real time.” No one had to imagine it is not a phrase heretofore associated with actors or the direction of actors.

Nicolas Cage is a good counter example to this kind of work, which as we see goes far beyond perfecting Hungarian accents. Throughout 2024, Cage spoke against AI every chance he got. At an acceptance speech at the recent Saturn Awards, he mentioned that he is “a big believer in not letting robots dream for us. Robots cannot reflect the human condition for us. That is a dead end. If an actor lets one AI robot manipulate his or her performance even a little bit, an inch will eventually become a mile and all integrity, purity, and truth of art will be replaced by financial interests only.”

In a speech to young actors last year, Cage said, “The studios want this so that they can change your face after you’ve already shot it. They can change your face, they can change your voice, they can change your line deliveries, they can change your body language, they can change your performance.” And he said in a New Yorker interview last year, speaking about the way the studios are using AI, “What are you going to do with my body and my face when I’m dead? I don’t want you to do anything with it!” All this from a man who swapped faces with John Travolta in 1997’s Face/Off with no AI required—and “face replacement” is now one of the main things AI is used for.

In an interview with Yahoo Entertainment, Cage shared an anecdote about his recent cameo appearance as a version of Superman in the much-reviled 2023 superhero movie The Flash. “What I was supposed to do was literally just be standing in an alternate dimension, if you will, and witnessing the destruction of the universe. . . . And you can imagine with that short amount of time that I had, what that would mean in terms of what I could convey—I had no dialogue—what I could convey with my eyes, the emotion. . . . When I went to the picture, it was me fighting a giant spider. . . . They de-aged me and I’m fighting a spider.”

Now that’s authenticity.