In a professionally lighted brick-walled space in Venice, not different from many other professionally lighted brick-walled spaces all over Los Angeles, actors routinely come in to have their photos and video taken.
The process is quick and unremarkable to anyone familiar with the studio-shoot culture of the city, where the backdrops change but the conventions stay the same.
Yet the similarities with a typical Hollywood shoot end after the camera switches off at the offices of this boutique firm known as Deep Voodoo. The images and video are converted into data bits and sent to AI-model experts employed all over the world. One in Eastern Europe, another in Argentina, a third in Vancouver. They work their machine-training magic, relying on compute from a data center at an undisclosed location. Eventually all that data gets turned into the desired object: a de-aged actor or deepfake or other synthetic image that can used for various forms of entertainment.
All of that would be intriguing even if the founders of Deep Voodoo weren’t South Park instigators Trey Parker and Matt Stone. But among all their taboo-busting, the Book of Mormon pair, it turns out, are also burgeoning AI pioneers. And for the past several years they have been quietly deploying their company to help production labels achieve their effects goals, making housecalls as doctors of deepfakes.
“I find that a lot of discussions about AI become tiresome. You know ‘put your taxes in and it can do them,’” says Stone, 54, in a rare interview about Deep Voodoo. “And it’s like, ‘cool, but a human can do your taxes.’ What we’re trying to do is something no amount of humans can do.”
Something like, well, the more than half-dozen viral projects Deep Voodoo was behind that you may not have even known it was behind. If you’ve watched that Kendrick Lamar music video from a few years ago where the rapper’s face morphs surreally into O.J. Simpson, Will Smith and Jussie Smollett, you’ve already seen its handiwork; ditto that Bill Clinton food-counter eruption in Ted earlier this month, or the Affleck & Co. ’90’s-revisionism for Dunkin’ Donuts at the Super Bowl last month, or the shocking Donald Trump full-frontal deepfake in the Season 27 South Park opener last summer.
But with Generative AI now poised to become a mainstay in Hollywood, Deep Voodoo videos could come at us even more often. If a studio or production house needs something shape-shifting or face-switching, chances are they’ll call Parker & Stone. And chances are something strange — and, perhaps even more surprising when it comes to AI, potentially ethical — will result.
***
Deep Voodoo wasn’t even supposed to be a company.
It only exists because of Donald Trump.
Late in the first Trump administration, Parker and Stone were developing a Donald Trump deepfake movie. Their plan: to graft his face onto another actor’s body and have him progressively lose his marbles, and then eventually his clothing. But the duo couldn’t get a studio to match the quality of the tech they needed. “A couple of effects houses in LA just kind of gave us the runaround. This has happened before in our career, where we go, ‘okay, well, we’ve got to go figure it out ourselves,’” Stone says. So they went online, rounded up some AI whizkids and formed an outfit to do it themselves.
The movie may not have come to fruition — it was scrapped by covid — but the team endured. One product: Sassy Justice, a Web series parodying public figures. A 14-minute episode with a deepfaked Trump went viral. While the visuals and audio seem clunky from the vantage point of 2026, they were downright renegade five years ago — good enough that Parker and Stone even used some of them for the July South Park season opener.
Another result: a full-blown company. By late 2022, Deep Voodoo was so established it had raised $20 million, in part from a CAA-connected venture, before many people thought a lot about AI in Hollywood.
The firm appears engineered to keep a low profile. Its two executives are even-keeled to the extreme. An animation veteran named Jennifer Howell, who once produced South Park and worked at half the studios in town, is its chief content officer, while its CEO, Afshin Beyzaee, is an unflashy lawyer who came to the job after serving for years as chief counsel at Parker and Stone’s Park County production company.
Neither is likely to dazzle (or rage-bait) you with Silicon Valley grandiosities; they are prone to saying strait-laced things like “it’s very inappropriate to be taking and making use of someone’s likeness without their permission,” as Beyzaee did in an interview. Let us blow up the Internet with Tom Cruise-Brad Pitt fights this is not.
But such hall-monitor vibes are how Parker and Stone like it. If you’re going to use generative technology that already has everyone all touchy, better to be the Mr. Mackey of the startup world.
Deep Voodoo prides itself in particular on licensing. The company won’t work with any studio that hasn’t obtained authorization from actors or estates (it didn’t have permission from the White House for that deepfake of Donald Trump last summer, but execs says they relied on fair-use images of the ubiquitous president).
“You’ve got a situation that some are paying to use or license IP and some are not and then people say ‘why should I pay for it?’” Beyzaee says. “To us, it’s this is about making sure that we’re providing this service, providing this technology, in a way that respects the laws and the protections and the rights that people have.” The company’s principals, Beyzaee says, will turn down jobs if they’re not satisfied with the level of footage-permissions that the studio or production company has received.
In that idea, perhaps lies the tantalizing contradiction at the heart of Deep Voodoo: the most subversive creators out there are trying to be the good guys of AI.
Instead of scraping the Web for images (or relying on a model that has done it), Deep Voodoo takes those licensed images either in its brick-walled space or those provided by a production company. The capture at the spaces involves nine cameras and a series of simple questions to elicit a range of facial reactions. It then then uses all the material to build a bespoke model for a particular production. This is a laborious process, especially for a one-off usage — a scale-minded Silicon Valley firm would have no truck with it — and can take up to a month and involve some 300,000 images. Still, what results is both something legal and suited to the case at hand. “It’s important that this isn’t us just going and scouring the Internet for materials and building it into our models,” Beyzaee says.
When the company de-aged (and then re-aged) Billy Joel in 2024 for the video to his comeback single “Turn The Lights Back On” — viewers re-lived the many stages of Billy as the lyrics took their own rueful look back — the cross-cuts between the decades were seamless because the video’s producers had something designed just for them, not plucked from Google and jammed into their workflow.
“Our goal is to make beautiful, cinematic film and television that that never pulls the viewer out because the effect doesn’t look right,” Howell says, attributing the mission to how “we were started by artists who are incredibly picky.”
The Deep Voodoo crew is aware of how much skepticism many actors and writers harbor about AI but believe that much of that ire should be directed at prompt-based materials, which often aim to conjure up content without an artist at the center or even the controls. Their own work, reality-bending as it is, stands apart, they say, because human actors are usually performing under what amounts to a facial mask. Stone notes how different the company’s M.O. is from the Tilly Norwood-style synthetic approach that has rattled so much of the creative community. “We’re not doing anything like that. We’re not typing in a prompt. It’s all capturing actors doing what they do,” he says.
He adds, “I mean, the magical part of the production is the puppeteer, right? The puppet is one thing — and the tools can create a great puppet. But the magic is the performer. Without that is just becomes wallpaper.”
***
Even with all those official permissions, Stone believes AI can be used for purposes we’ve only begun to imagine.
“Somebody’s going to make a scary fucking horror movie using this technology. Somebody’s going to make a really fucking funny comedy using this. Like really funny shit that couldn’t be made — that’s native,” he says. “What’s going to happen soon — what we’re capable of — is someone’s going to do a political show. Something very current, and they’ll use deepfakes not to look exactly like the person, although that’s possible, but to make some sort of grotesque mashup — to capture a bit of them in this phantasmagoric kind of thing. You’ll be able to do it as a weekly or bi-weekly show on an SNL-type schedule.”
De-aging has been a major use case for Deep Voodoo, and Stone says he sees it continuing. But he and Howell also point to a whole new realm, called “performance transfer,” that allows an actor to play their role in street clothes on a stage with only minimal shooting on location; the performance is then “transferred” so that it appears the actor was with the cast running through the streets of Paris or engaging in an intense showdown in Beijing, a kind of three-dimensional ADR.
That has the chance to speed up, and dollar-down, authentic-seeming production in ways previously unfathomable; the idea of shipping a bunch of stars and large crews to Europe or Asia to shoot an action movie may soon seem as antiquated as drawing an entire animated feature with your hand. “I really do feel like it’s going to touch on a lot of aspects of how things get produced,” Stone says.
That kind of advance won’t make physical crews or the places trying to lure them very happy, a point Stone concedes. And the deepfake usages, while clearly labeled as satire when Deep Voodoo does it, can still contribute to a culture of mistrust online.
But those downsides will come with an overwhelming number of benefits, Stone believes. And in any case, he says, while guardrails should be constructed, barriers can’t be. “This stuff is happening. We’ve already all watched stuff on TV that has utilized machine learning. It’s happening, and it is going to change the industry.”
And the big unspoken question — using it in South Park? Stone believes he and Parker will, and it could change the results on-screen.
“We’re doing [the show] every two weeks now. That has more to do with our age than technology, but the tech means that maybe we go home earlier, maybe we’ve got more options,” he says. “And it means maybe the show’s better.”
This story appears in The Hollywood Reporter’s upcoming AI Issue, out in April.
