AI documentaries won’t let the dead rest

Please, for the love of the visual arts, don't make your film look like a sloppy Facebook feed.

AI documentaries won’t let the dead rest

We are reaching AI saturation. Small restaurants lean on generative AI to create generic menus, neighborhood flyers regularly use its weirdly soulless fonts and graphics to advertise events, and social media feeds and search results have made it nearly inescapable. It’s also now fully encroached upon the documentary space, and it’s being used in more dubious and questionable ways that will mar at least a generation’s worth of nonfiction movies. 

One of the first major uses of machine learning in a documentary was in David France’s 2020 film Welcome To Chechnya. The nascent tech hid the identities of queer people from the region’s homophobic regime by swapping out their faces with those of U.S. activists. Since then, the technology has improved by many leagues, but filmmakers aren’t just using AI tools for subject safety. Now, generative AI is being used to jiggle the past to make the dead look like they’re in motion or pack the frame with superfluous slop, assuming audiences won’t look at the screen if there’s not constant movement on it.

When Ava DuVernay’s 14th premiered at the New York Film Festival, it followed a similar format as her previous Constitutional film 13th. Like her earlier film, the documentary is filled with engaging expert lecturers, historians, and archival footage. This time, though, DuVernay added a little extra to the archival material, using generative AI to give the illusion of a three-dimensional image. Not great, kind of uncanny, and still phony to look at, but it’s nowhere near as distracting or baffling as when 14th goes into a montage of historical footage where the photo’s subjects are made to move through AI. Generously, it looks like the magical photography in Harry Potter, where figures within the frame repeat certain movements like a GIF. More realistically, it’s reminiscent of when the Star Wars franchise exhumed the likeness of Peter Cushing to reprise his role in Rogue One, decades after his death. 

At least the Cushing estate gave Disney its permission. But with or without permissions, the effect of animating the dead still looks ghoulish. Documentaries have a long history of historical recreations, and those moments in 14th, mostly of historical Black social clubs and school photos, could have easily been staged. DuVernay could’ve also just accepted using the photographs her team had uncovered as the lasting records of these events. By cheaping out on hiring actors to bring these historic moments to life, insisting that every static photo must have movement in order to capture the audience’s attention, and resorting to generative AI to ghoulishly resurrect its subjects, 14th talks down to its audience—and the quality of the resulting film looks like something you’d scroll past on a Facebook feed, not expect from an Oscar nominee.

Similarly, Matthew Miele’s new documentary Vincent Price: What Keeps Us Alive also relies on AI to “jazz up” its biographical look at the actor, art connoisseur, and chef, but without reanimating his corpse. Instead, the doc utilizes AI to force artificial movement in the “world” around the archival pictures and video clips. Between the excellent interviews with famous Price fans like Bill Hader, Tim Burton, John Waters, and Patton Oswalt—as well as members of Price’s family like his children, Vincent and Victoria—Miele can’t leave AI alone long enough to let the clips of Price’s work, interviews, and appearances speak for themselves. What Keeps Us Alive is, ironically, constantly interrupting them with lifeless AI recreations of theaters, hometowns, beaches, and backstages, none of which have any visual cohesion in quality or color grading. 

It would simply be embarrassing if it wasn’t equally exasperating. Clips of Price’s movies are screened in empty theaters that don’t look like they exist in the same universe, because they’re all AI-generated images that make the film segments look even smaller in their cavernous frames. Great archival photos are skewed as if they were hung on walls, so once again, it hinders the audience from enjoying the actual details of one of Hollywood’s most beloved horror stars. Entire cities are rendered from AI, because towns like St. Louis, London, and Burbank famously never had any photos taken of them or stock footage shot of them. These janky decisions (and weird mistakes, like a bowtie that stands up on a desk by itself) make Vincent Price: What Keeps Us Alive as ugly and dated as in-camera VHS effects of the ’80s and ’90s dated amateur videos of their time. Outsourcing this much of your film’s visual identity to AI doesn’t just make for a bad movie, it renders it almost already obsolete.  

In these cases, and likely many more still to come, AI doesn’t enhance the sound or smooth the viewing experience, it actively distracts from it. Using this much generative AI can also cause the documentary to lose the audience’s trust, as they must now scrutinize and verify every image it’s trying to pass off to them. Seeing the same kinds of cheap AI parlor tricks used to manipulate older relatives on the internet deployed in documentaries is a dismaying trend, but one that’s going to likely only get worse as more ethically dubious filmmakers become enamored of the financial shortcuts and hollow look of its quick effects.

 
Join the discussion...
Keep scrolling for more great stories.