I made it out to the cinema recently and checked out Morgan Neville’s new documentary, Roadrunner , about the life, career and death of Anthony Bourdain. The two hour plus chronicle of the celebrity chef was compelling enough and a decent watch. Afterwards, I was shocked to find out that the voice I heard was a deepfake. Well, three separate narratives within the film.
In an interview with the New Yorker, Mr Neville said there were quotes that he wanted Bourdain to narrate. However, there were no recordings. The director went out to add that
“…It was a modern storytelling technique that I used in a few places where I thought it was important to make Tony’s words come alive.”
The AI used for this was scary good. I mean, I am no Bourdain scholar. I casually caught episodes of Parts Unknown and No Reservations here and there but it sounded A LOT like him. We’re only at the beginning of the age of the deepfake. However, doesn’t this move pretty much void the use of the category “documentary”? Isn’t there some kind of code of ethics directors should follow? The fact this is fake and admitted by the director, would make it work of fiction.
I’m sure from a storytelling standpoint it would be better to hear these words in Bourdain’s voice. But when a documentary needs to read something that was written by someone who’s passed away, don’t they normally just have a narrator read it? I can’t think of an example off the top of my head, but I feel like that’s kind of a standard practice, and that viewers understand the context of the words being written. This is a technique that works just fine even if it’s not 100% ideal from a storytelling perspective. Again, the words weren’t faked: the sound of him speaking them was.
The biggest problem isn’t using the AI dialogue (after all, a narrator or actor hired to read these letters is also going to interpret them, adding inflection that isn’t necessarily Bourdain’s), it’s using AI dialogue in a film that has actual recordings of Bourdain’s voice, and not distinguishing between the two. This guy wants a pat on the back for his ingenuity and for the fact that he doesn’t think anyone can tell the difference, when that’s the danger. It wouldn’t be difficult (or make the dialogue less “alive”) to put a chyron with “AI narration” on screen, so that the audience knows when Bourdain is speaking for himself, and when it’s Siri reading stuff Bourdain wrote in Bourdain’s voice. Otherwise, it’s like trying to pass off a recreation as archival news footage. No one cares how skillfully an FX team reproduced tracking errors to make it look like 80s-vintage news camera footage, they care more that it’s a lie.
In the same interview, Neville added “We can have a documentary-ethics panel about it later.” I think this is a great idea. When an artist is presenting something as factual, the audience has a right to know what’s genuine and what’s been created. Once the viewer knows it’s fake, they can admire the technical skill involved without being deceived.