New York-based Italian director Francesco Carrozzini is heading to the Venice Film Festival with “Be Brave” an 85-minute portrait of Renzo Rosso, founder of the Diesel fashion brand. The groundbreaking doc uses AI in a way that “can still live within a space of emotion and storytelling,” he says. “Be Brave” portrays Rosso as a visionary underdog who went against the grain and rose from poverty to owning a fashion empire. Carrozzini did not shoot a single frame of his protagonist, instead recreating Rosso at different stages of his life using AI.
“I want to put this [film] forward as an invitation for everyone to think about how AI as a tool can open up new creative possibilities for filmmakers,” Venice chief Alberto Barbera stated when announcing its inclusion in the festival. “Be Brave” is among the first AI features to launch from a top-tier film festival.
Popular on Variety
Carrozzini was first in Venice with “Franca: Chaos and Creation,” the doc he made about his mother, the late, influential editor of Italian Vogue. The director was subsequently on the Lido in 2022 with his first feature “The Hanging Sun.”
In the year-and-a-half that Carrozzini spent making “Be Brave” “AI has changed to a point where what we were doing in the beginning had to be redone,” he says. “And the moment we will be in the theater it will probably be old already,” he adds. This makes it hard. But “Working with a medium that is transforming at this pace is also really exciting,” he notes.
“Be Brave” is produced by Lorenzo Mieli, Mario Gianani, and Gabriele Immirzi at Rome-based Our Films, a Mediawan company.
Carrozzini speaks to Variety ahead of Venice about the challenges of making “Be Brave” and his thoughts on how AI can best be utilized in the filmmaking space.
Why did you decide to use AI for this project?
Lorenzo Mieli approached me with the idea of making a film about Renzo Rosso. Initially, I was a bit reluctant because I had sort of taken a step back from the world of fashion that I grew up in to throw myself into narrative film and TV. But Lorenzo insisted, and when he wants to sell you an idea he’s pretty effective. The idea was to create something a bit innovative and to experiment with AI, though it wasn’t quite clear what we were going to do. So we started testing the idea. In the very beginning, it was really about necessity. There is a lot of archival like TV stuff of Renzo Rosso, but there’s no good interviews. There are no really good intimate moments in his archives. So the idea became: first of all, we can recreate those from scratch; though we were worried about them not being believable. From the seed of that idea, then came the idea of: Well, let’s take AI and make it sort of the theme of the film and also make it the challenge towards something bigger than ourselves.
What was the next step?
We wrote a script. This film is fully scripted. We started toying with the idea of me finding Renzo; a rather well-known device in documentary filmmaking. But at the same time putting together the puzzle of who he is, just as he was running away from me because he was trying to trick me. We used this device, but then it sort of became the theme of the film. This film is not a thesis on AI. The only thing that is said towards the end – which I think is really important – is that the human has to come first. This is a reflection on AI in storytelling and entertainment. It’s not a reflection on AI tout court.
Let’s drill down on the nuts and bolts of making this piece. It’s not all AI because there are some so-called talking heads in the doc that are real people. For the rest, you had a reservoir of archive material to draw on. Talk to me about your workflow.
It was a nightmare, especially in the beginning, until we found the workflow, as you say. Because it’s trial and error. As you know, AI will never give you exactly what you’re asking. It will always give you a version of what you’re asking.
How did this process apply to putting Renzo on screen?
We used pictures of Renzo from the time [in which his scenes and interviews are set]. And then you have to have him speak. And you have to have him say what you want. And do it in the voice that you want. He also has to be dressed the way you want. Just boiling things down to that, and to the voice and the lip sync is painstaking. Regarding Renzo’s voice in this film, it had to have different ages as well, so that it would really be as real as possible. The first task is getting what you want.
And beyond that?
Trying to capture not only his physical age, but his behavior. For people who don’t know this process, behavior is really, I would say, the hardest thing. You can get the voice perfect. You can get the image right. But it’s the behavior that was really challenging in creating quite a lot of footage of Renzo.
What was it like for you to be directing yourself (as you appear in the film) done in AI?
Well, it’s all based on my pictures, and because it’s me, it’s the part that scared me the most. With Renzo sometimes I would think I saw a tiny discrepancy. But with me, I was like: “Oh my god, that’s real!”
Back to the workflow. Can you describe it in detail?
All this becomes a workflow problem. Labeling all this material as if it was footage, and making it accessible to the editor. Blackstone was one of the companies behind a big part of it. And then this genius, Roberto Beragnoli who was the AI artist, did the heavy lifting. It was really months and months of finding a way to get all that we wanted, in the order that we wanted it.
It looks like just a documentary. But the work that went into it is like the work of a big, heavy CGI feature film. The quality part was a big challenge, because there are so many different software tools that we used to create the material. In AI you work at very low quality, and you have to upscale it. It’s very technical. I don’t want to get too nerdy, but the whole process is quite nerdy.
You’ve talked to me about creating a person with AI. How do you create a place with people in it?
That was really a very exciting filmmaking endeavour because with anything I do that is scripted, I have to come up with you know, the same exact number of [visual] ideas. If the scene says: “Off Broadway theater. Morning. Ten people waiting in line.” I have to decide how do those ten people look? How are they dressed? Where exactly are they? In the beginning it was very frustrating because instead of casting and scouting and talking to a costume designer, you do it with AI.
Do you think it’s important for the viewer of this piece to know when they are watching AI and when they are not?
I want to say no. This film tells a story of a man who lived, who is real, and whose adventures are all quite true. And so, if we are able to entertain you, make you laugh and empathize with this character, then that’s what counts. What interests me is not whether they’re able to discern it or not, but whether the film has an effect.
One of the fears concerning AI in fiction filmmaking is that the medium of cinema could lose the value of the empathic connection people have with other humans. What are your thoughts about that?
That’s why I was interested in approaching AI through the medium of documentary, which is very different from fiction, although ours is a fictionalized documentary, because it’s written and created. It still has the form of the documentary, it has that filter that you put on when you watch it. If this was that beautiful documentary “Free Solo” [which captured rock climber Alex Honnold’s hair-raising ascent of Yosemite National Park’s 3,000-foot El Capitan rock formation] that would be a very different thing because we would be pretending that something that never happened had happened.
How would you feel about directing a fiction film in AI?
I think something similar is happening with scripted material in fiction. With TV shows and movies. We haven’t overcome that yet. And I’m not sure that we want to. The empathy you talk about, it’s generated by real people doing things that, despite being more or less made up, are still human. That emotion is human. That cry is human. That laughter is human. Once you know that it’s not human – and right now we can tell, even if we’re not told – the moment you start watching, after five minutes you know that you’re not going to get on board exactly because that empathy is lacking. Again, I’m not saying that we fully succeeded. People will tell us [whether we did or not]. But we tried to apply a form of AI that can still live within a space of emotion and storytelling, I think. Because you can’t discern it, it still works for its purpose. Right now I would not necessarily take on a project, such as a historic film or a fiction film made in the present day, where I don’t use actors.
This interview has been edited and condensed for clarity.