16 July 2024
Exclusives, Television/ Streaming
By TREVOR HOGG
By TREVOR HOGG
Protecting the identities of sources is an integral part of investigative journalism, with the usual techniques being blurring faces, shooting interviewees as silhouettes and modulating voices. HBO Documentary Films’ Welcome to Chechnya, which exposes the systematic persecution of the LGBTQ community in Chechnya, introduces an innovative approach combining machine learning and face replacements that earned a nomination for Outstanding Supporting Visual Effects in a Photoreal Feature at the VES Awards, and became the first documentary to make the Oscar shortlist for Best Visual Effects.
“There were so many challenges,” states filmmaker David France, who received an Academy Award nomination for How to Survive a Plague. “The first was to talk to the people who needed to be anonymous about my desire to film them. We came up with a strategy working together about how to protect the digital files and how to move them safely out of the country and how to erase the cards from the cameras. We were overriding those cards so in case we lost control of one of those, data couldn’t be harvested. Then I talked with the folks about wanting to find some way to disguise [people to be filmed anonymously] that doesn’t distract from their humanity. I didn’t know what I was talking about, but I was honest with them about not knowing what I was talking about. We agreed that we would come back to them. They gave me a conditional release. I would come back to them for these security reviews, and they would ultimately give the sign-off. I left with the footage well-protected and began that work once we got back to New York.”
The process of discovery began when producer Alice Henty (Buck) called veteran visual effects supervisor Johnny Han (The Nevers), who became the Facial Capture Supervisor on the project. “I said to Alice, ‘There are things being done now in the computer graphic community where we rely on whole other processes such as machine learning to put the likeness of one person’s face onto another,’” states Han. “I phoned Ryan Laney [Spider-Man 3] and asked him if he wanted to help me.” Capturing every possible facial angle and distance was not possible. “When watching the film and breaking down the shots, we realized that there was a certain comfortable distance for the average person when filming someone with their iPhones,” adds Han. “We decided upon two different lengths: a macro and a wide.”
“There were so many challenges. The first was to talk to the people who needed to be anonymous about my desire to film them. We came up with a strategy working together about how to protect the digital files and how to move them safely out of the country and how to erase the cards from the cameras. We were overriding those cards so in case we lost control of one of those, data couldn’t be harvested.”
—David France, Director
“I worried having no experience in this that we might be flirting with the uncanny valley,” admits France. “I reached out to Thalia Wheatley, who is a professor at Dartmouth College. She has a neuro-psych lab there, and I actually enrolled in a study to test the various approaches that Ryan had been developing to see if they work. In fact, the approach that we ultimately settled on scored a little higher than an uncovered face in empathy and in one’s ability to connect with the characters. We had been developing this idea called ‘the tell’ where there is a soft halo around the face that showed the viewer that something had been done to it.” The facial capture sessions with the LGBTQ activists took place in New York City with a nine-camera array. “We collaborated on the casting with David’s team to make sure within his pool of activists that we could find jawlines and general body composition that matched,” states Ryan Laney, who served as the Visual Effects Supervisor. “There were a lot of challenges in that. It was important for us to get something that felt natural when the veil was in place.”
With cameras not being allowed in airports in Russia and not wanting to draw attention to the production, regular film cameras were not a viable option. “We had rolling shutter issues from these consumer-grade cameras and low light where the grain was more than the signal,” remarks Laney. “We used some computer vision to find faces, so we got reasonable camera tracks from that. But there was a lot of manual tracks and de-graining sessions to mitigate these low-quality issues that came with some of the footage. In some of these cases where the grain was really prevalent, it sometimes showed up as micro-expressions, so the face would do strange things. It was interesting in how we dealt with each one of these complications so that we got a truthful representation of what was there originally. We ended up being all 2D. It was Nuke, After Effects, and edited in Avid. We relied more on talent than tools in this particular case.”
“After Ryan did his magic, we brought the scenes back to the subjects so they could see the entire run, and that is when we discovered that jacket was a gift, so we would have to change the jacket, or the writing on that shirt was going to give something away, or that piece of jewelry was a family heirloom,” remarks France. “We even found that one person’s head shape was a telltale sign of his identity. We worked on every part of the frame, like a room that would suggest what state, country or city that they’re in. Anytime we have an establishing shot looking out the window, that was often a different country all together. We wanted it to be confounding for anyone trying to use the film to find people.”
“We even found that one person’s head shape was a telltale sign of his identity. We worked on every part of the frame, like a room that would suggest what state, country or city that they’re in. Anytime we have an establishing shot looking out the window, that was often a different country all together. We wanted it to be confounding for anyone trying to use the film to find people.”
—David France, Director
“It was all done in service of truth. We did not create fiction. We created the ability for this truth to be told publicly through the use of visual effects. It’s about giving power to people who have been powerless. It’s about giving back humanity to people who have been chased into the shadows. It’s a phenomenal use of the technology that we have enjoyed for so many years in Hollywood filmmaking being put it into the defense of humanity for the first time.”
—David France, Director
An unveiling takes place as to the true identity of Grisha. “Maxim Lapunov had already done that press conference before we began this work, so there was a discussion initially about not covering him,” explains France. “But people needed to know how much risk he was already facing, so we decided to cover him until that press conference. Then we debated at what point in the press conference do we reveal him. [After a test screening] we decided to do it when a woman suddenly gives him a microphone and calls him Maxim Lapunov. Ryan tried a number of other ways to do it. There was just so much magic in Grisha’s brown eyes becoming Maxim’s beautiful blue eyes that invited you to feel the peeling back of the skin.”
The term deepfake does not apply to Welcome to Chechnya. “It is a similar style transfer of a deep-learning idea,” notes Laney. “But we think that deepfakes are inherently non-consensual, and we were careful about being open to everyone as to what was going on.” France is in agreement. “It was all done in service of truth.
“We collaborated on the casting with David’s team to make sure within his pool of activists that we could find jawlines and general body composition that matched. There were a lot of challenges in that. It was important for us to get something that felt natural when the veil was in place.”
—Ryan Laney, Visual Effects Supervisor
;
