Edson Williams and Shahana Khan explain advancements in face projection and select lens techniques in David Fincher's latest Oscar contender.
Leave it to David Fincher to push the digital envelop ever so slightly in The Social Network with two simple VFX techniques to enhance his storytelling. For instance, he suggested a "hockey mask" face projection method to create the illusion of the Winklevoss twins (both played by Armie Hammer), the arch rivals of Facebook founder Mark Zuckerberg (Jesse Eisenberg). In addition, to create the illusion of miniaturization for the twins during their dramatic loss in the famed Henley Regatta in England, the director recommended a new select lens effect based on tilt-shift photography.
To help with the face projection, Fincher enlisted Lola Visual Effects (under the supervision of Edson Williams), which worked on The Curious Case of Benjamin Button. They first thought of utilizing Paul Debevec's Lightstage method, but it proved unworkable with the Red camera.
"So we developed a similar system using traditional lighting (12 kinoflo 400s and 8 bi-color litepanels) and DMX stage lighting software," Williams explains. "We wanted to use the Red because the movie was shot on Red and it also gave us a higher res and better dynamic range. We're able to shoot realtime, 24-frames-per-second, with four Red MX cameras running, using a stand-in to match the lighting of the offline.
"We had the body double [Josh Pence] sitting in a dental chair with head restraint to minimize movement. During the shoot, Armie and Josh acted normally. We tracked about half a dozen dots on Josh's face. No HDR was needed: this takes time and can break the mood of the performance. In editorial, Fincher chose what take he wanted and we sent it offline to Lola. During evaluation, we analyzed the offline and determined main keyframe locations. During his performance, Armie watched the playback of the offline on the monitor wearing headphones; Fincher gave very precise directions for Armie's performance [enabling subtle nuances between the twins]; it took 10-20 minutes per shot to capture. We tracked Josh's face using PFTrack and then swapped his 3D face with Armie's. For the pre-grade, using Maya, we projected footage from the Red cameras onto a cyberscan of Armie's face. Back on Flame, we composited the face we received from the 3D department."
This method was used for 20 shots, with nearly 100 additional shots required for other split-screen work involving Hammer since they were unable to shoot at Harvard. Michael Watson, a camera operator on several Fincher projects, helped with technical aspects of the Red and the lighting as the internal DP. Since the flesh tones of Hammer and Pence are different, this required some additional tweaking. The process worked really well, according to Williams, allowing Fincher to get a very controlled performance in a very efficient manner.
Meanwhile VFX Studio a52 and their VFX Supervisor Shahana Khan worked on implementing the selective focus technique for the Henley Regatta sequence. "We were able to achieve a miniaturized look, but also showcase the Winklevoss brothers' unawareness of the world around them," Khan suggests. "They were always trailing Mark Zuckerberg's actions, and losing the race is yet another allegory to being one step behind. Everything around them is literally out of focus."
Not only that, but given the brightly lit, expansive location and the narrow depth of field during the actual race, the regatta seems very Old World, so the twins become fish out of water temporally as well as spatially.
"The first challenge was to create a process that we could use from shot to shot because the entire sequence runs about a minute-and-a-half, but there are 57 shots and each one was affected," Khan continues. "The first thing was to pull reference, some of which came from actual tilt-shift photography, while a lot of it came from sports photography. They shot during an actual Henley Regatta but not the race. It was important to make it feel like an action sequence because it's the only one in the film. It was also important to maintain the aesthetic, so we added a lot of grain to the footage because the Red provides a crisp and clean look.
"We applied an overall focus to each shot so it wouldn't appear jarring and the focus didn't change dramatically from shot to another. To achieve this effect all the subjects were isolated on their own plane. Articulate mattes were cut of each person in the scene. There was also extensive background replacement for many of the shots. We were fortunate to receive a lot of B camera footage shot at the Henley Regatta. This allowed us to accurately incorporate crowd, boat and additional objects. Additional set photography was also provided to create authentic matte paintings of tents and additional architecture, aiding to the chronology of the sequence. Silver clouded sky replacements played an essential role in creating an ominous mood.
"Each shot was tracked in 3D. Using this camera information allowed us to replace the backgrounds via camera projection and also create detailed depth maps. These maps were used to simulate the lens effect. We also created a custom kernel on which the blur was based.Noise displacement textures were also generated to add movement to the still backdrops, creating subtle movements and oscillations."
Khan worked round the clock with both day and night shifts, compiling an edit every day for review: it was very time consuming and meticulous work to isolate every object, implement the backgrounds, relight them and add blur so they didn't appear as still photographs.
"Typically, tilt shift is a technique most commonly used on wide shots or birds eye photography," she explains. "It is most effective in these circumstances, because it simulates a shallow depth of field normally encountered with macro lenses, making a large scene seem much smaller than it really is. This proved challenging in a sequence of 57 shots, the edit accelerates as the race closes in, cutting from a variety of wide, medium and close-up shots. Initially, we needed to define the lens effect across the diversity of these shots, creating an effective aesthetic that was both true to the idea and achieved the intentional plane of focus, without jarring the viewer."
Bill Desowitz is senior editor of AWN & VFXWorld.