From its inception, “Dunkirk” was never meant to be a VFX-intensive World War II depiction of the legendary evacuation from Northern France. Rather, Christopher Nolan’s plan was to shoot almost everything with IMAX cameras, and seamlessly combine VFX elements to deliver the immersive, doc-like action.
And with Andrew Jackson (“Mad Max: Fury Road”), Nolan found the right visual effects production supervisor to oversee the project with Double Negative. On Saturday, at the Academy’s annual “bake-off,” Jackson will make the case why “Dunkirk” deserves a nomination. “Chris avoids full CG shots as much as possible and uses live-action elements as much as possible,” he said. “The mandate was to combine visual and special effects and make it look gritty.”
Courtesy of Warner bRos. Picture
Diving into the aerial dogfights
After reviewing hours of archival footage and flying reference, Jackson worked closely with Nolan to develop choreography of aerial dogfights featuring Tom Hardy. Jackson then oversaw the previs. However, Nolan didn’t want to follow the previs too closely because he was afraid that the final result would contain unrealistic camera shots.
“So we stayed in that language of real aerial footage and used previs only as a guide,” Jackson said. “I distilled it down to a one-page drawing for each sequence, which I would take with me in the helicopter or the plane while we were shooting.”
Having done the previs and knowing what needed to be shot, Jackson then worked closely with the aerial unit and the SFX team (led by Scott Fisher) with miniature planes. They had three original Spitfires and one Messerschmitt along with a Yak plane dressed up with a Spitfire canopy. Fisher and his team built 1/4 models of the planes.
“I had a lot of input in [the aerial sequences], but it’s probably the one with the least amount of CG effects. Most of what I did was going out and filming real planes right at full size or large miniatures, and the post for that was really all about combining the practical plane elements with other practical elements of the cockpit and the canopy. And we had smoke effects on the planes and other pyrotechnic effects as well.
“We were really doing straightforward, simple comps of the foreground elements and background and keeping both plates as much as we could with scratches and the flared reflections off the canopy to retain that raw, gritty feel.”
For close-ups of pilots not shot in real planes, they set up cockpits on a gimbal in the cliffs of south L.A. “Some of the crashes were not deliberate, but we used them anyway at the end of the film,” Jackson said.
There were three sinking ship scenes: the hospital ship, the destroyer, and the mine sweeper and all of the explosions were practical (supervised by Paul Corbould). For the initial listing of the hospital ship, they used two cameras to simultaneously shoot stunt performers leaping into the water. This was comped and followed by half-scaled ships lowered into the water, augmented with minimal VFX.
The sinking destroyer at night was shot at Falls Lake Universal Studios using a full-size section of a ship on a hydraulic gimbal to simulate the action. The sinking of the minesweeper involved closeups of ships’s section on the same gimbal combined with a full CG ship and crowd simulation.
Crowd scenes on land
Most of the crowd scenes were shot in camera. However, Nathan Crowley’s art department built thousands of painted cut-out soldiers in groups of 10. These were then arranged as fences in rows behind the more than 1,000 foreground extras. The primary use of full CG crowd simulation occurred in high, wide aerial shots revealing hundreds of thousands of soldiers in the distance.
The takeaway for Jackson was accepting the reality of the shot and embracing its imperfections. “Looking at the film as if you made it 30 or 40 years ago was our starting point,” he said. “You’ve always got the full visual effects toolbox if you need it, but that shouldn’t be our first approach.”