LOS ANGELES — Captain America: The First Avenger tells the story of the iconic World War II comic book character Steve Rogers, who uses guts, gumption and noble courage to gain entry into a top secret government project that transforms him into a superhero that saves the world from destruction by the hands of the evil Red Skull.
Bringing that larger-than-life saga to the big screen took the skill of director Joe Johnston (Jurassic Park III, Jumanji, Rocketman), the team of editors Jeffrey Ford (
Crazy Heart, Public Enemies) and Robert Dalva (
October Sky, Hidalgo), and the combined talents of 16 visual effects companies managed by visual effects supervisor Chris Townsend (P
ercy Jackson & the Olympians: The Lightning Thief, Journey to the Center of the Earth).
“It’s a heightened reality of the 1940s that we’re portraying here” said Johnston at the Captain America red carpet premiere at the El Capitan in Hollywood. “I used Raiders of the Lost Ark as a template for this film as a way to do the period for a lot of reasons: it looks timeless even today, it looks fun from beginning to end, and I wanted this film to be the same way.”
The movie was shot by cinematographer Shelly Johnson (Hidalgo, Jurassic Park III) using primarily the Panavision Genesis with additional footage provided by the Arri Alexa, Vision Research’s Phantom, 35mm film and even the Canon 5D. With the Genesis camera, principal photography was recorded directly to the Panavision SSR-1 solid-state recorder and converted on set using Panavision’s digital transfer station to DPX log frames at 1920x1080.
“At the end of the shooting day we would send the DPX files to E-film and they would color time the dailies,” says Kiran Pagglegadda, first assistant editor to Jeffrey Ford.
Cinematographer Johnson supervised the color timing with E-film colorist Andrew Francis and encoded the DPX files to Avid DNxHD 36. “We would use the color-timed Avid media for cutting and back-up the raw DPX files to LTO and have them on hand to pull material for our 3D conversation house Stereo D and our VFX vendors.”
After nine months of shooting in England, post production went into full gear beginning in January 2011 at the Marvel Studios’ facility at Raleigh Studios in Manhattan Beach, CA. That became command central for Captain America’s editorial and visual effects team.
“At our peak we had 14 Avid Nitris DX workstations on the show,” explains Pagglegadda, “used by two editors, five assistant editors, three VFX editors and two stereo 3D editors. In addition, one Avid system was used for VFX review sessions and another was dedicated to exporting material for sound, music and digital intermediate turnovers. Our first assistant sound editor’s Pro Tools station was also connected to our Avid Unity to streamline sound turnovers.”
There was over a million feet of footage to work with but curiously no previsualization on the VFX-intensive movie. “Johnston has it clear in his mind the kind of shots he likes and how he’s going to shoot,” explains VFX supervisor Townsend. “He doesn’t want to be tied down to something until we get there on-set on the day.” A previs, says Townsend, can potentially block a spontaneous opportunity of getting something better than the previs. Without that guide for VFX shots Townsend caucused with the director to determine Johnston’s shooting strategy and carefully studied his previous directorial work. That way the VFX team would be ”as prepared as we possibly could on the shooting day.”
Meeting the movie’s July 22nd release date necessitated the efforts of two principal editors. Dalva had worked with Johnston before on Jurassic Park III and Jumanji. At Ford’s request his agent had approached Marvel about editing The Avengers and found that Captain America was looking for an additional editor. Though they never previously worked together the two editors instantly created a highly collaborative working relationship.
“We chose to do a leapfrog approach,” says Ford. “I worked on the beginning of the film, the end and a little piece in the middle; Robert worked on the pieces in-between.” Ford and Dalva broke the movie into around 10 sections that were “sequences of scenes.” More like “acts” than “reels,” these sections ran from nine to 20 minutes in length. “It made sense to do it that way. That’s how we got continuity and flow,” recalls Dalva.
“It was a lot of film because there were multiple units. There wasn’t a lot of coverage, but it was really smart coverage,” notes Ford. “Joe is efficient,” echoes Dalva. “He knows what he wants. He covers scenes, he doesn’t over cover scenes. He’s a great visualist. He understands where the camera should be and how it should work with the blocking he’s designed. Cutting is never easy, but he makes it easier. What he brought to Captain America more than anything is heart. That’s one of his great talents.”
“We were right next store to each other,” recalls Ford. “Which meant that only one of us could be working on a loud scene at one time because you could hear the machine guns and shield through the walls.”
Ford and Dalva would give notes to each other on sequences, recut them, then show the revised cut to Johnston. Sometimes they wouldn’t agree on all those notes, but they “always inspired conversations which led to making a better scene. When you talk about it that just leads to better ideas,” comments Ford.
“We would play off each other” explains Dalva, but ultimately everything they did was for the director. Johnston didn’t sit at anyone’s elbow but would make director’s rounds between visual effects and editorial.
From an editing standpoint both editors concur the movie was virtually problem free. “You always work on certain sections more than other sections, but not because it didn’t work.” The question about editing a scene says Dalva is, “should we shape it this way or should we shape it that way?”
“There’s always a scene in a movie,” says Ford, “where you get it to the sweet spot and it always plays; you never have to touch it again. There’s always a scene where you see more ideas until they rip it out of your hands. I’m willing to keep trying until the last minute. The problematic scenes you tend to solve; the scenes that you never finish editing are the ones that can be done a bunch of different ways, and you can’t decide which way is best for the character because they all work.”
VFX & POST-VIS
As the cut was being built, VFX shots would show in the timeline as backplates or slugs. The editorial team worked with The Third Floor and Scott Hankel on post visualization effects shots, which would be built in Autodesk’s Maya, comped together in Adobe After Effects and dropped into the scene. “The director is an artist himself,” describes Hankel. “He would draw up the storyboards of what he intended from the plates.”
Hankel would then take the plates, flesh that out by adding CG elements provided by the VFX vendors, animate and export the scene, pass that on to Townsend for approval, then to editorial to cut into the movie.
Over the three months of post-vis The Third Floor averaged over 30 shots a week with a three- to five-person team. “What we do is fast — turn things around in a couple of hours and give it to editorial. A single shot would get revised 10 or 15 times playing with different elements,” continues Hankel. The final cut and approved post-vis scenes would then get sent back to the vendors as a guide for building final shots.
Marvel Studios was also the review center for all visual effects work. Shots would come in from all over the world through the studio’s ultra secure file transfer and server systems.
Visual effects supervisor Townsend wrangled a global battalion of VFX companies and supervisors working on more than 1,547 VFX shots that included: Double Negative, Charlie Noble; Lola VFX, Edson Williams; Fuel VFX, David Morley; Method Studios, Sean Faden; The Senate, Richard Higham; Trixter, Allessandro Cioffi; Framestore, Jonathan Fawkner; The Base Studio, James Pina; Matte World Digital, Craig Barron; Look FX, Max Ivins; Peanut FX, Amelie Guyot and Peregrine Mccafferty; Evil Eye Pictures, Daniel Rosen; Luma Pictures, Vincent Cirelli; Whiskey Tree, Jonathan Harb; Rise Effects, Florian Gellinger and Hydraulx. Thompson also managed Stereo D and 4DMax, the two companies responsible for the Captain America’s 3D conversion.
“A lot of the photography defines the visual effect,” explains Ford. “The actors do their thing and we’ll work the effect in later. The tail was not wagging the dog here. It was about story and performance. The best stuff in this movie nobody is going to notice. It’s a performance-driven, character-driven movie.”
Lola VFX provided 350 shots, including Red Skull shots and the remarkable “skinny Steve” sequences where Chris Evans becomes the character of Steve Rogers as the 97-pound weakling. Evans told Johnston that he felt it was crucial for his character to do the actual physical action in those scenes, and the director concurred.
During principal photography, Evans played the shorter character by stooping down or playing to higher eyeline marks on the other actors. “On set we would shoot the full scene with Chris Evans and all the other actors,” explains Townsend. Then for reference they reshot the scene with Evan’s “skinny Steve” body double, Leander Denny, mimicking precisely what Evans had done. A third pass was done with both actors absent from the scene, and sometimes a forth pass with a greenscreen placed behind Chris Evans was shot to have the character as a separate element.
From those raw plates Lola used three different techniques to generate the effect. For the majority of the shots, they rotoscoped Chris Evans — separating him completely from the background plate, then shrank him down and put him back into the scene using the original plate plus clean plate elements around the shorter Evans. “That was the first choice from production. They wanted all of Chris’s performance, including his body, his face, everything. It was really important to them to maintain the original performance of Chris Evans,” says Williams. “In shots where we couldn’t shrink Chris down we had two techniques. One was a head replacement, where we would take Chris Evans head, we’d reduce the mass of it and then put it on his ‘skinny Steve’ body double.”
The last technique used the Lola Face Re-projection Rig pioneered in The Social Network to create the Winklevoss twins. Evans sits stationary surrounded by four cameras. Computer-controlled lighting changes around him to mimic the lighting in the principal photography. “So if he’s sitting and he looks toward the sun and looks back towards the camera when we shoot him he’ll be stationary but we’d move the lights around him to look like he looked toward the sun and back,” describes Edson.
Footage from those four cameras was then projected onto a 3D model of Evan’s face and that 3D projection sat like a hockey mask on Evan’s body double’s face. “So instead of replacing his whole head it’s a way of replacing just the facial features while maintaining Chris Evans performance.” Out of roughly 270 “skinny Steve” shots, notes Williams, “We did seven or eight face projections, probably 25 head replacements and then the rest were Chris Evans’ slimming down.”
“We do all our compositing in Autodesk Flame. Lola has 21 Flames now; it’s got the richest toolsets,” says Edson. With the Flame they had the power to scrub a full 2K plate and watch the shape track with the image and “chew through the really complex shots. For 3D tracking we use PFTrack, and for all our 3D elements we use Maya 2011.”
The Senate used Autodesk’s Maya, The Foundry’s Nuke and Adobe Photoshop to build and comp multiple set extensions and augmented crowd scenes for a number of key sequences. This included the Kruger Chase, turning 2011 Manchester, England into 1940s Brooklyn, New York, and the USO performance scene, where multiple small crowd plates were layered together to emulate the scale of Radio City Music Hall. They used 3D projections of digital matte paintings enhanced with CG elements such as cars and architectural details, including the Brooklyn Bridge and people from bluescreen shoots from other films in the archives.
“The most challenging thing” says VFX supervisor Richard Higham “was the number of different angles each street view had in the sequence — making sure that each extension was consistent to previous shots. Being true to the perspective of the live action in the foreground, as well as being visually different enough to allow for the illusion that we are rushing through many streets, not just the two or three from the shoot.”
They also had to replace Chris Evan’s feet in a scene where he runs barefoot chasing a bad guy. Evan’s flesh colored shoes had to be precisely replaced with anatomically correct CG feet complete with wiggling toes, flexing muscles and skin that could stretch, fold and react when making contact with the ground. The Senate created 167 shots overall.
Method Studios completed 28 shots on the movie, including making a digital double for Evans’ jump out of an airplane over enemy lines. CG flak explosions, tracers and clouds were generated in Houdini and rendered through Mantra. The digital Captain America costume was built using NCloth, and the entire digital double was rendered through V-Ray out of Maya. Shots were composited using Nuke and Flame.
According to VFX supervisor Faden, “What was unique about this project was that we were matching a period effects look. We had fun researching and matching authentic flak and tracers from rare color WWII footage. The overall look was to be real but subdued. Another challenge was to translate the gimbal/camera work from the stage into a believable plane motion. This often involved reducing the gimbal motion relative to the outside greenscreen and including CG clouds and jetstreams rendered outside the plane to contribute to the overall sense of speed.”
Look VFX did around 65 shots, three main sequences, with a 15-person crew in about six weeks. To make the car “float” more in the Stark floating car scene, Look first removed the rig under the car that raised and lowered the car. In one section they removed the car entirely, rotoscoped out the actors and built a clean backplate without the car. Then on a new CG model of the car they used the actual car photography, mapped it on to the model, then animated the new car so it had that floating feeling.
The major sequence was the montage section. “It’s Captain American fighting his way across Europe blowing up buildings with his squad,” says Ivins. One element had Evans on motorcycle jumping out of a building just before it explodes. That was shot in reverse on greenscreen. Look integrated the Captain America element into a plate of a real building that special effects blew up in Eastern Europe. “That was a pretty challenging sequence because every shot was different,” says Ivins.
The last sequence they did was the Captain America fight scene just before Red Skull soldiers with flamethrowers capture him. Shot in the rolling hills and oaks trees of southern California, Look successfully turned those shots into the Ardennes forest and built and then animated all of Captain America’s CG shields. Look used Maya, Flame, Nuke and After Effects.
MIXING & AUDIO POST
When the time came to do final sound mixing and audio post, the entire post enterprise, people and all the technology moved from Marvel Studios to Todd- AO West studios, says Kiran Pagglegadda. Todd-AO even converted two dub stages into color timing suites for E-film to set up a remote grading facility for the 2D and 3D digital intermediates.
[EDITOR’S NOTE: There are many more visual effects stories to be told but not enough space to do it. Please visit www.postmagazine.com for more VFX companies and their contributions to Captain America:The First Avenger.]