Visual effects supervisor Charles Gibson returns as second unit director and VFX supe to the final installment of the Hunger Games franchise, Mockingjay – Part 2. With the nation of Panem in a full scale war, Katniss (Jennifer Lawrence) sets forth to kill President Snow (Donald Sutherland) in this dark and gripping final second half of the third book. Joined by a squad including Gale (Liam Hemsworth), Finnick (Sam Claflin) and Peeta (Josh Hutcherson), Katniss goes off on a mission with the unit from District 13 to stage an assassination attempt on President Snow who, as always, is obsessed with her. On the way the squad encounters mortal traps or pods and attacking lizard mutts as they walk through the city – all complex visual effects work.
Directed by Francis Lawrence from a screenplay by Peter Craig and Danny Strong, the film used a range of visual effects companies, namely Double Negative (Adrian de Wet), Weta Digital (Martin Hill), MPC (David Seager), The Embassy Visual Effects (Winston Helgason), Lola (Edson Williams), Skulley Effects (Culley Bunker, Tsuyoshi Kobayashi), Exceptional Minds (Craig Seitz), Magnopus (Craig Baron) and Previz by Halon (Clint G. Reagan). The film had around 1200 visual effects shots.
Mockingjay – Part 2 shot in Atlanta, Paris and Berlin with both location and studio work. The film was anamorphic shot (2.35 : 1) on the Alexa (Codex ARRIRAW ~2.8K)) with primarily Panavision Primo C- and G-Series with some E-Series lenses by DOP Jo Willems. He shot all the last three films although they started on film, a point that would come back to be significant when revisiting some of the model assets, since the film requirements were less exacting than the newer digital frames and so the assets had to be remade for high resolution digital master.
One of the film’s most impactful sequences is the attack under the city by the Lizard Mutts. To bring the Lizard Mutts to life, the visual effects team turned to Weta Digital under the supervision of Martin Hill. The mutts attack and overwhelm our heroes. “There is this escalation and for everyone they drop another 2 appear,” explains Hill. “In the tunnel you can’t tell how many there are but you just know there are more and more coming. But in terms of their herd motion, they are not being particularly smart about trying to trap them (the heroes). They are just going hell for leather trying to eat the team.”
On set there was no actual motion capture, the stunt crew were dressed in white tracking outfits (to match the Mutts), but this was to provide reference and physical presence for the lead actors . “We really wanted to get one to one correspondence with the creatures,” says Charlie Gibson. “We wanted to film them in the water and get all the splashes and all the interactions.” These stunt performers would later be removed painstakingly by Weta’s roto and clean up team to be replaced with the fully digital mutts. Unlike the book which referred to these mutts as having tails, the film versions had primarily humanoid bi-ped forms, but were still very much blind manufactured evil creatures with particularly nasty dental work. “They are without eyes,” notes Hill. “When you are looking at the design of the characters you need a focal point. We knew their chins would be up and their heads back, so the mouth becomes the focal point of the face. We worked hard to make it compelling with all the correct stretching, lip compression, and the snapping of the teeth – working well from all angles. And the design inside the mouth was really important.”
The characters were designed before the filming of the first half of these last two back to back films and the design transitioned from more ‘animalistic’ creatures on four legs to the character that we see in the film. These were all designed in ZBrush, and that was used in blocking previs and also in a stunt-vis to make sure the stunts could be safe and what the director wanted. After principal photography the team went back and did “a very tight postvis with Halon for the sequence here in Los Angeles,” adds Gibson, “not only to confirm which parts we wanted to keep and what we wanted to change from the stunt performances but Francis was very focused on body placement and body count, because we littered the space with dead lizard mutts, as the sequence progresses, and they become part of the geography as we move through the scene. That was all quite tricky to work out.” Weta Digital did not actually start their work until 18 months after the sequence was first sketched out, and thus the turnover to Weta was very well planned and constructed. This allowed Weta to immediately get started and move very quickly forward effectively.
These stunt performers were shot, at all times, with the addition of two witness cameras on set. Weta did detailed measurements and scanning of each of the stand in stunt crew and the set environment. Since the mutts are blind their action is very physical and up close. Hill worked closely with Sam Hargrave the fight choreographer and stunt coordinator to get as much safe interaction between the stunt/mutts and the cast. “That gave us really good performances from the cast,” says Hill. “They are genuinely pushing away the mutts/stunt people. They are really naturally straining and pushing these mutts away from them, it was great. Jennifer and Liam were really good sports about it, they were really into it and having a good time!”
Even with the cast doing as much as possible, there were some dangerous shots, which needed Lawrence to be stunt doubled. For these sequences Weta did face replacement. Weta has carved out a global reputation for digi-doubles and realistic face work. Lawrence’s stunt double is a good match for the actress – Renae Moneymaker has doubled for Lawrence on several films including X-Men, so much so that for some very fast motion blurred shots, no head replacement was required. Martin Hill believes Moneymaker is a very good match in terms of body, head shape for Lawrence and “she is an excellent performer in terms of what she can do physically. There is an amazing shot where she gets thrown into this gantry. They did a lot of takes of that and I winced watching that on every take.” For the shots that did need face replacement they would have Lawrence perform the stunt in slow motion safely, providing lighting reference and sometimes allowing for a 2D solution. Then Moneymaker would do the scene, with tracking markers on her face, for later projection mapping or full digital face replacement.
The sequence was also that much harder as it was in a wet drain system with a lot of water splashing around and very confined spaces. The set was lit primarily with Kino flos and by flame bars, which was all enhanced in post by the Weta effects team. The aim was always to keep as much as the team could from principal photography and after each set up multiple HDRs were captured. It was important to have the stunt performers standing in during the HDRs for correct bounce light et cetera, given the mutts (and their stunt performances) were large white objects in the scene. “For every shot we would take an HDRi with the flame bars off and then one with them on, and also we’d be very careful where the film crew was standing,” outlines Hill. “If the cameraman was standing in front of a practical source, we’d make sure he’d stay there, and we’d take them at multiple locations around the set for spatial variation.” Any bounce off from an actor, say Lawrence, would also be important for the mutt’s lighting, so later a digital version of Katniss was made just to accurately add her bounce to the HDRi lighting solution.
The team shot a library of water elements in the set, and then these could be composited into the final shots as needed in addition to new fluid sims being designed and rendered for specific water interactions by Weta Digital.
Pods: oil and barbed wire nets
One of the most dramatic sequences on the way to Snow’s mansion is the oil pod trap that starts with vicious gunfire and explosions and ends with a massive oil-like flood. Shot in Paris in a housing development, the VFX crew not only had to introduce the massive fluid simulation but also the post-flood blackened courtyard. On the day there were no additional black material in the frame; the VFX team had to turn a bright grey colored lit square into a black tar dried environment without any help from physical props or effects. “DNeg did a fantastic job,” says Gibson. “And to make matters even more difficult for them, at that location it was hard to set up any blue screens and so they did quite a bit of it with just brute force roto and they did a brilliant job.”
The bounce light changes so much between the sun lit grass of the plate photography and the huge black space and what that meant to the kick back onto the lighting of the characters. “One of the most difficult things was just having to do a blackened version of that environment and how that affected the quality of light,” adds Gibson. “There was a lot of work, especially on some of the foreground characters to integrate them into the black environment.”
Without solving this, even if DNeg’s roto may have been perfect, the characters would still seem very out of context and stuck on the blackened post-attack space. DNeg had no other option than to manually relight the cast via elaborate compositing and skillful regionalized color grading. “It was really hard to imagine where to begin to get, effectively, negative fill light on them,” states Gibson. Even if the team could have laid down some black, the vast space meant a huge color grading lighting difference in terms of the the negative fill, the light wrap, tonal and edge qualities. “It was just great compositing, they did a brilliant job,” Gibson compliments. “Adrian De Wet, their supervisor, was there on set, and the key grip was standing by to do some quick blue screens if he’d wanted to, but that would have added to their own problems and the camera moves were fairly dynamic. Oh, and what you would not know unless you saw the breakdown is that – in that space there were a whole lot of trees they had to remove as well!”
The part of that sequence involving rocket fire was completely digital with a dust mortar mainly “as an actor tool”. Adrian De Wet and his team had worked on the majority of the previous Hunger Games films and they took over digitally the gun fire, rocket attack, building collapse and the destruction outside the windows where our heroes were hiding out on the other side of the courtyard.
Pods: meat grinder
More pods are triggered in the film when the team escape underground and set off the ‘meatgrinder’. Embassy VFX did this pod escape in addition to more hangar work, building on the very similar hangar work they had done in the previous film. It was important for Gibson that this sequence, happening so closely in the film to the Lizard Mutts, be realistic and not comical given the floor was churning up and light beams were zapping down from above, not unlike a video game. “The key to that was ripping through the sequence as quickly as possible, the environment was pretty much as photographed with the exception that it was digitally extended to three times its length,” he explains. Embassy created all the debris, bullet hits and the grinder itself. The explosions were a combination of live action elements and digital elements.
The sequence was shot on location in a parking structure under the Messe Convention Centre in Berlin. The specifics of the pod were not as important as the drama and energy it created in the story and how it moved the characters forward. Some pods needed more audience exposition but for the grinder it was enough to witness the threat and feel the urgency of trying to escape it, without needing to stop and explain the machinations of it in any detail. “Embassy came up with an elegant solution in terms of something that could be brought in one after another to create a sense of linear travel and the timing of it could be modulated to match the actors,” explains Gibson. “It is something that could have looked terrible, but it just looks like another terrifying moment that happens after the mutt attacks and everyone was really happy with it.”
Pods: Berlin flamethrowers
At the start of their journey to the Capitol’s center the squad come across a well flagged trap with the use of the Holo device. In reality the crew filmed real flame throwers with “a sort of shower head on it” which provided not only the “nice orange saturated flame” but the black smoke that came from the mixture of diesel and alcohol, explained Gibson. But the massive flames were still not quite huge enough, so Gibson got the team to bring in blue screens, and then by moving the camera around they filmed the rig at different angles. The flames were already very large and extremely hot, but the crew managed to get good additional takes that could then be used as part of the compositing by DNeg in NUKE to make the flames seen on screen.
Katniss and Gale avoid being revealed as they walk to the mansion by a surprise attack from the advancing rebel forces. Unfortunately even with months of location scouting the production could not find the right location with the correct mix of high buildings, arches and yet something that felt like it would lead to a mansion. So the sequence ended up being made up of two locations. At the first of these were some approximate building bases built by the art department, and some shipping containers clad in blue screen. The majority of the environment and the subsequent explosion, RPG, various Peacekeepers up above and attack debris were all added digitally.
“DNeg did a tremendous amount of digital architecture for the sides of the streets, above a certain height and up and down the street in both directions,” outlines Gibson. Because the location was late, the bottom sections that were built were not highly detailed and so work was required even for some of the lower set pieces. “They really did so much work in that sequence and the sequence that followed. They literally touched hundred of shots, and in that whole area of the film they did just a great job.” Gibson feels this sequence is one that he is particularly proud of, given that the plate photography is primarily “blue screen over containers – and in the film all that is digital environment – and you just don’t question it – as it should be.”
At the end of this long sequence the “girl on fire” was actually on fire following the parachute explosions. To film Lawrence on fire, her stunt double was filmed with flame gel all over her, but with the stunt woman wearing a face mask to protect her face and lungs. This base element gave the team something that could be tracked back onto Lawrence who was shot in the same location, but clearly not alight. The team then shot specific elements and then DNeg added contact lighting and modelling to produce the final result. The comp team at DNeg pieced together all the various passes and reference into one believable burning body. Unfortunately they did such a good job, the rating board MPAA made them “tone it back – as it was a bit too effective!” jokes Gibson.
DNeg’s Singapore office used Clarisse iFX as their primary renderer on Mockingjay – Part 2 after testing it extensively on Mockingjay – Part 1 on background buildings. “Clarisse is an intuitive tool, and it’s easy for the artist to pick it up quickly. That’s why we realized rapidly that we could use Clarisse more and more and be confident about it […] As the crew from Part 1 remained unchanged for most of it, and so already had a play with Clarisse, it felt like the right time to fully transition,” explained Eric Vezinet, rendering supervisor at DNeg Singapore in a release. “The movie was presented as a project involving massive environment, as well as some crowd work, thus large amounts of geo. It was a pretty big show!” The release notes that there were about 60 to 70 AOVs per renders, all at 3K resolution including some stereo renders.
Jay Grunfeld, visual effects supervisor from Cantina Creative, once again led the team on the screen graphics and the invaluable pod mapping ‘holo’ device. While there were less screens in this film, there were some and once again Gibson tried to make sure that there was always some monitor for bounce light or contact light on the actor’s faces – such as when Snow is reviewing an attack on our heroes. While these would be removed and replaced with holographic projections, having a real screen to view and to cast light was thought to be highly desirable. While the screens would aim to have the approximately right material, the monitors would have diffusion on them to better reflect the spill from the holograms. “Those guys are great designers and I think their material always looks great,” says Gibson referring to the various ‘mock’ user interfaces that Cantina has had to add to not only this film but earlier Hunger Games films.
Hovercraft and hangars
Throughout the film such as when Katniss sneaks away, there are a selection of military Hovercraft flying, fighting and landing. The Hovercrafts were primarily done by MPC, with the exception of the underground Hovercrafts which were done as before by Embassy.
Embassy’s reel from Mockingjay of the Hanger sequence in part 1
The Hovercrafts were also used in the mountain attack in District 2 , but in a classy move, the director decided to play at least the first part of the attack out of focus over the shoulder of Katniss, knowing full well that the audience was able to understand what was happening. Of course all the visual effects were done in full quality. The sequence was post-vised to find the right balance of what the audience could or could not see. “MPC did some lovely detailed simulation work – which of course we dropped out of focus and suppressed – but it is all there!” notes Gibson. It is rare to see a director so artfully not show everything, even when the work is spectacular and as impressive as the material was when first rendered.
Avenue of Tributes
DNeg had done the original Avenue of Tributes for Catching Fire, but that film was shot 2K anamorphic, the new film was full resolution Alexa. The new footage of course is free of weave and float or film grain and perhaps surprisingly the assets created for Catching Fire proved too low res to reuse as is in the new film and DNeg needed to rebuild many assets. “That surprised DNeg and they had to rebuild quite a bit of the city for the wide shots,” says Gibson. “For the big aerial wide shot of the avenue there was quite a bit more detail apparent just in that resolution bump to the Alexa – enough to make them redo those assets.” The Avenue had extras shot as a separate element pass in LA just a few months ago, but the wide crowd shots are all digital crowds.
For some of those crowd shots, DNeg turned to Clarisse iFX. Vezinet explained that “we had a shot with more than 100,000 actors and props on top of the entire city environment, and no geometry LOD for background buildings.” While this seems like a daunting issue, Vezinet says “we had to deal with brute force, huge Alembic caches loaded into Clarisse”, but “the nice surprise was that once loaded and live in the project, Clarisse was able to handle very easily the amount of geo. It was pretty easy for the artist to navigate into the entire city, replace their lights, do their layouts etc. Clarisse was still very interactive despite the fact that there were more than 220 billion polys in this environment! It was rather easy to navigate into both the raytraced views, and do proper artistic work on the crowd, like moving and removing actors, quickly testing different look alternates, etc, all interactively, with the real CG around.”
Oscar winning Craig Barron, the creative director for Magnopus, was also brought in to help with some of the wide matte paintings. He and Gibson had worked together in the past, and Gibson was clearly delighted to have Barron’s extraordinary talent contribute on this film. “We were looking to get some scope in the film – and he is the best…we were very fortunate to bring him on to consult.”
There is one classic shot approaching the Capitol that has featured in almost every Hunger Games film since Sheena Duggal supervised the VFX for the first film – it is an epic shot across a damn wall, over a lake to the Capitol. In this film of course unlike earlier films the shot is dramatically different as we can see the city burning and our heroes need to walk in, as the trains are no longer running. Gibson shot a new helicopter plate for this final reveal of the once impressive Capitol skyline. DNeg did the digital work building on the plate photography shot as all the previous films had done before at the Strom Thurmond Dam in South Carolina. The DNeg team removed traffic from the top of the Dam and tracked in the disused mag-list rail line and the entire city burning behind.