Pacific Rim: Uprising might just be an animator’s dream project; giant Jaeger robots fighting giant alien Kaiju. Whenever I watch a film like that, I always wonder, where do the animators start? Notwithstanding the fact that there’s already a Pacific Rim film out there, and perhaps plenty of other giant fighting robot films, something about Uprising’s animation felt different.
So I asked animation director Aaron Gilman to tell me. He’s now at Double Negative and is a veteran of such other effects-heavy films as Avatar and Iron Man 3 at Weta Digital. We talked about getting started with animating such large characters, finding performances, using motion capture and what Gilman calls his ‘Heft and Jank’ approach.
vfxblog: What are the first things an animator needs to think about when tackling large fighting robots and creatures?
Aaron Gilman: The first thing an animator needs to think about when dealing with large robots is if the scale of the movement relative to the camera is consistent with the overall feeling of weight and mass used across all shots. There is no fixed parameter an animator can fall back on to determine how slow or fast their action needs to be. A speed of movement may feel right for a 270-foot tall robot from one camera angle, but completely wrong for another.
For example, a close up of a Jaeger foot-passing camera needs to be slowed down dramatically in order to feel like it’s a massive object moving past the viewer. That same action with a camera much farther away, framing the entire Jaeger, would now feel slow motion. This is largely an interpretive process as cameras are always changing from shot to shot. The first question we always ask when looking at a blocking performance is are we getting the speed right?
Another thing we look closely at when first starting a new shot, is have we matched the client’s previs? We spend a fair bit of time comparing the two with each other. More often than not the studio gets very attached to the previs, and rightfully so. The director has worked hard on establishing the pacing of his film (often with another facility) and has determined the framing and composition of the shots during this process.
However, sometimes things can change, resulting in issues that need to be addressed because for whatever reason the shot no longer matches the previs, or because the previs might need to change due to other factors that have since come into play. Usually we will present a version of the shot early on that matches the previs, in order to confirm this is the route we are going to go, however, sometimes we’ll also submit an alternate version to propose solutions to an issue whether it be to the shot’s performance or camera composition.
vfxblog: What was your methodology for starting the animation process for the Jaegers? How was mocap and any other reference used?
Aaron Gilman: The first thing that needed to be done to get the animation teams from our global studios on the same page, was establish style guides for each Jaeger. Del Toro established a personality or style of movement for each Jaeger in the first film, and it made sense that we should do the same for Uprising, especially given that we had more hero Jaegers with a wider variety of design differences. As animation director it was my role to propose various options for each character to Steven DeKnight using film references and online illustrations that captured the essence of their persona. We analysed silhouettes, posing styles, and film characters to give the animators and our motion capture performers a sense of what was needed to distinguish each Jaeger.
It became immediately apparent at the start of the show that the volume of animation work we needed to do in an incredibly compressed schedule would make it impossible to key frame each Jaeger performance from scratch. We would need to use motion capture to establish the blocking of each performance and then work with the animators to retime the motion and introduce the required robotic nuances by hand. We cast our Jaeger motion capture actors very carefully, making sure to hire performers who could not only act like each type of Jaeger, but who could also fight, locomote, and even be proficient in the type of weaponry the jaeger used in the film.
For example, the Drone Jaegers transform from typical police Jaegers into vicious Kaiju who move and act like werewolves. We auditioned dozens of different types of performers, ranging from contemporary dancers to acrobats, actors, and even contortionists. Of all the people we auditioned we were only able to find one performer who had the right type of movement style and energy that captured what Steven wanted. In this respect, motion capture gave us the luxury of being able to focus on volume, rather than hand keying extremely complex and interpretive styles of movement such as this.
vfxblog: What reference or approach did you then look to for the Kaiju?
Aaron Gilman: The Kaiju, in contrast to the motion capture process used for the Jaegers, were entirely hand keyed. The Kaiju combine so many different anatomical forms that designing a movement style ultimately becomes an amalgam of different types of creatures. To establish this, we used the characteristics of their anatomy to drive the design process. For example, Hakuja has massive bony plates all over his body, and a head that looks like a giant hammer or pickaxe. At one point in the film, we know that Hakuja digs underground. So we looked to bulls and mountain goats as reference for how he could move and smash things. We needed to justify why his body has that particular design, so it made sense to reference animals that smash or charge into things.
In the case of Shrikethorn, we combined the mannerisms of a Silverback gorilla due to his body shape, with the locomotion of a large cat to justify the way the energy moves through his body down to the tip of his long tail. For Raijin, we looked at Jack Nicholson from The Shining and the alien in Predator. As a bipedal humanoid character, Jack Torrence was a great reference because of the hunched animalistic posture and the arms extended away from the body, poised to slash. With Raijin’s large outer face plates, the predator character was a great reference for how they could open and close. As mentioned earlier, the Drone Kaiju were animated based on classic werewolf movies so we looked closely at shows like Hemlock Grove which did werewolf transformations brilliantly, and films like American Werewolf in London and The Wolfman.
vfxblog: The robots and creatures are in full action in the fight scenes, but what were some of the more subtle things you needed to achieve in animation with the performances?
Aaron Gilman: When dealing with extremely large creatures that need to feel enormous and heavy there is always a risk that everything looks like just a slowed down version of what a regular sized creature would do. Taking motion capture or hand keyed animation and simply stretching it out to suit the camera and tempo of the movement is never enough. In the case of mocap, you may stretch out the original performance to get the general speed of the character in the shot, but once established the animator needs to retime certain parts of the performance in order to milk the contrast between poses, augmenting the massive weight and compression points in the action. Otherwise you risk having everything look soft, as if it’s underwater.
With large hand keyed Kaiju, having every pose 30 frames apart is going to look monotone and dull. The animator has to pay close attention to moments of compression in the limbs and impacts in the arms, elbows and knees. When the Kaiju catches its weight on an arm you need to feel the joints catch the weight and recover accordingly with the shockwave running up into the shoulder. The muscle simulations would also help with this but the animator needs to key it in so that the simulations pick it up. Secondary movement of metal panels and plates on the Jaegers was also something we constantly looked at.
What we called ‘Gizmos’ were also a critical aspect to showing that these giant robots have inner mechanisms driving and supporting their massive weight. So each animator would import a variety of hand keyed and automated behaviors for the various inner workings of the suits. For example, each Jaeger had a unique animation for how their foot mechanism would react whenever the foot is raised or on the ground in order to justify how they hold themselves up without collapsing. These sorts of animations go a long way to showing that the Jaegers are actual mechanical constructions with pistons and hydraulics driving everything.
vfxblog: What were the tools and techniques DNEG used for animation? Any particular new approaches or tech you employed for the work?
Aaron Gilman: I wouldn’t necessarily say we used any new tech or developed innovative tools in animation to do the work. One thing that was extremely useful however was a system we referred to as ‘Heft and Jank.’ Using Maya’s layer feature, we developed an extensive library of curves that animators could apply to any aspects of their performance to get a more mechanical and heavy feel out of the performance. Heft referred to large overlaps in body weight, specifically the hips and shoulders.
For example, when a Jaeger walks we need to feel the compression and overlapping action of the shoulders as they impact and recover with each step. These additive layers of motion, when offset to the correct places over top the mocap, would help amp up the feeling of immense weight. Jank referred to the high frequency shaking that would happen on all the smaller metal panels whenever there are body impacts. This type of subtle motion went a long way in making sure the Jaeger doesn’t feel like a single massive piece of metal.
vfxblog: Can you talk specifically about the Gypsy Avenger and Obsidian Fury battle in Siberia? How was that fight broken down? What were some of the toughest things to pull off in bringing that to life?
Aaron Gilman: The fight choreography for the Siberia sequence went through numerous iterations and redesigns over the course of post-production. There were a number of complex story telling beats for the animation performances. On one hand, we needed to make sure the fight itself was bigger and more intense than their first encounter in Sydney. We know that Obsidian is a stronger, faster and more powerful Jaeger, and it’s just a matter of time before he destroys Gypsy. So it became essential to design choreography that not only suited the different persona’s of each character, but also showed that the only way for Gypsy to beat Obsidian is by out thinking him. We had numerous creative brain storming sessions to work this all out.
We know that Gypsy needs to learn from his mistakes in Sydney and beat Obsidian purely from a strategic perspective. He manages to do this by attacking Obsidian’s most destructive weapon and ultimately the thing that powers his entire suit. We loved the idea that Gypsy breaks off a piece of his own sword in order to beat Obsidian, as this showed that Gypsy is willing to fight out-of-the-box as he will never beat Obsidian using purely traditional tactics. A lot of these ideas came about iteratively as we worked our way through many different creative approaches.
Perhaps the most complex shot in the sequence is when Gypsy and Obsidian fight while falling down an enormous crevasse in the ice. This was tricky because not only did we need to have dynamic choreography, but we needed to make sure the fight was fast enough that they can get through the choreography while falling in real gravity. But it couldn’t be too fast that it broke the reality of their scale. In this respect, we had to work closely with our environment department so the animator could make adjustments to the crevasse walls and the overall depth of the hole while working out the animation in their scene. Because they are free falling, this shot was entirely keyed by hand using filmed video reference of our stunt performers.
vfxblog: What was your most memorable or challenging shot in the final Tokyo sequence?
Aaron Gilman: For me, the most memorable series of shots of the final Tokyo sequence was how we tackled the transformation of Mega Kaiju. Given how complex the work was, developing the transformation lasted the entire duration of the post-production process at DNEG. We knew we didn’t want the transformation of the 3 Kaiju into Mega Kaiju to be a literal concept. There would be no real story telling value in showing the specific details of how each Kaiju is stitched together. Not only would this be incredibly complex and time consuming, but we were more interested in the mystery of what is forming before our eyes.
Steven and Peter helped us conceptualize the transformation by thinking of groups of shots as part of a specific phase of the process. We thought of the transformation as a kind of shadow puppetry where basic concepts such as severing limbs, transporting limbs, and attaching them formed the basis of the progression of the shots. We decided to let the backlighting of the sun and the smoke generated by the Rippers as they perform the operation to work in our favor.
By doing this, we were able to focus purely on creating abstract shapes that move in and out of the anatomy of the original Kaiju and gradually develop into forms that expose the final shape of Mega. Once we understood the shadow puppetry concept, the animators had a blast merging the different shapes of each Kaiju together to form cool and abstract shapes. Ultimately we’re left with a clear understanding of the transformation as it unfolds while also not fully understanding what we’re seeing until the final reveal of Mega in all his enormity.