'Zombieland' was a project CIS Vancouver started working on shortly after completing the visual effects work for 'Angels & Demons'.
As Visual Effects Supervisor on the job, I knew going in to 'Zombieland' that the creative opportunities were huge and that it had just the right measure of creative pizzazz to zombie blood-splatter that CIS was looking for. As this story goes online, I will be available to chat in the CGSociety forums as the resident 'Meet the Artist' for about a week.
We met with Paul Linden, the project's overall visual effects supervisor, in Los Angeles and discussed the various types of work that Zombieland was planning. In total, the movie was tracking about 400 shots, which we knew would grow by the time we reached the end of shooting. CIS Vancouver was awarded three groups of work totaling about 110 shots: a set extension showing a destroyed Washington DC, a slow-motion car accident, and several green screen scenes where the characters are driving across the United States.
Our visual effects producer, Janet Yale, and line producer, Marta Knapik helped set up the team for the show, drawing on individual talents specially suited to this type of environment work. Leading the CG team was Earl Paraszczynec who had played a key role in the construction of our virtual environments on 'Angels & Demons', and would bring key knowledge to the project.
Onset the cameraman retreats from hungry zombie.
The opening shot in the movie shows an empty street just north of the Capitol building in Washington DC. We start on the American flag, flapping in the wind, and as we pull back, we see that the flag is the one on the front of the presidential limousine, which is, in fact, upside down. We flip back to the right side up, and continue to pull back, panning around the scene, taking in some of the devastation. We then swish pan back towards the Capitol as a zombie jumps over the limo and chases us down the street, finally catching up and consuming vital body parts. Lovely.
We had planned to shoot the shot in such a way that we would be adding a matte painting of the US Capitol building into the deep background. But, as happens on set all too often, the shot bounced around the schedule and was shot in a location that made our initial plans redundant.
When the plates came in for our Washington shot, we realized quite quickly that we would need to replace nearly everything in the scene with a combination of 2D matte painting and full 3D environment. The shot had a considerable frame count in its first iteration, and although it did reduce from the initial 1800 frames, it was still over 800 frames at the end.
Working with a shot of this length greatly impacts matchmove and rotoscoping requirements. Our matchmove team, led by Peter Hart did a great job of breaking up the shot into segments based on camera action, reassembling the CG camera moves later on into one seamless move. Jessica Wan, our rotoscoping lead did an initial pass of the rotoscoping to enable our comp and CG teams to work with the plate and also for us to produce a temp version for some earlier screenings. Subsequent edge-perfect rotoscoping, was created later by Wan's team. Our last version of the roto came to us on the last day of production. Every version was a step up from the one before, with refined edges with accurate motion-blur. I was initially really worried by the amount of rotoscoping that this shot would require, but Wan's team did a precise job masterfully refining everything so our compositor had very little to adjust.
Editorial in Los Angeles wanted to add a 2D tilt throughout the shot to focus the shot on the zombie action as he runs at the camera. It was a good move, but it did add significantly to our rotoscoping requirements pretty late in the game. Normally, you would rotoscope to just outside your contractual boundary, for this 1:778. For this shot we just bit the bullet and rotoscoped every thing in 1:33. Our visual effects editor, Adam Estey worked with Zombieland's visual effect editor, Joseph Carson, keeping track of the various editorial moves that had been added and that needed to be matched by our visual effects artists.
Paraszczynec accompanied me to Washington DC to photograph the possible locations around the US Capitol that would work for this shot. We shot using two Canon EOS 1Ds MK2 digital cameras, shooting RAW images. For each location around the Capitol, we did extensive photographic studies at different times of day to give us variations in shadow positions. We took over 40,000 images during the texture survey, and although at least half of those were of angles we didn't ever see, the strategy was to cover everything possible
in one trip, and ultimately be in a position where we could create what we needed, from any angle.
Working with artists Christian Emond, Casey Rolseth, and Biren Venkatraman, Earl Paraszczynec worked out a plan of action that would allow us to put together the expansive environment within the eight weeks available. We had to cut a few corners on our build, allowing textures to do more of the work than we would normally want. But 8 weeks was very tight no matter what approach you take and we wanted the end result to look good, so even though we cut a few corners, we invested our resources and time into the areas that needed it most.
Starting the build straightaway, Emond used Google Earth to get an idea for the correct position for all the buildings in the location so we could start blocking out the camera move. Hart continued to work with Paraszczynec and Emond throughout the process refining the matchmove, fixing bumps or slips that showed up from time to time.
Ariel view of CG showing correct layout of buildings relative to set layout.
Paraszczynec and Emond also began the task of sorting out all the photographs into groups so we could organize our projection cameras. Any images that end up being used have to be cleaned up; removing anything that we don't want in the image. This usually means painting out people, cars, birds, shadows, etc. On hundreds of images this is a large volume of work that needs to be done quickly and also look flawless.
Emond and Venkatraman had roughed up a model of the Capitol building for our temps and together refined the areas that would be seen, adding more detail for our lighting to pick up on in the final render. The way we use the projections normally require us to paint out any real lighting and then have it cast back onto the texture from the 3D objects. It's always a fine balance between painting something out and regenerating it later. The key is always to know which end result will look better.
3D layout of scene including trees, roads and Capitol.
Some shadows are so complex on certain objects, it makes sense to leave them in the texture; at other times the texture may have a great shadow in it, but no matter what you do it's going to be coming from the wrong direction and needs to be painted out.
With cleaned textures and a modeled environment, the team set about putting everything together. There was too much to place it all into one single Maya scene, so Emond broke it all up into multiple passes to be reassembled later in comp.
Our base model of the Capitol was built in order to allow the textures to create most of the lighting. Subsequent layers were added after to fix projection issues.
One of the reasons we decided to build the US Capitol as a 3D model rather than treat it as a 2D matte painting was based on how much parallax we thought would be missing if we took the simpler 2D approach. We knew from match-moving the scene that the real camera mostly moved backwards on the road and didn't really move left or right very much. The little left and right movement that is present is enough to create the slightest of parallax, and your eye can detect when it's not there.
Stephen James, our lead compositor, provided us with motion footage of real trees that he had stabilized and given to CG with mattes. We added the 2D trees and bushes on cards and stacked them up facing the camera in the 3D scene. The slight motion on the real trees added greatly to the reality of the scene, your eye picking out the slightest of movements in the leaves and branches makes a huge difference to the final result.
Adding light poles, traffic lights, posts, and benches, all at varying distances to us,
Emond then broke the entire scene back up into render passes for comp and sent them to the farm.
Added street poles, posts, benches and road signs.
Composite supervisor, Marlo Pabon, worked with James to assemble all the pieces together, and to layer in passes of 2D smoke, fire, and ash that Linden had shot on set. We wanted to create a moody scene, not too intense and apocalyptic; otherwise it wouldn't match the style of the rest of the movie.
During the final composite, James layered hundreds of layers together to build up the environment. Combining 3D renders, filmed elements and an overall damaged home video camera look, James hit render on the final version and we finished the shot on the last hour of the last day of the D.I. No point having a few minutes left over not being used, right?
During Zombieland, the effects would be used to illustrate various "rules" that the main characters would adopt in order to survive. Some rules would be obvious such as "know your exits" or "good cardio" while others would reflect the comedic side "enjoy the little things" and "Wear your seat belts". It was this last rule that Linden asked us to help create a pretty horrific visual example of, for the audience.
The sequence consisted of three shots, showing a woman ejected from her minivan as it hits the side of a truck, throwing her and the dashboard's contents into the air in slow motion, eventually to fall on to the road with her head splitting open on impact.
Always wear your seat belts.
Paul shot the elements in several pieces. The first was the background elements of the van and truck colliding. He broke that up into four pieces, with the van coming towards camera with no truck, the truck coming into position with no van, the van hitting a stationary truck at full speed, and finally a clean plate so we could stitch everything back together.
Stunt woman element mid rig pull.
Jessica Wan, our rotoscoping lead, watched over the rotoscoping teams assigned to break apart the truck, van, and debris from each element. Stephen James and compositor Tiago Santos began working on piecing all the layers back together to make the sequence work.
Paul also shot an element of a stuntwoman on a wire rig flying through the air towards camera. She landed on a crash mat just in front of camera. Again, Wan oversaw the rotoscoping of the woman, and our comp team went about putting the elements together.
Stunt woman element at end of rig pull.
CG supervisor, Earl Paraszczynec, began work on the Beanie Babies that needed to fly from the car with the woman. Paraszczynec worked out a rig for the various Beanie Babies that would allow them to behave realistically as they flew threw the air and then landed on the ground.
Simulation after simulation perfecting the dynamics.
Finally the simulation started to behave realistically.
CG Beanie Babies in mid flight.
The basic problem was maintaining the volume of the object when it deforms on impact. Once we'd worked out a solution for the dynamics, we added a short fur texture based on each creature's design.
We knew early on that we would need to add additional glass to the van as it hit the truck, allowing the body of the woman to emerge through it as she launched into the air. Paraszczynec worked with various off-the-shelf plug-ins at first, but found that they didn't offer the tools needed to simulate the complexity that you see as a car window breaks. We had to build a custom simulation for the window, allowing the thousand fragments of shattered glass to initially break, land on the truck, and finally fall on the road surface. It was very easy to picture what we wanted, but much harder to actually make the physics work accurately on so many pieces.
Finally Paraszczynec built a rough blocking of the world, including the van, truck, and road surface for the simulation to occur in.
We mixed our renders across RenderMan and mental ray, using each for specific benefits it offered the different materials we were working with.
Added gore to the woman's face, as well as road damage in the foreground.
James and Santos worked on blending all the elements together in comp. I shot some HD elements of blood bags hitting the ground with help from coordinator, Riley McDougall. It's not every day you get to go home and answer the "what did you do today?" question with, "Blew up some simulated human head blood bags, and you?". We felt that we needed to add a little extra gore to the impact of the woman's head on the road. James and Santos also added some blood and scraps to her face in both shots.
For the above angle Santos added blood elements to the road and used photography of curbs, road cracks, and general road damage to give the shot a sense of scale. We warped the blood smear to match the ladies arms and legs as they flail about on the impact, also adding the CG Beanie Babies and glass from CG.
Extra blood elements were used to add the smear to the road surface.
Zombieland was a great project to work on. Linden had a visual plan from the start and knew what he wanted the shots to look like. Everyone told me I had to work on at least one zombie movie in my career, and honestly, I couldn't see what the attraction was.
But now, after shooting blood elements, throwing Beanie Babies around, and discussing in depth what a human head would look like if it blew up, I'm hooked.
Visual Effects by: CIS VANCOUVER VFX Supervisor: Mark Breakspear VFX Producer: Janet Yale Senior VFX Coordinator: Marta Knapik VFX Coordinator: Riley McDougal VFX Production Supervisor:
Dennis Hoffman Digital Production Supervisor:
Jason Dowdeswell Head of 3D and Technology:
CG Supervisor: Earl Paraszczynec CG Artists: Christian Emond Casey Rolset
Biren Venkatraman Matchmove: Peter Hart Pipeline TDs: Justin Rosen Mike Tuffy Matte Artist: Romain Bayle Head of 2D: Christine Petrov Composite Supervisor: Marlo Pabon Lead Compositor: Stephen James
Compositors: Daniel Cairnie Tiago Santos Thierry Muller Daniel Bryant Simon Ager VFX Editor: Adam Estey DFX Colourist: Zane Harker Roto: Jessica Wan Digital Production Manager: Kristin Dearholt VFX Set Construction: Barry Burke
The CGSociety is the most respected and accessible global organization for creative digital artists. The CGS supports artists at every level by offering a range of services to connect, inform, educate and promote digital artists worldwide