Alain Bielik visits The Ruins with Rising Sun Pictures, which devised a new system for animating the nasty, man-eating vines.
The Ruins (opening April 4 from DreamWorks) follows five vacationing friends who uncover Mayan ruins that are entangled with nasty man-eating vines. Obviously, the very nature of the film's creature constituted a major challenge for first-time director Carter Smith and Visual Effects Supervisor Greg McMurry. This was not your typical movie monster: no eyes, no teeth, no mouth… Just a bunch of regular looking leaves. Yet, the team had to find a way to make them look scary and deadly.
This task was awarded to Australia-based Rising Sun Pictures, which created 110 shots for the movie. The team included VFX Supervisor Ken McGaugh, VFX Producer Mel Jones and Production Manager Caroline Grubb. Using conceptual sketches provided by Rising Sun Pictures and Tatopoulos Studios, and onset props as a reference, RSP had to elaborate a look and feel for the vines, as well as developing an animation style that would evolve as the story progressed. "Creatively, it is always a challenge to create an amorphous character that is structurally different in every shot of the film," says CG Supervisor Malcolm Humphreys. "We had to go through a long phase of performance development to try to understand the director's vision about how the vines needed to read on screen. Our animation had to continually be tone down, as the animators always wanted to deliver a big performance…"
RSP used Maya for the hero vine animation and lighting, while Houdini was employed for vine mass and dynamics. Rendering was carried out with 3Delight through Maya, and Shake was the tool of choice for compositing.
Playing with Lego…
During pre-production, the team realized that Smith and McMurry wanted to have control of each vine configuration. Onset, the practical vines could be pulled apart and put back together in all different ways. A similar "Lego-like" system was employed for the digital version. "The 'Lego' system allowed us to add, change type, move and delete appendages along the length of the vine," Humphreys explains. "We had four main different types of appendages on the vine (leafs, flowers, buds and tendrils) with their own selection of variations. With all of the different types and variations, we ended up with 540 possible combinations for the vine character."
This amount of data became a large management task for the artists, and the team had to build a workflow to help manage this complexity. Each appendage on the vine also needed to bounce, sway and collide in a believable way, but still be under the control of the animator, who had to hit the director's key performance points.
"We came up with a system called 'Anim Layers' that encapsulated an animation rig that we called 'modules'," Humphreys notes. "We developed modules for different animation purposes like performance, secondary motion and collisions. The modules could be snapped together into a chain to produce a layered animation rig. We ended up with so many appendages in the Maya scenes that we created different bind modules that allowed artists to selectively choose a different geometry detail for each appendage depending on what they needed. This helped animators and effects artists remove/add some of the geometry complexity in Maya for different tasks.
"There was a 'blind data' module that allowed us to remove all the appendages from a scene while still maintaining the animation data inside these modules. It enabled us to swap appendage variations in a scene and recover the previous performance. This happened a lot during the production when Carter and Greg wanted to change the configuration of the vine during animation."
Developing an Animator-Friendly Rig
The vine body presented its own unique challenges. On a previous project, Brett Leonard’s Man-Thing, RSP also had to do vines. For that show, they developed a plug-in called ‘noodlez' that took a NURBS curve and created a lofted NURBS surface around it, with some controls for shaping and controlling the twist of the surface. For The Ruins, the team wanted to take the vines to the next level. The main problems with the 'noodlez' system were that it was hard for the animator to wrap a vine around objects, and the geometry had to be very uniform and could not have modeled details.
To fix these problems, RSP created a custom curve shape in Maya called Super Spline that enabled them to choose the type of spline interpolation they needed. "We decided to use Kochanek–Bartels interpolation as the resulting spline travels through each knot that is defined while producing a pleasant looking spline," Humphreys says. "Having a curve that passes through knots made it much easier for animators to snap and wrap the vine onto objects that it needed to grab. We also created a deformer called Super Spline Bones that allowed us to deform any object with a Super Spline. Internally, the deformer splits the spline up into virtual bones/clusters; it had the nice side effect of speeding up the deformer when using really long curves (which was the case with the vine)."
Putting all these pieces together to create an animator-friendly rig was a quite an undertaking in itself. Spline-based rigs create some very unique challenges, especially when the spline rig is very long and needs to have variable dexterity in different shots along the length of the rig. In the case of The Ruins, the vine had hundreds of control points along its length.
"The rig we devised was similar to a hierarchal modeling system where you have a low-resolution polygon mesh that controls a high-resolution mesh," Humphreys observes. "But for the rig, we did this with curves rather than polygons. The low-resolution curve had 30 controls while the high-resolution had 175. This allowed animators to do broad animation with the low-resolution controls, which would then pull the high-resolution controllers around. The high-resolution controls could then be used to do more fine detail performance or add organic shapes into the vine. This rig made working with the 205 controls bearable."
An Animation Challenge
Animating the vine was no easy task, as animators didn’t have any of the rig tricks that have been developed for quadrupeds, tetra pods and bipedal characters throughout the years. The artists ended up being a cross between an animator and modeler…
One of the key issues they had to address was leaf on leaf and leaf on ground collisions. The fallback option was to just animate it all, but with the animators already dealing with a highly complex rig, another way had to be found. "We began development on another 'Anim Layers' module, as we wanted whatever we came up with to plug into the existing appendage workflow," Humphreys continues. "Whenever you deal with simulations in 3D, most of the time, you will need to tweak the result. Having the collisions being dealt within the 'Anim Layers' workflow meant that we only needed to add a tweak module after the collisions module to fix anything that didn't look right. We knew we wanted to solve appendage collisions on a lower resolution mesh, so that we could have more collisions solved in a scene at one time. We needed to solve the problem of mapping the low-resolution mesh back into the 'Anim Layers' skeleton pipeline.
"We came up with a plug-in called Mesh Driven Bones, which can be thought of as reversing how you normally bind skeletons to a mesh that you want to deform. With Mesh Driven Bones, you can setup relationships between a skeleton structure and faces on a polygon mesh. The system would let you specify how much each bone in the skeleton could move or stretch to match the polygon mesh. This meant we could deform the mesh with any other available nodes in Maya, and the bones would move and align themselves as best they could with the resulting polygon faces. We settled on using Maya's new nCloth system to do the deformation, as it was really fast and stable."
The resulting collision module worked well for leaf on ground collisions, and for a few leaf on leaf collisions. But whenever the number of colliding leaves reached a critical mass, nCloth would explode or leaves would get really tangled. In the end, effects artists had to hand tweak some of the more complex shots.
Working Out Mass Shots
For the bigger shots, RSP developed another system that was called "Vine Mass." The vine mass was similar to a fur system where a Maya plug-in called "Surface Scatter" randomly created points across a polygon mesh, and these points would stick to the mesh as it deformed. The team then created a set of Maya deformers that could be used to add aim and movement to the points. There was a system that would preview the leaves on the random points, so the animators could animate the surface and the deformer, and get enough feedback to get the required performance. Many of these setups were layered up to get depth into the mass of leaves.
"For dynamics, we exported the points from Maya into Houdini where the points would be extruded away from the surface into a dynamic hair," Humphreys notes. "We then could run dynamics on the hair to give the vine mass leaves some secondary movement and bounce. The points were then exported again to be read in by Venom, our scripted rendering framework based on Python. We created a Python script specifically for the vine mass. It would take the points that were exported from Houdini, and instance leaves at these positions during the render. We also used different Venom scripts to inject the hero vines into the renderer as well."
Additionally, RSP created some creepy shots in which the vine is seen infecting people under their skin. First, the team prepared a library of 2D and 3D vine animations. These elements were then tracked and warped over the performer's skin. Bulging, lighting and bruising effects were finally added in post to better integrate the animation.
"We came on very early, which was a lot of fun," Humphreys concludes. "Having some time to take a breath and regroup between the stages of production helps you fix some of the dirty hacks you do early on to get ideas out there, and finding the balance between doing enough to brainstorm ideas without painting yourself into a corner is a great skill. The industry is always changing, and we have to continually adapt to new requirements. Amorphous characters like the vine are really hard: it seems you never have enough time, but you do always find a way before you run out of it."
Alain Bielik is the founder and editor of renowned effects magazine S.F.X, published in France since 1991. He also contributes to various French publications, both print and online, and occasionally to Cinefex. In 2004, he organized a major special effects exhibition at the Musée International de la Miniature in Lyon, France.
