In The Final Destination, the fourth installment of the popular horror franchise, Entity FX contributed an all CG — and stereoscopic — x-ray sequence for the film’s dramatic surprise ending. Visual effects supervisor Mat Beck talks to fxguide about how the sequence came together. Warning: this interview contains spoilers.
fxg: It’s definitely a surprising ending – what direction were you given, especially in terms of the x-ray look?
Beck: The production decided in post to give the audience more of an ending resolution than was provided by the last live-action scene. Our brief was to execute an ending that involved a considerable amount of mayhem, but with the stylistic remove of an x-ray look similar to that of the title sequence. Eric Brevig was serving as a special advisor and brought the team to Entity FX for a meeting. As it happens, we have some experience melding an x-ray look with a practical sequence, and also zipping in and out of “bullet time” while doing it. Producers Craig Perry and Sheila Hanahan Taylor asked us to strike a balance between explicitness of violence and a kind of schematic abstraction, to show just the right amount of blood et cetera.
fxg: What sort of concept work or previs did you undertake or look to for the sequence?
Beck: We started out with a cool-looking previs sequence that was supplied by The Third Floor. It was based on the original live action, augmented by some motion capture of real people. Mark Helfrich, who was cutting the sequence, laid out where the ultra-slow motion pieces were, linked by transitions at accelerated rate. We began by redoing the previs in our own system; it was necessitated by a number of factors: 1) We needed a heavier, more complex rig on the skeletons in order to control the various kinds of violence to be visited upon them. 2) We had a skeleton in house that we could more easily add hints of skin, muscle, etc, as necessary. 3) We needed a more precise hookup with the live action of the characters for the final compositing 4) We also needed more detail for vehicles and a closer fit to match the live action.
Most importantly, the action changed. The sequence of figures and what happened to them changed as the production worked on exactly the right sequence; a challenge was to make it clear to the audience what was happening to whom as the environment became more and more chaotic. So basically we continued the previs process that had been elegantly begun, only now the adjustments were in exactly the same form that the final render would be in, and the final action was achieved through keyframe animation. While refining the action, we were also continuing to refine the look of the x-ray look and submitted a number of tests before getting the feel that the production liked.
fxg: What plates or elements were shot?
Beck: There was a green screen composite of the characters on a set, a practical vehicle, and a street background that was added. We lined up to a finished composite that was done by someone else.
fxg: What was involved in transitioning from the live action to the ‘x-ray’ look?
Beck: We did a relatively simple transition into x-ray space with characters that lined up with real characters, as well as x-ray furniture, props, and a large vehicle that all started out lined up with the original photography. Everything had to be placed in the right spot in 3D space. We discussed doing a dimensional wipe through 3D space, but in the end a simple cut with a flash was actually better. Because the transition was such a quick visual shock, it gave us the opportunity to have more motion and energy in the B side.
fxg: Can you talk about the modeling, rigging and animation involved for the skeletons and other geometry?
Beck: We replaced the skeletons that were in the previs with our own skeleton rig that was also outfitted with muscles, nerves, etc. Ultimately, we ended up using bones, and just a hint of skin and flesh without specific nerves or muscles. The other components were deemed to provide too much detail that detracted from the schematic style that was desired. The original previs included motion capture data that drove the skeletons; we referenced some of that data when creating the final moves for the characters, but, as the moves changed and as the fate of two of the characters was swapped, all of the characters’ performances ended up being key framed. There was a lot of use of inverse kinematics to help the character interact with parts of the scenery.
We worked entirely in Maya, and conveniently the previs was in Maya as well. We used an x-ray shader that we had fine tuned in previous projects. Because it simulates an x-ray, there is no traditional light source in the normal sense; the bones were effectively self illuminating based on density and angle to camera. The brightness of overlapping bones combined for even greater density, much as a real x-ray would. We added a bit of fractal noise to the bone surface to help give a little volume and dimension to the bones; we also added some more density in the center of bones so that it was more than just an edge effect. We were continuously tweaking transparency to adjust how much detail was seen in layers that were further from camera. We used depth of field and stereo information to convey help convey depth.
The flying glass and blood did have source lighting in order to give modeling and cool-looking highlights. We did not worry about honoring the direction of the original lighting of the scene – it was not a concern. The glass was broken using a blast code simulation which was then tweaked for maximum stereo impact. Our vehicle was extensively modeled to give the sense of layers of moving parts. The tires and grill received a special amount of attention because they were elements involved in especially intimate (and messy) interaction with characters.
The energy of the impacts was heightened by manipulating time. Normal time with lots of quick movement repeatedly ramped into hyper slow motion bullet time which allowed the camera to lovingly dwell on the mayhem, only to zoom again to the next bit of crunch and snap. In order to allow for the greatest creative freedom, we initially rendered whole shots in hyper slow motion so that editorial could do their own speed ramps and give us a template for what to dwell on and what to zoom over.
fxg: How did you accomplish the blood effects?
Beck: The blood effects were accomplished using Maya 2009 nParticles, which allowed adhesion between the flowing blood particles as they spread. Because of concern about the film’s rating, we had to choose our moments with restraint. Not too much blood, not in every shot, no blood pooling on the ground. Those were some of the rules. We went with the aesthetic that the only saturated color in the scene was the deep red of the blood.
fxg: What were some of the stereoscopic challenges?
Beck: The first task was simply modeling the entire set and matching the inter-ocular and convergence that was built into the plates. As we moved into the purely synthetic world, we of course had complete freedom with stereo parameters – our initial guidance was to use a bit of 3D restraint and have most of the action take place behind the screen. Nonetheless, there were a number of selected moments in which we had elements extending out into the audience for impact, and we were generally able not to have them break frame awkwardly while doing so. We did use the occasional floating window to help with the frame line interaction. At the end of the last shot, we had a “tear your eyes out” impact moment in which something flies right at camera and we tweaked inter-ocular and convergence pretty extensively.