Alain Bielik gets the lowdown from ILMs Roger Guyett on M:i:IIIs stealthy but significant CG work.
The first one was a stylized spy movie; the second one was pure action extravaganza. For the third Mission: Impossible movie (released by Paramount on May 5), producer Tom Cruise and director J.J. Abrams set out to firmly ground the story in reality. This time around, you wont find a helicopter chasing a bullet train in a tunnel, or any acrobatics on motorbikes. In M:i:III, the action scenes needed to look absolutely believable.
For lead visual effects vendor Industrial Light & Magic (ILM), this meant invisibly enhancing the action in some 550 shots. This was especially true for the nighttime sequence in which agent Ethan Hunt (Tom Cruise) jumps off the rooftop of a skyscraper in Shanghai, China. Shooting the stunt on location presented major problems in terms of visibility, mainly because of haze and pollution, recalls visual effects supervisor Roger Guyett, who oversaw the project with visual effects producer Shari Hanson. So, we decided to recreate Shanghai in the computer in order to produce as spectacular a backdrop as possible for the sequence. Tom Cruise really jumped off a building, but it was a set piece built on top of the façade of a parking structure at Universal Studios. The entire Shanghai environment was digitally added later on.
Building a CG Shanghai
To start with, associate visual effects supervisor Russell Earl and a small crew spent four weeks in Shanghai photographing all the downtown buildings in high- resolution. The high dynamic range images were photographed at night with long bracketed exposures, as to capture the trademark neon lights and signs of the Chinese city. For each shooting session, the Canon digital camera was placed on a tilemaker, a programmable computer-controlled head that moved around a few degrees after each shot. Still after still, the camera captured the entire environment of each location at 360°, building up a spherical representation of Shanghai from each particular point of view. Since we were shooting with long exposures, the lighting conditions had changed quite a lot by the time we got to shoot the last image of a series, Guyett adds. When you looked at the stills, you could actually see the level of haze or pollution changing the dynamics of how far you could see in the city. The visibility basically changed from one day to the next, and some time, from one hour to the next! So, we had a lot of grading to do before we could actually tile those images together.
The tiled photographs were complemented with hundreds of additional stills of the hero buildings at closer range. The last step consisted of shooting high definition footage of all the main locations, as to capture the moving elements of the city. Since traffic was very limited in Shanghais office district at night, the ILM crew filmed busy streets in other areas of the city and later dropped those images into the main plates, creating the requested lively backdrop.
Back in California, the Canon raw images were converted to 32-bit EXR files, ILMs new open format. Using in-house photogrammetry tools, the CG department led by Gerald Gutschmidt and Patrick Tubach took those images and managed to recreate the basic 3D geometry of each hero building. Architectural features were then added by projecting the Shanghai images onto the corresponding 3D models, a task that was accomplished in ILMs proprietary Zeno: a hub that handles file conversion for multiple applications. The result was a three-dimensional high-resolution matte painting in which the camera could move around.
Using Maya and RenderMan, we first did a rough layout of the city, deciding what we were going to keep, what we were going to delete, and where the hero structures would be located, Guyett notes. Based on that layout, we put more efforts into the modeling of the buildings that were going to be closer to camera, as to obtain different levels of parallax whenever the camera moved. One of the advantages of creating the sequence in the computer was that we could re-organize the city, so that every angle revealed a spectacular backdrop of Shanghai. Using matte painted elements, we also often modified existing buildings to add a sense of danger to the scene. The various layers were mostly combined in Shake, but lead compositors Mike Conte, Robert Hoffmeister, Sean Mackenzie and Todd Vaziri also used ILMs proprietary CompTime.
In the trickiest shot of the sequence, the airborne camera travels across the Shanghai skyline and ends up on Cruise, perched on a rooftop, high above the city. Almost a minute long, the ambitious shot actually combines a helicopter plate of Shanghai, a digital representation of Shanghai downtown and a Spidercam plate of Cruise shot at Universal Studios. We first previsualized the scene, and then shot the plate in high-definition in Shanghai, flying the camera helicopter as close as we could to the building on which Ethan Hunt was supposed to be, Guyett explains. Then, lead matchmovers Giles Hancock and Terry Chostner took the last part of that move and extended it, using Zeno, with a CG version of the skyscraper and its surroundings. Finally, we took that digital camera move and used it to shoot the Spidercam plate of Tom Cruise on a rooftop set piece, which gave us the last portion of the shot. It took us almost five months to put it all together and create the illusion of one continuous shot.
Destroying a Landmark Bridge
ILM also created the entire environment for another key sequence of the movie: the attack of a convoy on the Chesapeake Bay Bridge on the East Coast. When shooting at the real location proved almost impossible, the filmmakers opted to recreate a 600-foot long section of the bridge in L.A. The massive set was built six feet off the ground in a location surrounded by a low horizon. This particular set-up provided plenty of opportunities to shoot the real sky as the background and thus, to get many shots in camera. Whenever the camera angle revealed the actual fields surrounding the set, the ILM crew simply blocked the view by raising a series of computer controlled moveable green screens that covered some 400 feet along one side of the bridge. Designed by Guyett, the rig could be raised or lowered within minutes and proved extremely useful during the shoot.
Once the plates had been photographed, Guyett and crew calculated the camera angle and time of the day of each shot. We went back to Chesapeake Bay Bridge and, using helicopters and various pieces of equipment to match the original camera angles, we photographed the background plate of each shot with the bridge and the ocean. We first thought that we were going to create the environment in CG, but once we got to the bridge to get reference photographs, we realized we could just shoot real plates. The bridge was built about 80 feet off the water, which meant we always saw the ocean from a distance. It turned out to be really easy to roughly match the original camera angles with a helicopter. Also, from a budgetary point of view, the cost of renting a helicopter was significantly lower than the cost of building the complete environment in the computer for each and every shot. In the end, this 2D approach proved surprisingly simple and effective.
After the foreground and background elements had been combined, animation supervisor Paul Kavanagh and his crew added many cars to the bridge, and augmented practical explosions with digital smoke, fire and debris created in Zeno. The main shots of the bridge explosions involved building very complicated, three dimensional matte paintings what I called digimattes, Guyett continues. Its one of these sequences where the line between a 2D matte painting and a full-on 3D environment is becoming very blurred! We basically built a simpler version of the bridge and painted all the elements directly onto the model to create an enormous amount of detail. We then used Mayas rigid body dynamics to create the destruction, combining it with the practical explosions that had been shot on set. Finally, we also added CG rockets and a digital unmanned aircraft. The whole sequence was a huge combination of many different elements, and it was of paramount importance that all our work was absolutely invisible. That was probably the hardest thing to do
Flying among Windmills
The other large-scale action sequence in M:i:III involved a helicopter chase in a windmill farm in Palm Springs, California. Most of the nighttime sequence was shot for real, with ILM adding the necessary destructions and extending the farm with CG windmills. We also did all the green screen composites featuring the actors in the helicopters, Guyett adds. This type of shot is probably very uninteresting to your readers to hear about, but it still is one of the hardest things to do in visual effects! You have a group of actors shot on a soundstage and youre trying to make it look like theyre truly flying. Some time, creating huge destructions is much easier to deal with. The simplest things often prove to be the most difficult to make look real. In this case, it was all about the subtle changes of lighting that occur when youre traveling. I was very pleased with what we ended up doing. One of the helicopters had a searchlight that we used to dramatically change the lighting conditions in the other helicopter. This simple trick did an awful lot to add excitement to the sequence.
In the most spectacular shot of the sequence, one helicopter is cut in half by a windmill. The chopper was actually digital right up to the moment of impact, at which point it was replaced by a 10-foot-long model. Positioned exactly at the same angle, the miniature was hit by a windmill blade model and blown up in camera. Compositors then created a seamless transition between the two helicopters.
Behind the Mask
On top of its many duties on the action sequences, ILM also tackled Mission: Impossibles signature effect: the mask reveal where one character removes a mask to reveal its true identity. In a shot that required no less than five months of work, Cruise puts on a mask and becomes villain Philip Seymour Hoffman. We first shot Tom putting on a prosthetic mask of Philip that had been created by the make-up effects department, Guyett recounts. Then, we got Philip to match Toms movements and, finally, we created a CG mask to blend between the two. The whole trick was to hide the transition between Toms practical mask and the CG mask, and then, between that digital prop and the real face of Philip Seymour Hoffman, all in one shot. It required an enormous amount of modeling, rendering, and compositing to make it happen.
For the 180 ILM crew members who worked on the movie, M:i:III presented some unique challenges and great opportunities for signature shots. There was an incredible range of work, from the reality-based effects of the bridge sequence to the challenge of making the mask effect look authentic, Guyett concludes. Plus, Im very happy with the image-based rendering technique that we used for the Shanghai sequence. Its a very powerful tool for a director to be able to recreate an entire actual location in the computer. It allows him to shoot the plates of the principals in the controlled environment of a Hollywood studio without ever having to worry about the logistics of a highly complicated shoot at a foreign location. J.J. Abrams could obtain shots that would have been impossible to achieve at the real location. It was a real asset for the production.
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 and occasionally to Cinéfex. Last year, he organized a major special effects exhibition at the Musée International de la Miniature in Lyon, France.