Computer Graphics World

use to move objects toward the audience or push them far away are the distance between the two cameras (one for each eye) and the convergent point for the two images. e choice of camera lens affects depth as well. "Let's say you want to use a long lens, maybe a 100mm lens," Neuman says. "at gives you an unsatisfactory result in 3D: ere is a large separation in depth between the characters from foreground to background, but each character looks like cardboard. To compensate, you can increase the interocular or interaxial dis- tance [between the cameras], which in- creases the internal volume in the charac- ters, but the gaps between the characters are also magnified." And that could quick- ly use up the stereo budget. To solve this problem, Disney and other studios use multiple sets of cameras—perhaps one for the foreground, one for midground, and one for the background—and dial in the depth to create the internal volume and roundness for the character. en, they composite the layers together. With that kind of control available, di- rectors can choose where to place charac- ters in scenes and how much depth to give characters based on the story they're tell- ing. "One metaphor we build on is equat- ing the emotional depth of a scene to the depth of the character," Neuman says. "We might increase the depth until we get a nice, round, full character for an emotional beat. A second literal metaphor we use is equating emotional separation to depth." For example, when directors want the audience to connect with a character, they might place that character on the audience's side of the frame; that is, bring the character out into the audience a little. When they want the audience to feel de- tached from the character, they might push the character back. To do this without changing other char- acteristics in the scene, Neuman sometimes uses another stereoscopic technique called the floating window: e stereoscopy crew puts a black mask on the edges of the im- ages and then varies the thickness of that mask between the two eyes to change the perceived location of the theater screen. "We create a virtual screen (a window) New tools from several software companies now help artists working in 3D. For example, Autodesk's Lustre speeds the work of colorists who are grading 3D films. The Foundry's Ocu- lar plug-in for Nuke helps compositors who paint one eye to create the second eye in the correct depth. Quantel's Pablo system aids artists matching the colors on the left and right. And, Tweak's RV software gives artists the ability to see images in stereo 3D on standard monitors within RV viewing software. The animation and visual effects studios are also creating 3D-specific tools, and those tools provide a glimpse into the state of the art for stereoscopy. "We've had to come up with a whole new workflow," says Matt Welford, head of compositing at Weta Digital, where work is under way on James Cameron's Avatar and Steven Spielberg's Tintin. "We use [Apple's] Shake and Nuke for compositing, which are both node-based. Initially, we made a left tree for the left eye and a right tree for the right eye, so we had two trees per shot rather than one, as in traditional films. To reduce the number of files and make file management neater for artists, we worked on a way to represent stereo images in a single image file. To do that, we formalized the names of the left and right eye within the EXR file format." Weta called the new format SXR, and worked with the OpenEXR community to make it broadly useful. "Over the last six months, we've seen companies starting to adopt, or at least support, this format," Welford says. "Now, within Nuke, for ex- ample, there's a left-eye and a right-eye button in a single file." "We have a whole suite of tools that take care of the nuts and bolts of stereo for the artists," says Phil McNally, stereoscopic supervisor at DreamWorks. "If an animator has a scene with a character running up a road straight to camera, we would want the character to feel distant at the beginning and literally get closer to us at the end. We give the artists controls so they can dial the stereo as they watch the performance." Artists can have a different stereo setting on every keyframe without worrying about distortions, and without knowing the nuts and bolts. "We also have something called the multi-rig," McNally adds. "In CG, it's easy to have three, four, five stereo rigs all pointing to the same scene in much the way we have multiple lights put- ting rim lights on one character, for example, but not another. Unless you're doing a 3D science project, why not have every- thing up for manipulation and artistic interpretation?" "We developed a camera rig that uses a results-driven paradigm. I can define how far out from the screen I want the closest part of November 2008 16 n n n n Stereoscopy

• Cover