Compositors work to solve vfx puzzle

Innovations boost 'Holmes,' 'Avatar'

Compositing isn’t the most glamorous of visual effects tasks, but it’s among the most essential.

It’s compositors who take the distinctive digital characters and headline-grabbing innovations and piece them together with greenscreen shots of the actors, smoke, water, fire or explosions, some of them computer-generated, some live-action.

As a result, it’s the rare vfx shot that isn’t assembled by compositors, who are typically among the last artists to touch a film before it’s delivered.

Complex shots often have 100 layers or more, which must be meticulously assembled so each object appears in exactly the right place relative to the others. It’s like assembling a jigsaw puzzle — with the pieces separated from each other in space.

For example, in one slow-motion shot in “Sherlock Holmes,” Holmes (Robert Downey Jr.) runs from a burning warehouse straight toward the camera. Explosions surround him. He catches fire. Downey was shot separately from the fire; to complete the shot, compositors spent seven months layering objects, explosions and flames onto the footage of the actor.

Traditional compositing meant working with each frame as a flat image, like a photographer retouching a still picture. Now, though, “The paradigm is shifting,” says Chas Jarrett, who supervised “Sherlock Holmes'” 850 vfx shots. Compositors set up each shot in their computers in three dimensions. They can view the shot from the side, revealing layer after layer of flat pictures receding from the camera, like flats on a stage set, with computer-generated three-dimensional debris, boxes and barrels flying between them.

“Having compositors working in 3D space with images and 3D geometry is faster than having the 3D team render objects at the correct depth,” he says. “And it keeps the 3D team focused on issues that only they can deal with.”

But Jarrett isn’t talking about stereoscopic 3D (S3D), like that seen in “Avatar.” S3D compositing is more difficult because those flat elements — a common “cheat” that saves time and money — look flat in S3D. Almost everything in the shot has to have depth. If there’s smoke, for example, instead of a hazy flat layer in front of a character, it has to be a 3D cloud with space carved out for the character.

For “Avatar,” Weta Digital had as many as 50 compositors assembling everything into final shots.

In stereo, we have a pair of images for every element that goes into a shot,” says Erik Winquist, compositing supervisor. “And, if the left-right disparity for each element isn’t correct, your brain explodes.” A mere half-pixel difference could mean one eye sees an object deeper in space than the other.

Tools gave the artists a starting point to help reduce the workload. “We had to start compositing as soon as we got shots in rough form so the lighting artists would have something to review in dailies,” Winquist says. “And then the shots would evolve and grow. When you’re creating an entire world, it’s important to maintain consistency from shot to shot, the brightness of god rays, the treatment of bugs flying around, the color of fog and so forth. It was two years of hard labor, but it was entirely worth it.”

Popular on Variety