Ask the Expert

The purpose of computer graphics in the movies is not to just duplicate reality. When a filmmaker sets out to make a movie, their goal is to tell a story and to express their creative style. Sometimes that requires realistic graphics, but sometimes that is better done as a cartoon, or with effects that bend the rules of reality, or by creating a scene that is more ideal than real. I often find myself explaining this to students, who start their work thinking that duplicating what we see is the ultimate goal of our field.

Before joining Stanford I worked at Pixar, the maker of Toy Story and Cars. Their creative team has been very successful by defining their own animated style. What is memorable about their films are their characters and stories, not the technology.

Of course it is still useful, and intellectually interesting, to develop technology to make things look real.

My favorite uses of realistic computer graphics in the movies are in situations where they help recreate the past. A particularly good example is the deadly dinosaurs in Jurassic Park. Another good example is in Titanic, where the computer generated effects gave you a visceral feeling of what it would be like to be there as the ship sank. Without computers we could never realistically reproduce such amazing creatures or places.

Among the techniques I have developed are methods for making natural-looking skin. This is a tricky challenge because, as it turns out, until you really get it exactly right, computer generated skin looks really creepy. This frustrating phenomenon is called the “uncanny valley” —as you get closer to imitating a person, they become more and more eerie.

Sometimes creepy is okay. The skin of Gollum in the Lord of the Rings trilogy was eerily realistic, but Gollum was supposed to look sickly. It is more challenging to simulate skin so as to make a character attractive. Using subsurface scattering, a technique we developed, I think it is safe to say that we can now make attractive skin.

Another area that I work on is lighting simulation. Lighting simulation is one of the most technically challenging areas of computer graphics because we have to simulate photons (light particles) as they stream around the environment from light sources to the camera. Lighting simulation can consume massive amounts of computer time. But lighting is also an art. Lights are used to emphasize the main character and set the mood. To achieve these goals, the cinematographer positions lights to make things look right. The lighting on set, in a sense, is fake. One of the biggest challenges in computer graphics is giving artists control over the simulation. Artists want to be selective about reality. For instance, they might want a scene to be lit realistically, but without inconvenient shadows. So when reality gets in the way of telling the story, they want tools to bend the rules of physics.

The idea that motivates me more than “realism” is “visual complexity.” This is the idea that everything we see has all sorts of amazing detail. Real materials have complex patterns. Wood, for example, is not mirror-smooth. Instead it scatters light in characteristic ways. If it is outside, as in a fence or a bench, it should look weathered and dirty in places. All the leaves on a tree aren’t the same color or shape, but instead some are yellowish, and others have been chewed up by bugs. Every time I look at something, I imagine all reasons that cause it to look the way it does. The power of computers is that we can model all the complex processes that underlie what we see.

When I helped make the rendering below in 1998, this scene was about as realistic as we could make it. Yet, looking at it closely today, all the plants are too perfect. It is clearly fake. There is still a long way to go.

It is up to each artist to decide whether realism, idealism, or outright fantasy will be the best style for telling her story. The goal of our work is to give them the ability to employ computer graphics to suit any of those choices.