John Dykstra interview pt. 3

Martin Anderson

When we set out to make a film, we have to use techniques that we're uncertain of, or we will have an obsolete film when we're finished

Published on Nov 2, 2008

Continued from part 2

Were there new challenges on Hancock for you, or did you feel that it was all within your experience as soon as you read the script?
Well, it's always an effort to come up with a new look for a description that you've heard before. It used to be that the script invariably had a line someplace in it that was 'unlike you've ever seen before' [laughs]. That was a standard writer's euphemism for 'You've got to come up with something great for this!'.

It seems to me that the focus at this point has become interpretation of what's written on the page, as opposed to simply a mechanic of getting it done. It used to be that we had to figure out how to fit the camel through the eye of a needle. And now we can do that. So it comes down to whether or not fitting a camel through the eye of a needle is what best tells the story.

So in every script that you read, no matter how much material is in it, the challenge - at least, the personal challenge that I like to make - is to come up with something that's in the unique interpretation of that written word.

How does that manifest in a practical way in a film like Hancock?
It's an interpretation of the character. [Hancock] is an oddball character - that idea of a ner' do well superhero is kind of a strange combination of things, sort of an oxymoron. So the idea was to bring across, along with the personality of the character as played by Will Smith and directed by Peter Berg, to come up with some visuals that would interpret or evoke that same sense of an oxymoron.

I think that's true of the way he flew - his erratic flight patterns, his non-graceful positioning during flight and the whole kind of cadre of things that we had to come up with for him when he hit the ground. He made holes when he launched, he left debris behind…that kind of thing. It involved coming up with stuff that was fun and had a sense of humour to it.

Did it involve much collaboration with Will Smith, seeing as you're contributing to his performance and persona?
It's a collaboration in every way. The movie is Peter Berg's movie, it's his vision of what he wants to have happen there, and - as with all directors - he then collaborates with all of the people working on the film to come up with a way to express that visually. Understandably we used Will for as much of the material as we could, so that he was actually on screen interpreting the character. SO much of the material that you see is blue-screen of Will Smith doing his own interpretation of what Hancock would be like if he were flying or whatever. So he has that input to begin with. And that creates the paradigm for what we do for virtual versions of the same character, in terms of how he moves and positions himself…his head position, whether he looks or doesn't look.

If another effects company needs a virtual Will Smith in two years' time, is that a model that you're going to be selling on to them?
Never. Built-in obsolescence. When we set out to make a film, we have to use techniques that we're uncertain of, or we will have an obsolete film when we're finished. That's changing significantly with the advent of digital imaging, but the truth is that some of the things that occurred in the more physical world of optical printing are also true of digital imaging: you push to the limits each time to invent a new 'look', or a new technique to create a look. With computers, you write code. The code, just like the operating systems and the computers themselves, becomes obsolete at an incredible rate.

So the model that we created for Will Smith for this movie will probably not be applicable in two years...no, 'probably' is wrong: it will not be applicable in two years, because it will either be the wrong kind of geometry, or it will be what we call 'heavy' geometry - where you have to put in too many vertices to define a very detailed surface, and you create this geometry that's incredibly heavy. The parallel for that is that we used to make everything out of polygons, little geometric forms that were linked together. And then they came up with spline technology, NURBS, and that allowed us to make much more complex characters with smooth and detailed surface with much less computational resource.

And that's continuing to happen. This stuff keeps re-inventing itself, it keeps being put together in different ways - different rendering systems, different kinds of shaders, all that stuff, all those complex components. You've got to remember, you're creating this object from whole cloth. So it's going to change - it changes with every show that you do. And reverse-compatibility is not hot in this particular business. You can use old models, but you end having to limit what you can do with them, with the ability of the software that you're currently using, to interpret.

Could I ask what software was used, for instance, for Hancock? We're familiar with the likes of Maya and Lightwave, but I presume you use custom-made applications…?
Well, okay, you can ask, but I won't tell you [laughs]. No, Maya is the primary software that was used on Hancock. Renderman, I believe, is the rendering engine that was used, although we render different things using different engines depending on what the final look has to be on screen. There are very few facilities that do cutting-edge visual effects that don't write their own plug-ins, and one of the beauties of Maya and Wavefront, one of the advantages of their software is that the architecture's open so that you can create plug-ins to take what they have created as a foundation and develop your own architectural interpretation. Which means that you can do new ways of shading, new ways of animation and create your own GUI, so that you've got the elements and tools that you want your animators to work with. There are so many interfaces now, what with motion-capture and all the other elements that come into play, that you really have to be agile when it comes to software. Taking stuff off the shelf and having only things that will plug in, without interpretation, produces relatively obsolete material.