The fabled Medusa was created by Framestore, with the biggest decision being how to animate the 50-odd snakes in her hair. Considering the short post schedule, they thought the best solution was keyframe animation. Even so, it was very time-consuming and difficult because of "the collisions and colliding and crossing over within the snakes." Actually, she was based on Russian supermodel Natalia Vodianova. "We did a photographic texture shoot of her, we did body and facial scanning, some basic performance capture with her and then, from that, she became the basic model that we created Medusa from," Davis explains. "And then on top of that, we wanted there to be a change within her when she goes into her petrified state, so we built a secondary model of her face that becomes a lot more demonic with her brows being pushed and her eyes blazing and her teeth becoming snake teeth and the entire texture of her face changing into more reptilian scales."

For Mount Olympus, Framestore was pulled into a completely different design. "One of the challenges of Olympus is that we shot on a set that in post-production everyone decided that they weren't that keen on anymore," Davis remarks. "We came up with a concept of the floor being a map of the world with moving clouds on it. This makes the Gods glow and it's much more of a surreal environment. We took the artists off the original set and set them in a new environment and actually rotoscoped every single shot. Framestore had to redo hundreds of shots in a few months."

And what was Davis' reaction to Warner Bros.' last-minute decision to go 3-D? "We were in shock. Here we were getting near the end of post-production, where we were in the thick of it with reshoots and changes, and, suddenly, we had this announcement that we were gonna do it 3-D [in less than 10 weeks]. So it was a bit of a challenge thrown everybody's way. But I have to say it's been a very educational process. I've been involved for the last couple of weeks, working here [in L.A.] with Prime Focus and this post stereoscopic conversion definitely has its place and has a big future."

Still, there's a lot riding on this conversion, given the industry's sudden appetite for everything 3-D. Yet despite all the carping about Clash sight unseen and post conversion in general, Prime Focus has an opportunity to distinguish itself with its proprietary View-D process.

"We really attempted to keep the original cinematography intact in every way that we could," Bond emphasizes. "They wanted depth, but not in your face gimmickry, and didn't want to further miniaturize creatures or environments.

"In Clash, actors have fly-away hairs and tassels on their clothing, and what I've seen with other conversion techniques is a tendency to cut these things out. But we've kept all those pieces so you'll have a foreground actor [with fine detail]."

With View-D, Bond says there is an underlying depth generation system, which is very fast and offers lots of iterations. "There's not a huge calculation time and it's very deterministic, meaning that it's consistently repeatable, so that on this film, in order to work on it in the timeframe we had, we worked on scanned plates and received graded plates later and ran them through the same tools and got the same results. So we were able to work with different pieces of material during different stages of the post-production process and not lose a lot of time. And there were days when we saw 19 minutes of footage. But normally we got to the point where we were handling 10-12 minutes of iterations a day. It's a very efficient process and we hope to make that even more streamlined."

Clash was split up into three phases: selection, depth generation (conversion) and convergence adjustment. Selection, which determines what elements are worked on in 3-D space, was spread out among Prime Focus' offices in Vancouver, Winnipeg, L.A. and India. In fact, the bulk of this phase was done in India, with a peak of 200 staffers, because of the extensive crowd sequences requiring roto work to maximize dimensionality. Depth generation, or the actual View-D step, was done completely in L.A., with a staff of around 40 making use of three stereo screening rooms, a Dolby Digital 3-D system and two RealD systems. Convergence adjustment was also done in L.A., with a couple of people overseeing shot finals plus 10 coordinators, a producer, a production manager and an editorial team of five.

"We'll have to leave it up to the audience to see whether they enjoy the experience," Davis concludes," but I have to say that I'm very impressed with the volume that they're able to create and the speed with which they've been able to do it."

Bill Desowitz is senior editor of AWN & VFXWorld.