Scanline's founder and vfx supervisor discusses the new mindset for working with Clint Eastwood.
After raising the bar for the wild fluid sim ride on 2012, Scanline VFX had to shift gears for the more real world demands of Clint Eastwood's Hereafter. Stephan Trojansky tells us what new challenges this entailed.
Bill Desowitz: What has this experience been like?
Stephan Trojansky: It's been a crazy ride. I never thought this would happen at all. When we finished the movie, this seemed like a really good piece of supporting visual effects for a Clint Eastwood movie that's very emotional about finding yourself and thinking about the hereafter. And suddenly there was this big bombastic sequence at the beginning of the movie -- a tsunami -- and we delivered it. But when it came to the awards season, we never thought of this as a visual effects movie. And then when we heard we were in the long-list, I was like, "Oh, my God! Well, somebody saw something in it that needs to be added to the plate." And then when we were still in the running, I thought, "A few more people think it was an interesting piece of visual effects work." And then when we were in the bakeoff, I thought, "OK, we're not like a $200 million movie, and the only thing I can emphasize is that we did a lot of difficult work to actually create supporting visual effects for a movie."
BD: And here you are.
ST: Look, it's refreshing and I'm happy that it shows that it's not only about the budget and the size and the number of people and the time that you have. It's an opportunity on a normal movie to do something exceptional and get rewarded for it.
BD: This is a direct consequence of expanding the nominees to five.
ST: It shows how a diverse range of visual effects affect today's movies with cutting edge level of technology. That's how I see and what makes us proud, independent of the fact that it was the only movie that was actually created by a sole visual effects vendor and it means a lot to us because all artists live for this.
BD: Let's talk about what makes the tsunami sequence exceptional, in light of your previous work with fluid sim with Flowline.
ST: We developed Flowline and improved it in ways to control the behavior of the physics of water. But to do something that it could not do in the real world. But now, in Clint Eastwood's film, it was suddenly a task that existed in the real world: the 2004 tsunami, where the speed of the wave, the impact and the locations had been seen on TV many, many times. So what I realized was this time we needed to match the real world and need to use our control to match to the live action. And that became clear when Clint decided he wanted to have the most realistic and emotional performance of the actors and didn't like going into a tank in a greenscreen studio where everything is safe and you can switch the water off when needed. We had our original actress [Cecile De France] for the hero shots, so we did some close-ups of tank footage in Pinewood Studios in London, but then when it came to the shoot in Lahaina, HI, Clint had the idea of shooting in the ocean directly. Basically, it was just the camera, the ocean and the actress and it immediately felt so different from a studio shoot because she was struggling in the waves and even the camera was so natural because it was swimming in the water, too. And it created genius plates, but for the integration later into the tsunami, what we had to do was have water in the foreground that was live action and the farther we get away from the character, the more it had to be CG water. And ultimately this was water that was rushing down the street and was interacting with buildings and cars that were stuck and would be swirling around.
So this time we had to teach the water to behave exactly like the water in the plate. So it was exactly like creating a CG double of the water and we used all these control techniques we had developed in the past to actually constrain the water motion to reproduce exactly what the water did with the actress. That way the motion vectors of the flows were the same. And then we would loosen up the constraints when it was like 10 feet away to create a water simulation of what it would really look like down the road. And in compositing they could just blend it together. And the integration was all done with rotoscoping because it was in the open water. We all believed it kept the emotion of the actress real.
BD: And the biggest challenge was having a different mindset for this integration?
ST: Yes, we did a kind of reverse simulation where we said, "OK, if at the end of a shot, we have the tree hitting here in the CG set, then where does the actress need to be in the beginning so that she ends up there?" So we took like 10,000 tennis balls and threw them into our simulation and would follow their path and until we would find the tennis ball that actually hit the tree. So let's follow the ball backwards from where it was dropped into the water. Then we knew if we linked our actress to that tennis ball, then she'd have the correct motion to ultimately end up [where she needed to be]. And then we would combine this with the workflow we formally had developed already where we would do this on low-res, very fast feedback simulation, but then up res our simulations until they were more and more detailed and we would get the spray and all the fine scale ripples and water structures so that it looks photoreal.
BD: And it's all about the harrowing performance.
ST: Yes, when Michael looked at it, he realized this is all about her -- the actress -- in the water. And we see way too much of the destruction in the background. And in reality, it would be out of focus. So for a moment, we had multiple versions where everything in the background was blurred, and the whole team was disappointed because you couldn't see all the work that we did. And, ultimately, it didn't end up being de-focused in the background, but it showed that it wasn't all about the big bang: it was just there to support her emotional near-death experience in the water.
Bill Desowitz is senior editor of AWN & VFXWorld.