ILM’s Ben Snow Talks Pirates, Mermaids and 3D Pipeline
B.S.: Absolutely, yes. Working in stereo, you have to be a lot more careful in layout and animation. People get used to cheating. Of course we always try and beat that out of people, because if you are doing any sort of physically-based simulation, like a particle simulation or a water splash, the 3D space is important anyway because the way the character interacts with the physics of the simulation. So we had situations on Pirates of the Caribbean where we were match animating a body to a mermaid, and it just wasn’t working. We discovered that the head of the mermaid was actually two feet behind her shoulders somehow, and it was some cheat that had happened in the layout phase. The problem is that you really have to be on top of your quality control, because if it makes it into the pipeline, what sometimes happens is, the next artist comes along, and says, “Well, you know it looks like it’s working so I can just band-aide this on top of it,” and it actually goes quite far down the pipeline. And [then you find] this just isn’t working.
Once you open up the scene, oh my God, it’s two feet off the body, you know, and so, then you have to go back and re-do the work. We had to really be on top of that, and work with the artists and say, “OK, you have to quality control that, everything has to be proper, in proper space.” Animators can’t animate to cameras as much, so it [stereo] does make a big difference with that [animation]. Generally a rule of thumb was that lighting wouldn’t change a whole lot. Animation, we didn’t think would change a whole lot although the issue that I mentioned does add the fact that they can’t cheat as much. Layout, definitely it added and rotoscoping it added. Compositing, where it really got more complicated was when you were doing things like, you’ve got a water surface which is uneven, and it is going back in stereo space and you are adding three mermaids splashing around, and each of them have splashes, and you somehow have to make it look like those splashes are sitting on this uneven water surface. That created a lot of nightmares for people.
What oddly didn’t create nightmares was one of the issues that you get when shooting as we did with a mirror system. When one camera is looking down at the mirror, and one is shooting through the mirror, even just the fact that the eyes are offset, you get different reflections in different eyes. The polarizing filter, and the mirror, has an effect on things. So there would be shots of the mermaid coming up in a little sea pool for example, and you get reflections in one eye but not in the other. That is something we experience in real life and is certainly common in stereo photography. The tendency is to want to fix some of that stuff. But we found that when we started trying to paint out reflections that were in one eye and weren’t in the other, the water stopped looking wet. So, you actually want to leave in some of these imperfections.
D.S.: Were there any significant pipeline changes you needed to make to handle this film?
B.S.: I think the big thing was definitely the color pipeline, getting the RED camera data through our pipeline, how we dealt with that, how even we dealt with the facility that was handling the overall DI and the color correction. We did a fair bit of work with them to work out how to get the data from the camera set to us, what happened when we got the data. What we would do is bring in the plate say, “OK, is there any error in the stereo between the different eyes?” We’d try and correct and offset that. “Is there any color difference between the two eyes?” We’d try and correct that. We built a different front end to our color pipeline for the way we brought data on. Also, because we were dealing with an over size image that we were going to crop, [we] changed all of our display techniques and the way we would look at dailies and then the way we would give data back to the client. So the whole image pipeline changed quite a lot.