All about ILM’s adoption of several new technologies back on JP3, including ambient occlusion, virtual sets, Maya Paint Effects and dynamic flesh sims.
So groundbreaking had the CG breakthroughs been in the original Jurassic Park and then subsequent films worked on by Industrial Light & Magic that the release of Joe Johnston’s Jurassic Park III in 2001 did not arrive with such a huge degree of VFX fanfare. Yet, in many ways, it heralded several major steps forward for ILM and digital visual effects.
The film was, of course, full of major effects work, not least of which were some meticulous ILM miniatures for the thrilling light plane crash, as well as Stan Winston Studio’s animatronic dinosaurs. But it was also ILM’s adoption of several new technologies that made Jurassic Park III a game-changer.
It was here that ambient occlusion – first developed on Pearl Harbor – was adapted to be run per-frame on animated creatures and with heavy displacements to provide far greater detail on the dinosaurs. It was here that ILM tackled dynamic flesh simulation for the creatures. And it was here that foliage simulation (care of Maya’s Paint Effects) and full virtual environments were also relied upon.
A few years ago for vfxblog I sat down with Christophe Hery, a CG supervisor at ILM on Jurassic Park III (and now research scientist at Facebook Reality Labs), to discuss some of these important visual effects developments. Here’s the piece republished for the 20th anniversary of the film.
b&a: Tell me about ambient occlusion on Jurassic Park III. What was being done in lighting at ILM right before this?
Christophe Hery: Well, previously, we had point light sources that we were just placing judicially. Reflection occlusion is something that we had worked on. Dan Goldman and Stefen Fangmeier had worked on that prior to Jurassic Park III, including on Speed 2. So then ambient occlusion was the next step, and it was led by Hilmar Koch, Ken McGaugh and Hayden Landis on Pearl Harbor.
They came up with the idea of extending reflection occlusion essentially to a full diffused hemispherical domain. It was a push from Joe Letteri in many ways when he was at ILM. On set, we would acquire the grey ball and the reflection ball, and we had had tools for years to extract them and try to do things with them directly. But ambient occlusion is what made it reality where all of a sudden we could directly use the map and drive diffuse and have semi-correct shadowing, essentially, or self-shadowing. That was a big breakthrough for those guys for Pearl Harbor. And then in parallel to when Pearl Harbor was still going, Jurassic Park III started. I was CG supervisor, with Jim Mitchell as the main visual effects supervisor on the show.
I grabbed the idea of this ambient occlusion technique. It was much more of a leap of faith, what I had to do. They were doing it statically once, ie. for a static model for the planes in Pearl Harbor. For Jurassic Park III, I was running it through animation per frame and also with heavy displacement. I had decided that we would put way more detail on the dinosaurs, so we had a lot of displacement maps. There was a lot of questions about is it going to be stable in animation? Is it going to be acceptable? Is it going to react properly to displacement?
Plus, we were writing the ambient occlusion pass in Mental Ray at the time, because there was no raytracing in PRMan yet, which we were doing as a pre-computation. We would merge all this in RenderMan – we would read the pixels as a texture, and then just multiply the illumination by the darkening of those shadow values.
b&a: There was a signature shot, wasn’t there, in Jurassic Park III, what seemed to show all this off. And that was the Spinosaur in the water with the fire?
Christophe Hery: Yes, well, it went further than that because the creature was wet, obviously. It’s not just diffuse, it’s not just ambient occlusion, because we really had to run reflection occlusion as well. We were essentially running the plates as a refection map on him. This is, again, prior to raytracing, and prior to physically based lights, and shading models. I used the plate and extracted the section that was interesting where I would see the fire growing, and re-projected that on him during his specular evaluation. It’s not just ambient occlusion. It’s perhaps in the forest shots where there’s more of a canopy of trees, it’s more diffuse lighting, which are much more representative of what ambient occlusion is. For example, the Spinosaurus against the T-Rex.
b&a: What do you remember as the reaction inside ILM, or from Joe Johnston, about this step up with ambient occlusion?
Christophe Hery: Well, Joe Johnston, obviously, is an old ILM guy. He was storyboarding and all that back in the day. He appreciates all that stuff very much. I think his main goal artistically was – well, first, they had discovered that dinosaurs were not really reptiles and that they were more like birds in a way. And so there was this potential that they had features and likely even had colours on their skin.
I think that’s the big contrast from this perspective between Jurassic Park: The Lost World and Jurassic Park III. So, in number three, the dinosaurs have way more textual detail, both in terms of colours and in terms of displacement on them. That’s what Joe was pushing. So the ambient occlusion, the reflection occlusion, all these techniques were there to enhance that. I think from Joe’s perspective, because we had the detail there, we were then able to pick up lighting from every direction. He didn’t really know how it was done, but the fact that we preserved this detail was crucial to him. I think it helped overall.
b&a: Another advancement were the skin and muscle simulations. How wer you actually doing flesh sims and secondary motion at the time?
Christophe Hery: The skin simulation was much more subtle. We did it especially on the raptors. You could see and test the muscles rolling underneath the skin. For Jurassic Park, they had, in Softimage, what they called secondary enveloping. It’s kind of skinning – they had deformation – but for the secondaries they would put some high level controls. There would be a primary deformation that would happen from the IK bones directly. They would preserve some sort of smoothness and this was a lot of ad hoc things that were put in for every point, trying to react to deformation at the time.
The secondary motion would take those already enveloped points, and apply another layer of enveloping on them, another set of weights that were much more general, that were smoother. The animators would jiggle by hand, essentially, some spheres in the belly. It didn’t respect any topology or any of the way the skin would be moved, or any muscle anything, but it gave this impression that there was this added ballistic that they were adding to the skin manually, mainly for heavy creatures.
They used this approach a lot on the T-Rex in the first movie, and I think that technique lasted, really, up until Jurassic Park III. By then, scientists such as John Anderson and Sebastian Marino had been doing a lot of simulation work for ILM. And it was about writing solvers. A lot of it is about meshing. It’s about how can you create a mesh of points with the correct springs between them and have that stabilise and not explode during the simulation, so you have dampening, you need to put all those forces so that it doesn’t continue expanding all the time.
We figured out a way to run that meshing process on a simplified model of the dinosaurs. We had simplified muscles as well that were physically modelled and that mesh was attached with springs to those muscles. The ballistic now was automatic, in a way. The animators wouldn’t jiggle anything. They would run their animation and then we would do a pass of simulation at that stage. Now, the rig for animation was more complex because, you normally animated the bones that would deform, the IK bones that deform the skin directly, but now, it would deform the muscles. Then, the skin would be attached to the muscles, and jiggle accordingly, and react to that.
That, then, was a big a leap as well. It was fairly subtle to some people, even internally. I think it was technology that was very sound and something that other people have done more on since then. Now we start modelling from the inside out. You see that a lot in ‘making of’ movies now where people show actual skeletons, and then the muscles, and then the skin. In the past, this was more or less a cheat, now it’s the real way you do it.
b&a: Another thing I had forgotten about Jurassic Park III was the use of digital foliage to help with the interaction of the dinosaurs and plants. Can you talk about how that was achieved?
Christophe Hery: In Maya at the time there was this thing just starting called Paint Effects. We basically started looking into, how can we anchor the creatures better into place? In the past, this would be done by running mattes essentially that would cut out the legs or the feet and then we would comp over actual real grass from the plates.
So the animal would never step on the grass or deform it, really, unless it was rigged that way, but it’s very complex because you would have to plan your animation a crazy amount ahead of time for all that stuff. We did the opposite. We had a plate, we had the grass, and some elements. We essentially covered it all with Paint Effects in Maya. It was more or less an experiment at first where we were saying, ‘Can we drive the grass?’ Essentially we had anchors where the feet would go from the dinosaur because the animation would go, we would extract that position and use that as a force field essentially on the brushes in Paint Effects and deform that. Chris White who is now at Weta Digital was the one who developed a lot of that stuff, and I was supervising him for that.
We started as an experiment – ‘can we do it just sparingly?’ – just to add a little bit. We ended up covering a tonne of stuff at the end! In fact, just like we had covered a lot with dust in the past, because the dust is much more forgiving in that aspect, but now we were more up to the point where a lot of plants were regenerated at the end, like that.
b&a: There’s digital plants, too, in that aviary canyon sequence, but that was also where you had a virtual set, wasn’t it?
Christophe Hery: Yes. There was a plate, though. Now, I don’t know if you have been to the Universal Studios tour in LA, but not too far from the Psycho house is this large bubble of cement, a half dome that you see from the outside. A very famous movie that filmed there was Jim Carey’s The Truman Show. At some point you see him sailing and he hits this wall. Anyway, they used that for part of the aviary sequence in Jurassic Park III. It’s partially submerged, there’s water at the bottom of the canyon. They had dressed some of it to look like a canyon. So the opportunity for movement was very limited and you obviously could not do big aerial camera moves in this studio space.
But after that kind of semi-success with the Paint Effects plants in Maya, we thought, why don’t we create, with fractals, some base geometry for the cliffs and start doing procedural generation of rocks and different colours to try to represent the canyon. Then, we’d add plants on it. We used the same sort of technique from Paint Effects, but we dressed physical geometry in with proceduralism and instancing.
We painted some density marks and some regions where we wanted plants to be. Most of the canyon that you see is CG, and it’s also under heavy fog. The pteranodons are flying through the fog, so I had to write a technique to render volumetric fog. It was very simple, compared to what we do now, it was nothing. I was driving a series of spheres. In each sphere I would run some different fractals that would intersect each other. I would make some fake fall-off from them so that you wouldn’t see the intersection between them, then I could animate them, so I could give the noise a little bit of movement. The camera would fly through them, but they would also be floating around in a way, like clouds, or banks.
There’s also a guy parasailing in there, and he was a bluescreen element, but pretty much the rest of the environment was CG. There’s nothing from the plate except for him, who was a bluescreen element that we put on a card to animate through that as well. I think that’s a pretty successful first from a CG environment I think.
b&a: I guess this essentially was a virtual set – at the time, was that something people recognised as a step forward?
Christophe Hery: I think Jimmy Mitchell, who had been at ILM before I was, he had a clear vision about that. He sensed that it was possible to do it. He trusted me through it. I assembled a team for it, and I developed the tech and I don’t know if there was even a plan B. Had we failed, I don’t know what would have happened.
b&a: Perhaps a further aspect to all of this was a confidence in the fact that CG could hold up in a film, even though there were also practical things, like the animatronic dinosaurs, to shoot with.
Christophe Hery: Yes, there were a lot of builds from Stan Winston. It was amazing to see those things and that gave us a lot of reference. They had shot them in the plates, but for a lot of the shots, we did replace them. We erased them and replaced them. They were present in the sense that they were an amazing reference.
I remember there’s a couple of shots in the forest as well, in the jungle where there’s maybe four or five raptors coming together and sort of communicating to each other. I think in those shots the ones that are in the background who are not moving are still the animatronics. All the other ones were covered back in CG. This is mostly because we had reached a point with the displacement with the rigging where we could free the performance, You could get more breathing happening in the dinosaurs, more expressions, more blinking. The practical dinos could do amazing stuff, but they they would have to stand still for the most part on the four legs, or on the two legs, they couldn’t move forward, not easily.
So there was a lot more energy I think in this film because of the CG dinosaurs. Some of the big creatures like the T-Rex and the Spino, they had a full scale model. I went on set and they were amazing to see, amazing. They were remote control, essentially. With those you could get some movement, but you could barely get walk out of them. A walk would have to be done probably with somebody in a costume. I think some of the pteranodons at the beginning of the aviary sequence, when you see them walking through the mist, when you’re really close up, those are someone in a suit.
I had a game, by the way, about this. I did a presentation, where I asked the audience is it animatronic, or is it CG? Among all the shots, 90 per cent of them were CG. It was amazing what the practical guys had done, but it’s hard to get the right amount of motion necessary compared to what CG can give you in some ways. But those animatronics gave us the grounding for lighting big time. We also had scale models that you can still see at ILM. Those were used for painting reference when we textured the CG models.
b&a: Were there any other developments from Jurassic Park III that you wanted to highlight?
Christophe Hery: Crowds. In Jurassic Park III, we did a few hordes or crowds of animals. We were driving through them, or flying over them in helicopter shots, and these CG dinosaurs were simulated. You’re driving through helicopters, and those were simulated. We had worked this out in Episode I, where we used Maya to drive animation cycles. We used that and we added a layer of simulations to the dinosaurs so they would actually recognise the terrain. They were driven essentially by a force field and by running cycles that they had. The terrain would influence how they would bank, potentially left or right within their body, left or right when they turn.
b&a: It did feel like in the early 2000s there was this major turning point, including in Jurassic Park III, where CG work could really be thought of as holding up in a big way. Is that how it felt back then?
Christophe Hery: We sensed that, even within the company. Somebody like Jimmy Mitchell, who had been at ILM for a long time, it was natural to him. He saw this evolution, it was pretty clear. We had an actual discussion, I remember, with Joe Johnston on the stage. Those practical creatures, again, are amazing, but it slows down the shoot. You need to reset them, they’re heavy machines. Once they’re wet, it’s terrible, they can’t move anymore. There’s a lot of conditions around them – a lot of people are working to make them work. Then, you wait for them to be ready, essentially. There were a couple of times where we asked Joe, is it really useful that we even do use them in the shot? He told us, at the time, that even if he was getting one or two shots of a practical animatronic, that that was enough of a pay-off for the practical animatronic. I don’t think it would be like that today.
Illustration by Maddi Becke.Become a befores & afters Patreon supporter.