ILM is on fire with Deepwater Horizon


You might not have seen Peter Berg’s Deepwater Horizon when it hit cinemas. But I’d implore you to check it out, if only for the remarkable combination of practical set photography, fire simulation and compositing involved in depicting the catastrophic – and real life – oil rig explosion that occurs in the film. I had a long conversation with overall and ILM visual effects supervisor Craig Hammack about the vfx effort in which we discussed what it really means to capture real fire, how to simulate it digitally, how some of the key shots were achieved, and about snaffling the last remaining LED wall screens in Louisiana for the shoot.

vfxblog: I’m curious as a visual effects supervisor what your first conversations are like on a film like this?

Craig Hammack: The very first conversation I had about the film was a meeting with Pete Berg where we had brought down samples of fire from other movies – CG fire from moves, practical fire from movies, practical fire elements that were shot, and basically going through them with him and letting him just tear them apart and saying, ‘That doesn’t look integrated to that, this fire looks safe to me, it’s not visceral enough, this fire looks calm, the people next to it look too safe,’ you know, that kind of discussion which early on set the tone for, ‘Okay this is gonna be a pretty high standard.’ And that was pretty understood given the very seriousness of the subject, it kind of demands a certain kind of respect.

vfxblog: What did that mean in terms of planning for the shoot?

Craig Hammack: So there was that early-on discussion, and then we showed up to New Orleans and we started discussing what was possible. It wasn’t a matter of early on, ‘Would this be better done in visual effects and this would be done in practical?’ It was a matter of, ‘What’s possible practical?’ And the reality is there are a lot of constraints around practical work on sets. My mindset was still, let’s do as much real as possible, and we’ll find the constraints there and then we’ll work based on what’s smart, what’s financially reasonable.

The Deepwater Horizon set included an almost full-sized build, along with greenscreens and a LED screen wall for interactive firelight.

Now, you say that and that’s great, but you have sets that are 60 feet up in the air – there are a certain amount of constraints around just how much propane you can pipe up 60 feet in the air, and about what’s safe to have on fire up there, and how much fire and water you need to put out fires if anything goes wrong. You can’t just flood a tank full of gasoline and light it up. There are constraints around proximity to actors and the safety of the crew.

All that gave us a built-in box to live in for the practical effects, and we pretty much pushed it right to the edge of that box. I didn’t put a lot of constraints on it myself as far as, ‘Well, we’re not gonna be able to extend that fire so just put a practical light there and we’ll do it all CG.’ That was never the mindset. It was always, get as much as we can because there’s a nature to fire just in the light that it throws which is unique and really hard to simulate. And also it just does so much to set the mood of the scenes when there’s real practical heat coming at you and torches burning propane tubes. They get flared up and it really helps the actors get in the performance.

vfxblog: Real fire on set is also something for the crew and the actors to play against, isn’t it?

Craig Hammack: Yes, I had a real desire to make the DP, Enrique Chediak, have to expose for the people and the fire, to have a real gauge for what the fire should like in the frame when we’re seeing people’s faces. I think that’s one of the bigger faults of some fire in film is that you know you don’t let the exposure go the way it does in reality. You try to capture too much detail, you maintain too much detail in your fire, or you basically expose the scene and the people up too much so they don’t live together in a natural world. So that was another huge desire of mine to have real fire out there as well. So knowing the subject matter and knowing that it’s such a big part of the story just how much a catastrophe it was and how engulfed that rig became, it was an obvious choice to do as much practical as we could.

vfxblog: And in terms of practical effects, they really did build a large part of the set, if not most of the set, in which to film a lot of action. Can you talk about that and just how that was built and staged for filming?

Craig Hammack: There were three main sets. There’s the overall footprint of the rig, the broad 250 by 300 foot platform that’s out in the middle of the ocean that is this rig. One of the huge that we were undertaking was building a section of that to 85 per cent scale in the middle of this parking lot in New Orleans and doing it basically entirely out of steel. So it’s 30,000 tons of steel that they brought in to build this set that is 85 percent scale. So it’s 60 feet in the air instead of 70, that kind of scale. What that allowed us to do was, we also had a water tank, a large water tank, at the base of it that held like two million gallons of water. And that allowed us to keep the section of the movie that happens on the lifeboat deck as practical as possible. It meant we could shoot down into the water, we could follow the lifeboats down as they got lowered practically, we can do a little bit of walk and talk along that deck to establish the rig at the beginning. We can also shoot up from the water and have basically the front face of the rig. That’s what that set was mostly used for.

Then there’s a strip that was maybe 15 feet of the deck above that which was dressed, and the bridge which sticks out the front of the rig. And that’s where the helicopters land. That was constructed to land a real helicopter on, which was the most impressive part to me. It wasn’t a small helicopter, either. It was one of the largest helicopters I’ve ever seen. It’s designed for transport of 30 people out to the rig, and they flew it in and landed it on the set. So that whole area, as well as the helicopter pad on top of the bridge, was dressed for in-camera work.

vfxblog: Now even though it was built at such a large scale, I’m guessing there was still a lot of extension and extra CG work required?

Craig Hammack: Well,there’s issues everywhere you look. It doesn’t take long to get off the edge of the tank and even the tank water was ‘tank water’ so it’s not quite the right nature of ocean water that you need. So much of that gets replaced anyway, but the ability to have water at the base so we could actually put water, or boats in the water, and people in the water, was great. They ran propane tubes under the water to get the fire on the water – they would basically just bubble up propane and ignite it as it reached the water. The ability to have all that together at a decent scale – it wasn’t full scale – but it was enough of the elements that it allowed for a certain amount of in-camera work.

vfxblog: What other set pieces were constructed?

Craig Hammack: We also had what we called the main deck set which was built I eight feet off the ground. If you’re looking top-down on the rig it’s the forward deck. So if the rig is split in half where the derrick is, it’s everything forward of that derrick. And that was mainly built for a few scenes where they first arrived to the rig and they’re walking around talking, we follow the actors in the daytime a little bit, and then it’s used as journey scenes during the destruction as they’re trying to get to the bridge, then get to the generator room, and back from the generator room.

It was an impressive set that was dressed with real objects from salvage yards. The good thing about being in New Orleans is there’s salvage yards that just have real working rig equipment that they were able to, in some cases, get the actual set dressing from. So that was a pretty good sized set that you could do some walk and talk, even though everywhere you look you’re shooting into greenscreens. We had a 270 degree 40 foot high container wall that had a curtain wrap on it that could be green or black or white. So we had quite a bit of flexibility there.

The last set is the drill deck set, where the derrick is, the tall derrick is right over that. And that’s a fairly self-contained set that was again built with working machinery and steel plated, like it would be. And they built the little housings that all the operators sit in. That was built basically to do the practical effects, the mud blow out, and to run all the stunts that happen on there which is quite extensive.

vfxblog: There’s a great series of wide shot approach shots of the helicopter coming in to land – can you talk about those and the model of the rig that had to be built?

Craig Hammack: It was a very, very heavy model. Very data intensive and detail intensive. One of the early discussions with Pete was about building this huge asset, and there was a little bit of a lack of definition of where the scenes needed to take place on the rig. So it was determined early on that we would build this thing to stand up to extreme scrutiny and we’d have to build almost all the rig to that level. At ILM when we have these kind of assets, we have these things called nested assets where you can build them in sections and basically assemble them very easily when it comes time to run a shot so that you only need to load in what you need for your shot. And that just became a series of very detailed model shots carved off to be manageable chunks at any one time. It just takes a lot of manpower and a lot of time to put it together.

Original plate.


vfxblog: What reference did you look to for building the rig?

Craig Hammack: Usually with oil rigs you can find documentation for, but that specific rig wasn’t all that well documented. We had probably just a handful, five or six, photographs to go on that the production art department had found that were high quality and were helicopter approach level shots. We had hoped to make it out to the sister rigs. There were only two rigs that we could find that were that configuration, and its sister rig the Nautilus was still in operation. So we had hopes of either going out and landing and doing a bit of documentation or at least being able to shoot approach plates or aerial plates around it to use and just swap out details of it to be the Deepwater. But it turned out the oil industry wasn’t really favorable to the project which was completely reasonable, and so permissions never came to do any of that. And by the time we wanted to do just fly-bys that rig had been brought in to port and so we couldn’t even access it that way.

vfxblog: During shooting, I noticed there had been a large LED screen used on set which I’m guessing was for interactive light. Can you discuss how it was used?

Craig Hammack: Again in those early discussions we talked about how we were going to simulate the firelight on the sets. Enrique was in favor of bringing in these LED walls, which is great because I’d used them in the past on Tomorrowland. So they got as much LED wall as they could in Louisiana and I think they ended up going to a rock concert promoter to get some LED walls because all the rest of them were being used on some other projects. So they got in basically a 30 by 30 foot wall of it. The idea was, we would pipe real fire through it, so we shot some fire elements and we took them back to ILM and manipulated them and scaled them and re-timed them to be a reasonable proxy of a large-scale fire, and we just piped them through the video playback on the LED wall.

It worked well for getting broad firelight that’s animated with the right frequencies. One of the problems, though, was that the idea was to fly it on a crane over the 60 foot tall rig in place of where the derrick would be, and we quickly found that the weather in New Orleans was not all that predictable and not all that calm. So we couldn’t fly it, which meant we had to basically anchor it on set which then puts it in a lot of shots and also had it casting light from the wrong spot. You never quite reach the brightness of real fire, but you do get the idea and you get the motion that feels correct for the shadows and the light play. It still ended up being a bit of a nightmare because you couldn’t keep it out of the shots. They were also shooting very handheld, just following along chaotically to build the drama, which means there was no specific planning for where the camera was going to shoot. And so there was a lot of clean-up that was involved in using it, but it was still a lifesaver for broad surfaces with firelight on it.

vfxblog: It seemed like this was also part of the learning process about what fire really looked like on film?

Craig Hammack: That’s right, and, you know, you get a certain amount of practical fire on the sets, which is great, but again it’s the clean fire. It has a very kind of ‘propane yellow under control’ nature to it which is the opposite of what we need. We need the ‘orange filled with black smoke, streaky really just nasty’ fire. So every time you end up seeing the propane fire you either have to dress over it or replace it completely or blend in in a really delicate way.

vfxblog: When they did shoot fire on the set presumably that also included smoke. But how clean did you want your plates? Then again, did that real fire give you the interactive lighting you needed? What were the trade-offs?

Craig Hammack: It was a debate in my head for the whole of pre-production. But, like I said, I’ve always been in favor of as much reality as possible, even if it makes our work harder – at the very least it gives you great reference. Even if we’re replacing the whole frame it’s a lot easier if you have something you can reference that was shot on the day in the lighting conditions with your actors. As far as real fire, again it’s an acting thing – where we needed the performances that lent itself to chaos and danger. And playing it safe for visual effects reasons, it didn’t seem like the right choice to make.

vfxblog: Let’s talk about how you actually did CG fire simulations. What was your approach to this?

Craig Hammack: Well, I think the fire that is most memorable in the movie is during all the scenes near the end – all that fire is pretty much all digital. We do of course have a quite extensive fire library, but again it’s not always the right nature of the fire and this needed a consistency of chaos and a consistency of nature of the fire that doesn’t lend itself to shooting elements. First of all, these days it’s not even legal to light those kind of fires unless you go to other countries, but also it’s such a storytelling part of the movie and the fact that the camerawork is all so frenetic, they don’t really lend themselves to 2D element work.

Also we are quite proud of our 3D fire software here and our tools that we have and have developed over the years. It just made the most sense to put our energy into refining that and honing our skills with that. And also it was really benefitted by the fact that this is a contained location. So, you take your time and map out where on the rig all the fires need to be, and then you spend just a ton of time developing fire for those locations. And then you can pretty much use it throughout the movie.

Original plate.
Final shot.

vfxblog: Is it Plume that was used for the fire simulation?

Craig Hammack: Yeah, it’s still Plume. We’ve advanced the development of Plume and of course we’ve advanced the ways that we source things into Plume. We can now basically generate sources in Houdini which then feed into Plume and stuff like that, but it is Plume and at its base it’s still the core software that we wrote, oh boy, a few years ago. Quite a few years ago. And it’s just amazing software to interact with. It’s the iterations out of it that are the real benefit, and its stability. It’s incredibly fast on our GPUs, and it just becomes this real artistic endeavor. It’s stable enough and it’s fast enough that you don’t – I mean, I’m sure some effects artists would shun this idea but since I was one and used it quite a bit – you don’t really have to know the nuts and bolts of the solver because it’s so fast. You can just try things, it gets to be, like I say, an artistic experience and turnaround.

You can run out 12 sims a day and see them churning along the way as the sim goes because it sims and renders at the same time. So it’s constantly spitting out final quality frames at you. And you can judge it along the way, and kill it and restart it and it just becomes a little more artistic. And so you can get to those flavors of the fire that you need, which for us was something that we haven’t really had to tackle before – this continuous fire that is continually fed by multiple sources and needs to be here in frame and feel like it’s continually bellowing out heat and methane and oil and rubber material. It’s just a good piece of software to explore those different looks in.

vfxblog: How did you study fire to then accurately replicate it?

Craig Hammack: One of the beautiful things about fire is it’s self-illuminated. It’s not light-rig dependent or shot-angle dependent. It’s also such an organic element in the frame, especially the way this needed to be, and the fact that it was nighttime and all the lights were out on the rig, it’s very forgiving for integrating it into surfaces. As long as you can cast light from it then the actual contact points and stuff like that are very forgiving because again it’s so organic. But one of the other main benefits of it, of doing it through our software, was we were able to develop a way to, for every simulation we did, spit out basically a proxy that could be used to light with. So that goes back to the decision of, do we try to wrestle real lighting into the set or you know do we just replace it? It’s a fortunate fact that we were pushed into developing a very strong asset for the rig itself and replacing most of the fire, because it meant that those decisions become a little easier when you have an asset already built that pretty much stands up to any angle and you have a fire system that very easily gives you all your lighting sources.

vfxblog: Clearly you were making a film where the heavy action sequences involved a lot visual effects but I assume they needed something to cut with and they need to see work in progress. How do you manage the process of simulations or tests or postvis to give a ‘feel’ for the fire before you’re finaling shots?

Craig Hammack: Well, that’s incredibly tricky.

vfxblog: Or do you just final them?

Craig Hammack: You kind of just final them. You know, we had postvis, and we actually had postvis artists sitting with production and editorial from ILM. So we had a very close contact there. And we gave a postvis artist a certain amount of practical elements that would simulate the size and the nature of the fire that we would eventually be generating. Those guys are just incredibly good at dressing out scenes and blocking out stuff with just 2D cards. They give you the right flavor to it. As far as showing work in progress it’s just almost impossible. It’s just not going to look good until you spend enough time to get the motion correct and even after that it’s just really hard to judge until you spend the time to integrate it into the plates. There were also not a lot of back and forth iterations that I would have with Pete over it. It ended up, fortunately or unfortunately – it’s hard to say – but it ended up being kind of an exercise in trust on Pete’s part, that once we got the right feeling of the shots as far as the chaos of the framing and the blocking of which events would happen, that at the end of the day we would have made it look real.

vfxblog: There’s something else to the actual fire of course and that is the haze and embers and light wrap that comes with it. Can you tell me about some of the effects and compositing challenges there?

Craig Hammack: Everything had to have a certain amount of heat distortion and most shots would need this atmosphere of embers and soot. It needed to feel like just a real nasty place, like you would just never want to spend any time there. And your lungs would be hurting, and you might be shredded apart from spending any time there. Pete was constantly saying this just needs to be nasty, it needs to be visceral, it needs to just feel incredibly dangerous in every aspect of it.

While we were shooting we did a few things to help ourselves there. We would do heat distortion tests with various propane burners to try and develop a kind of language to the heat distortion that we liked that Enrique and Pete were involved with just to give us reference. It wasn’t practical to shoot with, but just to give us reference and we went to the trouble of mapping it against a grid so we would know frequencies and scales and that kind of thing. And for as much as we all wanted a certain amount of that darker soot, smoke and embers in the air, it wasn’t going to be safe for the actors to continually blow that stuff around. So a decision was made to do all that stuff in post, and so for pretty much every time you see a scene that has just a blanket atmosphere of our recipe which was heat distortion, smoke and soot and embers, those are all added after. And we were able to work with Iloura VFX there in Australia to develop a whole toolset to be able to add those into those shots. So it became a very quick thing to do and it was a great interaction and workflow we had with them.

The trick is to actually drive one from the other, so even if the smoke is not visible, you’re basically driving heat distortion through smoke and you do the trick of basically blurring through the heat distortion in areas to give you the optical smudging you get from real fire optics. And then mixing in some amount of smoke into certain currents of the heat distortion, which gives you the idea that there’s parts of sooty smoke that get trapped in the same currents as the heat.

It all plays together pretty convincingly to give you this extra kind of flavor to the shots. It was very important to always feel like they’re in this oven – it’s a furnace out there. I remember having a conversation early on with Mike Williams, who is the real life guy who Mark Wahlberg plays in the movie, and he was one of our on-set consultants. He was talking about being up on the helideck and not being able to turn and face the fire because it was just too hot. It would immediately start just burning his face. It had to feel like it was that kind of intensity, and just putting our layers of CG fire and smoke would only take you so far. You have to feel it in proximity to the actors and camera.

vfxblog: One of the more dramatic shots is of Mark Wahlberg doing that final jump from the helideck with fire and embers encroaching on him. How did you pull that off?

Craig Hammack: There’s always one but that was the shot that was started earliest and last to finish. It was imagined as this continual experience as he jumps off and the camera goes with him, and you splash down with him and you go under the water and things are raining down on him under the water, and then you come up above the water, and then he has to go back under to go find Andrea. It was always going to be a nightmare, and everybody knew it, but it was the climactic shot of the movie. They were the last people off the rig and it’s their story.

So we had the helideck there which was already 70 feet in the air. We had a water tank under it. We thought, let’s try to use them and try to get that travel and that jump for real with descender lines and stunt guys. And those guys worked a long time to try to get that right, and in the end it came off beautifully but it’s pretty tricky to try to organize and orchestrate a stuntman going off, and a cameraman going off seemingly right on top of him.

Then we did many, many elements of shooting off explosions as the camera goes down, embers as the camera goes down, sending the camera all the way down to inches from the water to get that travel, letting the stuntman actually splash into the water in a few takes to be able to get some interaction, and then going to a stunt tank, for the actual splashdown and recovery and finding him under the water. All the stuff that rains down with him.

Which means all this becomes a safety thing as well because he has to be pretty severely harnessed up. You can’t really have any fire anywhere close to him or on the water or under him. Once he’s in the stunt tank you can’t rain down things into the stunt tank around him. So all that becomes very manipulated on our part, and again it’s all pretty much handheld cameras. As much as you try to line things up so that you can stitch plates together, everything’s in motion and in action. Nothing’s static, there’s no opportunity to settle the camera so you can do easy blends, and it just becomes a lot of hand work and patience in compositing and layout. It becomes basically the whole show for a group of people. That’s all they do.

Bonus before and after – ILM’s digital oil-covered bird. Original plate.
Final shot.

All images and clips copyright 2016 Liongsate.