Fast & Furious (4), directed by Justin Lin, had the biggest box office opening of the year when it was released in early April (now eclipsed by Wolverine and Stark Trek). Double Negative’s effects team built three tunnels different tunnels for chases in the film. Ian Failes talks with them and also Rhythm & Hues’ Bob Mercier about R&H’s contribution and the opening exploding petrol tanker sequence.

Double Negative’s Jordan Kirk (3D Supervisor) and Jon Bowen (2D Supervisor) talk to Ian Failes.

09May/ff/O_tr_055fxg: Can you give me an overview of Double Negative’s work for Fast and Furious?

Kirk: There were two tunnel chase sequences in the film. We had to build three different types of tunnels that they race through. When they first break into the tunnel, they break into a dark, unlit miner’s-type of tunnel which is very tight and winding and full of splits and turns. Then they burst through a double-width tunnel – it’s still a miner’s tunnel but it has lights in it and has a long straight run.

And then once they’ve gone through that section, they move off into a third tunnel which has supposedly been set up by the drug smugglers – we referred to that as our “smuggler’s” tunnel – and that was full of hand-turned rock. They must have knocked it out themselves with rough tools. And finally the guys blast through that out to the other side of the border. Overall, this meant that the three types of tunnels had to have three different looks to them. Our challenge was to come up with a look that wasn’t repeated. We didn’t want it to look like we were going past the same thing over and over again.

Bowen: They shot the practical cars in a sports hall, I think.

Kirk: Yeah, it was some kind of warehouse.

Bowen: They set up the routes with shipping containers. When they shot the cars they were already in a very dark environment with, at most, one little fill light. This was done rather than shoot them on greenscreen and pull a key, which would have meant they’d have to have lit the environment far more brightly than you’d expect in the film. Instead they used a plain canvas cloth to cover the details of the shipping containers. What that meant was when we came to composite the shots, we already had the cars in the environment with all the realistic reflections appropriate to driving down a tunnel. We then roto’d out the cars from the shipping containers, brought in the walls so that it was a little bit narrower than they were able to do, and for a couple of shots we sped up the environment.

Kirk: When they laid the shipping containers out, obviously they only had a certain length to run until they ran out of space, but these tunnels had to look like they were thousands and thousands of feet long. This was the reason why we couldn’t make a single layout to copy their layout of the containers exactly. That would have been a bit contrived. They were shooting up and down and backwards and forwards the length of the containers supposedly to represent different lengths of the tunnel. So what we did was have these LED tracking markers on the containers to track the shape of the tunnel. But then it was up to us to make it different with extra splits and turns.

fxg: Before you got to modeling the tunnels, did you have some concept or previz you could look to?

Kirk: We had some concept artwork that defined the different tunnels. It let us know that there were these specific looks for the miner’s tunnels and then smuggler’s tunnels. After that there was a little bit of free reign to play with the look. One of the things we did early on was look at some caves in Kent near us, actually. Chislehurst Caves, they’re called. We went down there and shot a whole lot of reference photography.

Initially we were trying to derive some of our geometry from that. We have tools where we can use multiple pieces of photography to create geometry. Then, most importantly, we used the photographs as texture reference for the rock and then mixed it up to create lots of mattes and runs of pieces of tunnel to stitch together to make one giant tunnel. In the end there’s probably somewhere between two and three thousand feet of tunnel created to keep the shots convincingly different.

Bowen: Even though it was very dark in the final film, we had to make our tunnel stand up as if it had been brightly lit. The information is still there in the dark values. The big thing that gives CG away is when you have these big, flat empty, black spaces where the shadows are. So we paid a lot of attention to make sure our shadows were full of detail wherever we could.

Kirk: Yeah, in CG we had to make a whole lot of paraphernalia – bits and pieces like spades and shovels and buckets…

Bowen: Wheelbarrows!

Kirk: …and oil lamps. They all had to go in. It’s one of those sequences where the shots go by so fast and you barely get a lock-off. But when those things were absent you could really notice it. It just had a generic look to it, but then when you put those things in it helped sell the shot.

Bowen: Also, I think visually you’re expecting the cars to be CG because they’d been CG in some of the other movies. So when you put the whole thing together with the real cars and the headlights and our tunnels with dust and interaction with the environment, we were pretty happy with how it was looking.

09May/ff/O_et_825fxg: What techniques did you use to model the tunnels?

Kirk: We use Maya in-house for our modeling. So we had a bunch of talented artists sculpting in Maya for quite a long time to create that look, working from reference as much as possible. We also have a proprietary system called dnAssets, which is a tool that lets us store vast amounts of data ready for rendering. We can lay the models out in a scene in a crude, proxy fashion as simple geometry that later on in render time becomes full-res.

We used 3D renders of match moved cyberscans as mattes. There was always a slight difference here and there, depending on who had been hammering in the dents on the on-set cars on the day. We worked to make sure that our cars matched up to that 3D outline. So our cars were clean models of the cars which gave 2D a good outline to start with. Then it was a case of further enhancing the roto to make sure it matched the outline of the cars.

Bowen: The big reason for using these 3D mattes is that once all the effects were laid in, all the dust layers, that would define the shape of the car. So from a 2D point of view, it was useful to have a proper 3D shape that would have an accurate profile of the car as it turned. It’s very difficult to roto that, anyway, but doubly so when you can barely see it above the grain.

fxg: How difficult was it to incorporate dust into the shots?

Bowen: Very! Especially when you’re talking about focus. You have a big semi-transparent element that’s got a lot of depth in it. The plate has to have a defined plane of focus and a fall-off. So it was quite a challenge rendering the dust in different layers. Obviously, if you’ve got a semi-transparent object it’s impossible to blur it all properly.

Kirk: How many layers was 3D giving you, Jon, for effects?

Bowen: For effects, it was the headlights, the tail lights, keylights and dust – per car. And then there were 5 cars in the tunnels sometimes.

Kirk: So it was really quite a lot. The other thing was that the director sometimes wanted to change the look of things. It meant we had to be quite careful in the way that we worked with those elements. It’s all very well, you can go down the road quickly trying to get the shot finished and you tie all your pieces of dust together to make the edge of the car looks great, but then all of a sudden if you need to just change one of them, everything needs to be re-balanced. So you have to be clever with the way that you work.

Bowen: The final touch was adding camera shake to lots of shots. Myself and Frazer Churchill (Double Negative’s VFX supervisor) went through one of the earlier edits and did a bit of direction of the shake, how it should build up and how it should act throughout the sequence. Because we thought there’s no point having a fully intense camera shake for the entire sequence and having that set as the gold standard. The camera was in different mounting locations from shot to shot. I think probably the clearest one was on the approach. Just before they break through the concealed entrance to the tunnel system, they’re driving along – camera shake, camera shake – and there’s a dip as they go down. So we had to do a little floaty moment as the suspension became unweighted and they were just about to dive down the hill and through the door. We also sampled some real camera shake from real cars, which we snuck in there!

fxg: What kind of reflections did you have to add?

Kirk: All that stuff had to go back in. For all the greenscreen plates, the cars had been shot with the glass removed. So that had to be replaced. From a 3D perspective, we had to render passes of the 3D tunnel that were comp’ed back in. Things like dust on the windscreen went back in…

Bowen: With wiper marks as well.

Kirk: …yeah, with wipers. And also, in what we referred to as the exterior shots, where the camera was mounted outside of the car watching the action, a lot of times then the rear windscreens had been taken out so we had to replace them and have them reflecting the overhead lights.

Bowen: A lot of the 2D was about taking cues from what was in the footage. When the greenscreens were shot, they put some interactive lights to simulate the tunnels they were going through. When we matched the 3D environment to sync with those interactive lights, it really helped to tie them together. Whatever cues we could find to extract from the plates, we’d try and incorporate that into our 3D.

09May/ff/O_et_1065Kirk: The director wanted the sequence to look frenetic and ‘high-octane’. Some of the shots we got, where the guys had been driving real cars at fairly restricted speeds in between the shipping containers, needed to look like they were going significantly faster. That’s why we had to build so much tunnel and why we couldn’t mimic the layout exactly. That tunnel had to travel past the cars faster than they were going to make them appear they were going 70 to 80 miles an hour instead of 20. It got to the point in some shots where the action that they shot just wasn’t working or wasn’t quite what the director wanted. So that’s when we turned to our 3D re-projection techniques, and used digital car models. We’d built some of them up to hero level.

We’d take the source material from the real photography, re-project it onto the geometry and then after that we’d do some clever stuff with our animators and play with the cars. So some of the shots ended up being entirely digital with the environment and the replaced cars. One of the digital shots is where Fenix is coming over the jump in the early parts of the tunnel and makes the jump with a little ‘yelp!’.

Bowen: In the shipping containers shoot, they obviously couldn’t dig a little section for the cars to do that jump.

Kirk: Yeah, the cars were all shot driving along a flat floor. That was another thing we faced – how do you make cars going along a flat floor look like they’re driving on rocky, bumpy terrain and being thrown around looking really dangerous? Also, things like where Vin Diesel cleverly times the opening of his car door before jumping into the other guy’s car. The doors are open and get smashed on the pole and come flying to camera. Obviously they didn’t want to rig and do that practically because it was quite dangerous. So we built them digitally.

The final areas of fully digital work are when at the end they come up to the last section of tunnel. Brian and Fenix both burst out of the tunnel and have a mid-air collision. At that point, there were fully digital cars coming out into daylight. We’ve got a fairly large library of HDRI images at Dneg and we used one of those, built some geometry for the mountains and did some matte paintings and 2.5D projection work. There was also a lot of debris for that shot. We used a tool of ours called Dynamite and a bit of Houdini. We actually used Houdini to pre-break the tunnel. When the tunnel breaks and falls apart, we built it into its component parts. It doesn’t automatically have a back piece, usually a tunnel is just a shell. So when it breaks and falls apart, you don’t want it to look like pieces of paper falling down. So we added depth to the rock.

fxg: There was also some Maya cloth used too, wasn’t there?

Kirk: Yeah, chicken wire was strapped to the walls of the tunnel to hold up loose rocks. So for all that collapsing we had some cloth simulation done with Maya cloth. That end sequence was quite involved. Just the sheer amount of geometry that had to be set up to collapse was quite a big undertaking. Plus mattes for all the dust. So all the rocks had to hold out the dust correctly. Jon, how many passes were there for that shot? I think it was over 115.

Bowen: It was the equivalent of about five shots in one, really.

fxg: One of the things I like most about your work for the tunnel sequences is that it started out real with real cars being shot. Is that something you like about visual effects in that it has to be entrenched in reality or it just won’t work?

Kirk: Exactly. The good thing about it is you don’t get away with starting from scratch and making everything up. It has to be based in reality. At least as much as the action dictates if it has to be crazy and out there. You can’t just go for anything with CG. You’ve got to tie it to reality and make sure the elements you do have actually look like they all sit together.

Bowen: There’s a great amount of satisfaction in compositing to appear invisible. To achieve seamless effects. Almost the bigger spectacle in the shots, the more you’re trying to make it look as realistic as possible. I think we pulled off some shots where the question of real versus digital doesn’t even occur. That’s the gold standard, that’s what you aim for.

Kirk: You’ve got to make it look real, but still make it a spectacle.

Interview with Bob Mercier – Rhythm & Hues – Fast & Furious

Rhythm & Hues visual effects supervisor Bob Mercier discusses what it took to create digital environments and an exploding digital tanker for the opening of Fast & Furious.

fxg: Could you give me an overview of Rhythm & Hues’ work for the film, and what your role involved?

Bob: Rhythm completed the opening sequence of the film; the gasoline tanker sequence. Primarily, we did the second half of that where the tanker flips over. It’s burning and tumbling and Dominic (Vin Diesel) is trying to find a way to drive underneath. As the visual effects supervisor for Rhythm & Hues, I was in charge of our work.

09May/ff/ff42hBg

Original background frame


fxg: Did you get involved early on to work out the look of the shots?

Bob: Yeah, we were involved before principal photography with the guys at Universal, Thad Beier, visual effects supervisor, and the director, Justin Lin, to try to figure out how we were going to shoot the plates. And then we were involved all the way through.

09May/ff/ff42hLightMainTankerWireEnv

Wireframe


fxg: What sort of issues came up in your early meetings?

Bob: Well, an example was whether to have the gasoline tanker be chrome, as they sometimes are, like stainless steel and be highly reflective, or have them have a matte finish with flatter coloured paint. We looked at what was common in the Dominican Republic; which would be more interesting to look at; which is more realistic; and which would be more difficult in terms of carrying off the effect.

09May/ff/ff42hCmpMainMain

final frame from the comp


fxg: What sort of concept work was involved?

Bob: They had the look for tankers pretty much worked out, because the production actually built a working land-train. They built a rig where they could have five full-sized gasoline tankers driving around on the highway. Obviously they couldn’t flip over and there were great limits to how fast they could drive, but the look was worked out.

Our visualisation had to do with the environment. In the second half of that sequence the entire environment – the road-way surface, the ravine that they’re driving in, the sky – it’s all synthetic. So we had our art department do a bunch of design work back and forth with Justin to come up with a look that he liked, that we felt represented what he wanted in the movie.

fxg: How did you go about modelling that synthetic environment?

Bob: For small sections of it, we had LIDAR for the things that had to match principal photography. We use Maya primarily for modelling at Rhythm & Hues. We used Zbrush to do some concept work, but ultimately the final model was done by hand in Maya.

09May/ff/ff59bBgGs

Greenscreen plate


fxg: How were the shots finally executed?

Bob: The shots that had live-action tankers were photographed about an hour north of Los Angeles in an area called the Templin Highway. It’s an empty, relatively remote stretch of mountainous highway. We went there with laser survey equipment and surveyed about a kilometre of road as detailed as we could. From that we made the model so that it would fit reasonably well with the plate photography. For a given shot we would roto off the tanker and replace the entire environment.

We would then used that roadway survey as the basis for our model and then extend it to the sides. For example, there’s a shot early in the sequence where they drive over the crest of a hill on the roadway and the road drops off steeply away from them. That was all synthetic. Justin wanted it to feel very treacherous, suddenly. But at the very beginning of that shot, we had a real roadway we had to match and a real tanker. So that’s an example where we grafted our CG environment onto a very detailed survey of a real environment.

09May/ff/ff59bCmpTankerEnvOnly

Tanker and the environment only


fxg: In the crash part of the sequence, there’s a lot of fire and smoke. How did you accomplish those shots?

Bob: Well, we still don’t feel like it’s easy to do fully synthetic fire. So we had some practical shoots to generate a variety of practical fire elements for the shots. For the most part where you’re seeing flames, there’s between 10 and 20 compositing layers of pieces of fire that are stitched together to stick to the CG tanker as it’s rolling.

For the look that we wanted, you couldn’t easily photograph smoke and fire elements that were going to work in that environment. So the main challenge for that sequence was to try to marry our volumetric smoke elements to the practical fire elements that were composited onto our tanker. In particular, our smoke is all driven by a complex fluid simulation that’s partly our software and partly Houdini. It took a tremendous amount of compositing work and some crafty Houdini tricks to get our CG smoke to match the practical fire. It had to look like the smoke was coming from whatever direction and velocity the edges of the fire was travelling, The smoke you see when it’s born off the edge of the fire has to seem like it’s coming from the same thing.

09May/ff/ff59bCmpWireframes

Wireframe


fxg: What kind of reference did you look to for the tanker crash?

Bob: There is actually a large number of examples from professional work – Twister, there’s a great Bond movie with a burning tanker, and searching on YouTube! Just searching on YouTube turns up some interesting oil fires and gasoline fires. We really did try and make it look like something that would be, although not perfectly accurate, something that people would definitely recognize as an oil fire.

09May/ff/ff59bCmpMainMain

Final frame


fxg: Fast & Furious is quite frenetic and always moving. How did that impact your approach to the effects generally?

Bob: Well that‚ is Justin’s aesthetic. He’s known for almost a music video look. For us, it meant that no matter what we showed him, he wanted it faster. Faster, faster! And we pushed our fluid simulations and our rendering software to the limit. It’s not easy to convey high levels of motion blur in volumetric rendering. It’s expensive to try and get that look. With fluid simulation software, the faster you run it, the more time slices you need to make it look accurate, which greatly increases ren-time and memory requirements. So Justin pushed us to our limit. That aspect of the look did have quite a large impact on how difficult it was to do the work.

fxg: Finally, are you now scared of driving on the road next to gasoline tankers?

Bob: Well, not CG ones! Nah, it’s fine. It was great fun to work on. They were a great team. I couldn’t say enough good things about our compositors. Difficult as the 3D was, the 2D was even harder and our compositors were fantastic.


Thanks so much for reading our article.

We've been a free service since 1999 and now rely on the generous contributions of readers like you. If you'd like to help support our work, please join the hundreds of others and become an fxinsider member.

Leave a Reply