Visual Effects Supervisor Josh Simmonds runs down the many ways in which his team amped the verisimilitude in George Miller’s latest ‘Mad Max’ franchise outing.
From terrifying car chases to toxic dust storms and deadly sinkholes, the world of Furiosa: A Mad Max Saga is not an ideal place for a tranquil vacation. A prequel and spin-off to 2015’s Mad Max: Fury Road, and the fifth installment in the Mad Max franchise created by George Miller, who has directed all the films, the latest iteration of the iconic series presents the origin story of the Fury Road character Imperator Furiosa (originally played by Charlize Theron, and here incarnated by Anya Taylor-Joy and Alyla Browne).
As for the chases, dust storms, and sinkholes, they were created by the Melbourne branch of the redoubtable animation and VFX studio Framestore, led by Visual Effects Supervisor Josh Simmonds, who previously served as lighting technical director on Fury Road. We talked to Simmonds about his second tour of duty in Miller-land, and the tricky business of making dust, dunes, and gnarly vehicles.
Dan Sarto: When did you come on the film and what materials did you have to work with?
Josh Simmonds: We came on board in January, 2023. The film had been shot by that stage, mostly on a vacant lot south of Sydney, as well as some location work out in Broken Hill, which is kind of the Outback. Pretty much all of our sequences were shot either on that vacant lot or in a studio.
DS: I understand that the lion’s share of your work on the film involved a number of key chase scenes. Can you talk a little about those?
JS: Our first big sequence was when Furiosa is abducted as a child, and there's a chase scene across the desert with some of Dementus' bikers, and Furiosa's mom chasing them on horseback and taking them out with some very good sharpshooting. Sydney caught some incredibly inclement weather during that time, so they were shooting in-between thunderstorms, very dark skies, very wet grounds, none of which really lend themselves to a bright desert environment. So, in many cases, we had to take over the entire environment right up to where their feet or their vehicles are touching the ground, which effectively made it live-action components, both vehicles and actors, on an entirely CG environment.
Working with those plates was tricky for a number of reasons – not only because we had to do animated tire tracks, and footprints, and so on – but also getting the lighting right, because they did have to add quite a lot of onset lighting because it was so overcast. Often, when you're looking to match bright outdoor sunshine with artificial lighting, it's a really difficult challenge. So we were using all sorts of tricks, like making animated cloud shadows, CG shadows that would sort of drift. That was one of the more difficult aspects, just getting it all to sit in that environment.
Creating the actual dunes themselves was also a big part of the challenge – diving deep into how they're formed and making sure you're getting the angles, and the prevailing wind direction, and so on. All that kind of stuff actually makes the background feel legit, versus random sand dunes plonked everywhere. That made a big difference to the workflow.
DS: The second big chase scene is at the end, with an adult Furiosa. Can you walk me through that one?
JS: After the montage section of the big battle, Furiosa goes chasing after Dementus and his enforcers, and they head out across the desert. Very similar technical challenges in terms of completely replacing environments for the same reasons. In this situation, we had to do a little vehicle work as well. The actual vehicles that they built for the film were amazing, but in some cases, for safety reasons or otherwise, the vehicles were static and we had to simulate travel on them. We leveraged plates as much as possible, but often working with quite different takes of one vehicle and then a different take of the next vehicle, and trying to make a consistent camera across all of them.
One of the more challenging aspects of working on Furiosa was that the edit was just never locked – it's always being refined, and different live-action elements were being stripped in to help tell the story. I think [director George Miller] doesn't really feel any limitation about exactly what has been shot – he likes to use those elements to build the story and build it in a way that he thinks works best. And so it's our job to interpret both the postvis and sometimes the Avid comps into a continuous shot, and make sure that it works. So, we might have two vehicles in live-action, but shot on different lenses and different cameras, and we have to find ways to blend those together. Sometimes it involves switching one to CG partway through the shot and making sure it's as seamless as possible.
DS: My understanding is that in Fury Road a lot of the vehicle stuff was done at slow speeds, and then sped up for the final cinematic version. Was that done on this film as well?
JS: There was certainly components of that. We had sections during the chase scenes where we had motorbike riders doing 15-20 kilometers an hour, and we had to speed it all up. Sometimes that involved moving them further forward, especially if you're tracking with the camera – you're moving the camera faster as well just to make sure everything is sliding through the world at a faster speed. Given that we kind of owned the environment, it makes it a lot easier because we can just effectively make sure that the environment's moving faster, and that in turn makes the bikes look like they're traveling faster.
DS: In these instances, were all of the assets provided or did you have to create them?
JS: In the case of the Cranky Black, which is the hot rod that Furiosa is driving at the end, we built it completely from scratch. That was probably our biggest build. We had lots of great reference photography and some scans to work with, but you're effectively rebuilding the model and making sure that the textures are looking right, that the chrome responds in the same way that the live-action chrome does, and so on. And it's especially challenging if you're transitioning from a plate car to a CG car in the shot. I think we did a pretty good job of disguising those transitions. Also, for Dementus’ chariot, with the three bikes pulling it, we had a fairly low-res model. It was only when we were looking at these shots that we realized we had to get up pretty close to it. So there was an extensive build getting that up to spec as well.
DS: In addition to the two chase sequences – which sound like they gave you plenty to do by themselves – Framestore also created a toxic dust storm and an impressively terrifying sinkhole for the movie. Can you talk a little about those?
JS: We knew we had to do the dust storm fairly early on, and we started doing some development work on that right at the start. They told us that the dust storm is not nearly as much of a player in Furiosa as it was in Fury Road. In fact, they gave us percentages: "It's a 40% dust storm at the start of the film, and then maybe it works up to a 70%." We used a lot of the same techniques in Houdini as we did in Fury Road, simulating individual components – very large detailed volume sims combined with lots of fine particulate – and then layering it up to help create the composition of the shot.
The sinkhole is one of the last big effects in the film. Furiosa has been picking off Dementus' lieutenants one by one and, at the end, she's chasing a character named Smeg, who’s riding in Dementus' chariot as a decoy. The actor that plays him, David Collins, is also a comedian, and it was actually quite a funny sequence. The idea is that he's so busy trying to explain to Furiosa that he’s not Dementus that he doesn't see this giant sinkhole coming. So when he hits it with the chariot, he goes flying, flips through the sky, and lands on the other side of the sinkhole, before getting sucked into it.
On set, it was basically the actor being pulled down a gentle sandy slope, so we had to digitally build the entire sinkhole. We started with a sculpt based on references of sandy sinkholes, and then we simulated billions of particles on top of that. Again, it sort of started off with individual components that we could have breaking away. But the way that the sinkhole works is like it has a plug in the bottom of it, and when the chariot drops down, the plug gets released and the sand starts getting sucked from underneath up towards where the actor is. Making sure that we could get the detail that we needed, both in wide shots and also extreme close-ups, was one of the more challenging things. So we had a base sim, and then we just had to really fill it out with billions and billions of points to get it looking believable.
DS: Have those kinds of labor-intensive tasks gotten easier with the advent of AI and/or other advances in technology?
JS: Our very early temps used some AI roto just to help do junk maps, because we were cutting the characters out from the background. We needed to do temps for various screenings along the way, but there's no way that any of that would ever be close to production quality. I think we ended up with about 6,000 days’ worth of roto that just went into cutting the actors and the vehicles out from the background, because you're dealing with flyaway hair and so on. And it's not just about roto, it's about the compers using every trick in the book to try and retain as much of that hair as possible when you've got a very busy background behind them and we have to replace it with nice smooth sand.
But as far as environment creation, one of the cooler pieces of tech we used was a dune solver, just to get the sands flowing the right way, and to get the overall shapes of the dunes. But then it's just heaps of custom textures and lookdev work to get the details looking right. There are a lot of great resources now, like Megascans, that will give you normal maps and full-detailed tileable textures that you can use for those sorts of things. But our lookdev artists really dove pretty deep into that and made some incredibly complex shaders that, while they were challenging from a rendering perspective, really did sell the level of detail that we needed.
One other component that was a fun piece of tech was using VDBs, or volumes basically, to create animated tire tracks, that kind of thing. And Framestore has its own renderer, Freak, that works with a whole bunch of useful little utilities. One of them allows us to feed a volume in to create a displacement that you can use again in lookdev to combine divots in the ground with a crumbly sort of surface, so that as bikes are animating around through the scene, they're effectively drawing down the tire track live.
DS: Looking back on the work you and your team did, what was most rewarding, or most challenging in a good way?
JS: For me, it was seeing what can be done in terms of blending different takes and performances. For George, the story – and making sure that an audience can follow action – is everything. And so there were situations where we'd be working in individual shots, and then suddenly we'd be looking at completely rearranging some of them to make sure that, from cut to cut, things will match and the audience will know where to look. And the same holds true for color, in that it's very much designed to direct your eye in an almost graphical way. So you introduce contrast where you want the eye to go, and sort of flatten things out elsewhere. So that was one of the more interesting discoveries – how George works to make what could potentially be quite a confusing action scene very clear to the viewer.
Dan Sarto is Publisher and Editor-in-Chief of Animation World Network.
