Dealing with dogs, wolves, a river rapids crossing, a plane crash, and titanium teeth.
Directed by David Ayer, Heart of the Beast follows retired Special Forces veteran James Belmont (Brad Pitt) and his retired combat dog, Odin, on a harrowing 58-mile trek across the unforgiving Alaskan wilderness after their light aircraft crashes into a lake.
A primary German Shepherd named Uber—a former search-and-rescue dog—was cast to handle the expressive hero shots, while three of his sons served as specialized stunt doubles tasked with specific physical demands like wire work, swimming and running. Production shot in New Zealand, standing in for Alaska, with on location and studio and backlot filming all part of the mix.
While the practical dogs provided the heart of the film, digital creature work—entirely handled by Framestore—was also crucial in realizing a number of shots and sequences with Odin. VFX also came into play for the plane crash, the river rapids crossing and for encounters with several other creatures such as a bear and an aggressive pack of gray wolves.
Here, befores & afters goes deep into the film’s effects with visual effects supervisor François Dumoulin, who hails from Framestore. He explores the art of delivering complex VFX for a film that was not meant to feel like a VFX film at all, starting with the ‘Pepsi challenge’ showcasing the real vs CG dog, orchestrating a highly realistic plane crash, coming up with a unique way to shoot the river crossing, and dealing with puppets and performers for key animal scenes.
b&a: How did you come on board this film?
François Dumoulin: I got approached to work as the overall VFX supervisor on Heart of the Beast around December of 2024, just when I was wrapping Prehistoric Planet with Framestore. I was coming from this massive documentary, which took two years and a half of my working life at Framestore, doing basically a masterclass in creature work with paleontology, and really pushing our creature pipeline at Framestore, to a level of excellence that I think we never reached before.
All of that made us a go-to choice for Paramount because they knew going in that there were a lot of animals in Heart of the Beast, and they were looking for a company that would be able to execute those animals to a really A-level. There were also a lot of questions about how to handle the interaction between an A-list actor and CG characters. Framestore has demonstrated a really strong history of solving these kinds of problems on different types of show, whether it’s fantasy creatures or realistic animals.
I interviewed with the director in late December 2024 after having read the script for the first time, and I got really emotional reading the script. For some reason, that story in all of its simplicity—it’s really bare bones, there’s no fat in that script—but it really touched something inside of me, and I really wanted to do it. And when I got into the interview with David, the director, I instantly knew that we had the same vision about that story. The fact that on the surface it’s a survival adventure, but deeply it’s really a story about that bond, the unconditional love between the man and his dog. That opens up the movie to a very different level, which, in my opinion, makes it more like a spiritual journey rather than a traditional survival adventure.
b&a: What were some of the early conversations you had about how they would realize Odin? I mean, clearly, it’s a real dog, but I imagine you would have been part of those early conversations about, well, how much can we film with a real dog, and how much might be needed to be done with any augmentation or CG?
François Dumoulin: When I joined the project, there was a lot of anxiety around everything relating to the dog because one thing that you need to understand is we had an extremely short preparation time. We ended up having less than two months of prep.
There is this great animal wrangler, he’s from Canada, his name is Andrew Simpson, who flew to New Zealand. The second the production said, ‘Okay, we are going to shoot in New Zealand’, well, there’s a lot of restrictions that come with shooting in New Zealand. There’s a lot of rules. You can’t import animals, or if you import them, there’s a quarantine period that would have completely screwed our shooting schedule. So the first thing was, they would be sending this animal wrangler into a country that he doesn’t know, and he needed to find the dog and he needed to train him, but we only had two months.
So, when I joined, there was a lot of pressure around, well, if the dog doesn’t behave, can we do anything with visual effects? What’s the range of things we can do, and can we afford to do them? Also, the dog is a three-legged creature. One of his legs is a prosthetic. So there was an initial discussion about whether the prosthetic should be CG all the time, which we quickly figured out would be cost-prohibitive. So we turned to Wētā Workshop in New Zealand to build a custom prosthetic for our hero dog that would sit on top of his real leg. We still ended up doing a CG leg in a dozen shots, because the prosthetic was sometimes getting in the way of the dog’s performance.
We were then extremely lucky with the casting of that dog. I think the success of the film is in really good measure thanks to the dog we found, which was incredibly cooperative and a very good actor. It was actually not just one dog, it was four dogs. The main one, called Uber—you can see him everywhere in social media these days, he’s getting very popular—is the dad. And we also had his three sons: Seeka, Ryker, and Hondo. Given the very short time we had to prep and train those dogs and getting them used to the location, what we decided to do was train each of them with specific scenarios in mind.
Uber was particularly trained for his chemistry with Brad, so he got to spend more time with Brad during prep and develop that bond. For the other dogs, one of them was extremely good at running fast, one of them was good at fetching, one of them was good at swimming, etc.
But then we knew going in that preparation time was lacking, and that we were shooting on location, so there could be all kinds of things that could happen on the day. So, I had to be prepared to come up with solutions on the fly as we were shooting scenes if something would just not work on the day. Which means that obviously from the get-go, I knew there would be a need for a digital version of the dog. Very early on, we knew we would need a replica of the dog that would be a completely seamless match down to every single fur detail, and we would require that dog for many different scenarios. Just reading the script, there were a few scenes where the level of danger or violence was so extreme that it was obvious we would not be able to get that with a real animal.
Then as the shoot progressed, there were moments where we were not exactly getting the right performance or the timing was off, so we also started to talk about hybrid solutions where it’s neither a full CG dog or a practical dog, but a hybrid where maybe we keep the body language of the real dog, but we replace the expression of the face, or simpler things where maybe it’s a re-time or a split-comp where we take Brad from one take and the dog for another take. So, the scope of work on the dog is really broad.
b&a: What sort of things did you do to bring those dogs into the digital realm, say in terms of photogrammetry, reference gathering, and watching their gait and movement?
François Dumoulin: It’s pretty standard in terms of how we acquire data for a character, whether it’s an animal or a human being. We put the dogs into a scanning booth and calibrated that booth for the dimension of the character and captured polarized, cross-polarized and different lighting scenarios. We tried to acquire the best possible textures. Obviously, when you’re scanning fur, it’s quite challenging because you get a lot of noise. So we also did scan the dog with matted fur and unmatted fur, with and without the prosthetic scars, bandana, jacket etc We scanned all four of our dogs, but knowing that one of them, Uber, would be the hero asset, so that’s the one we put more effort into the scanning and acquisition of data.
And then a lot of it comes down to performance. Framestore has an amazing creature pipeline. I mean, they’re in my opinion best in class when it comes to creating a character that is anatomically correct from the inside out. So they will make sure that every bone fits exactly in the right spot before attaching the muscles and the fascia and the skin, and all of those things that are simulated on top of each other.
I knew that technically it was a no-brainer to achieve a perfectly good-looking and good match to our dog. What I was most worried about was getting the performance correct because our dog is so unique, he’s got such a strong personality, there’s so much going on in his eyes and in his expression, and that was really the hard part to nail. So that comes down to me just spending the entire shoot observing that dog and getting to know the personality of that character and then handing that to a brilliant team of animators at Framestore, under the guidance of animation director Meena Ibrahim.
Towards the end of the shoot, I also organized a shoot specifically for VFX reference where we got our hero dog Uber on a platform in front of a bluescreen with multiple cameras so we could get different angles. We got his favorite trainer so he could feel completely comfortable, and we had him run through the widest range of motion and emotions that we could get within a single day without exhausting the dog. That produced hours of footage that I gave to Framestore.
That was really the foundation of the work and our way into establishing trust with the director and all of the studio people that we were going to be able to execute not just great action scenes with the dog, but also great emotions and be faithful to the emotion of the real dog.
And what Framestore did as a test very early on before we started turning over any sequences was, they grabbed a 20-second fragment from that big shooting session of reference that I did on a sound stage with the dog, and they did a very detailed, extensive match-move of that dog. They ran that through their simulation system and they did a first pass animation of the facial features. I showed that to both the director and the studio executives on a big screen side-by-side: the real dog on the left, the CG dog on the right. The game, of course, was: which one is real?
Our director, David Ayer, was very, very, very reluctant to do a CG dog. He really wanted to capture 100% in-camera. And for him, CG would only be good enough for quick shots where we’re panning with the dog, or he’s biting a wolf, but it’s very dark. He was completely opposed to the idea of doing any emoting and close-up shot. But the fact that our asset was so convincing and that we got that trust very early on with brilliant work from Framestore, it suddenly opened the door to doing much more than those action shots.
I won’t reveal exactly which shots, but there are some really key moments of performance and emotion that are entirely digital. And that’s really a testament to the quality of the work, because I think they’re completely indistinguishable from the real thing. I’ve been looking at the comments on the internet, which I should never do, but so far I haven’t seen any people pointing to the fact that there are close-ups of a digital dog in that movie.
b&a: Well, I certainly couldn’t tell. I mean, as you say, there’s a lot of action scenes where I feel you wouldn’t put a real dog into that scenario. I’m going to ask you about some of those in a second, but just one quick thing: we do get a close-up view of his titanium teeth at some point. Was that a prosthetic or is that a digital mouth replacement?
François Dumoulin: It’s a digital mouth. And actually, it’s a little more involved than that. This shot has a funny story because the titanium teeth are part of the character, and we wanted to show them to tell the story that he’s been through war, and just like James, he’s damaged. And so the reveal of the titanium teeth was always an important thing to get to tell that story. But when we shot with the real dog, our dog was just too nice! He was just a happy fellow enjoying time in the outdoors with a big group of friends, and we never got him to look angry or snarling. It’s just something he would never do. Whatever we would try, the trainer was desperate, but the dog was always happy.
We tried this thing for the shot of the reveal of the teeth but it just looked completely stupid and rather hilarious. So what we got out of that shot was a good body position, a good composition, the timing of the way he’s turning his head is correct, but the expression was entirely wrong.
Then, at first we explored doing a completely CG version of it, but for some reason, there was something about the performance of the body that we wanted to preserve, and that just felt better than trying to reanimate it from scratch. So I asked Framestore to do an extremely tight body match-move of the real dog and line up all of their facial rig and skin deformation and everything so that we could keep the eyes and the ears and all of the rest of the body from the real dog, but everything from the jaw up to the nose and the entire muzzle is replaced by a digital asset. So it’s a Frankenstein’d dog, which is very interesting, and it’s something we’ve done in a couple of shots.The good thing is those titanium teeth are only revealed in a handful of shots, so the volume didn’t make it cost-prohibitive to approach it as a VFX solution.
b&a: In the film, we’re in Alaska, and it’s just beautiful landscapes, but as I understand it, you shot mostly in New Zealand. How were you dealing with any kind of landscape augmentation, or could it be basically New Zealand for Alaska?
François Dumoulin: We chose New Zealand because of its beauty, and we never said, ‘Okay, let’s shoot in New Zealand, but the landscapes are going to suck and we’re going to have to replace everything.’ That was never the philosophy. We really embraced the landscape of New Zealand for what they are in all their stunning beauty, and I think those landscapes are an important character in our movie. There were initial conversations of, maybe we’re going to need to add distant snowy mountains, and that was not just to make it look more like Alaska, but more to give our story some visual anchor points.
Now, when we talk about environments, when I joined the show and I had this first interview with David, and I was just coming out of Prehistoric Planet, so my mindset was very much documentary: let’s make everything look completely real, like the real world, and let’s base everything on acquiring data and reference and shooting plates, etc. And David told me something that really stuck with me for the duration of the show. I don’t remember if it was about the animals or the environment, but he said, ‘François, I’m not interested in doing a documentary at all. I’m doing a movie here, and I want to do really a classical movie, a piece that is timeless, and it’s not exactly about Alaska, and it’s not about like some generic wolves in Alaska. The world of the movie is unique to our movie, and it’s somehow a fantasy world in a way.’ And that’s something that I really got when I was reading the script. Actually, my first impression when I read the script was maybe James died during that plane crash at the beginning of the movie, and the entire journey is him carrying his dog in that land between life and death, and one of them is going to survive.
I spoke to David about that, and that kind of almost mythological dimension of the journey, and we were completely on the same page. Maybe my vision was a bit more French and romantic than his, but I think there’s definitely an aspect to our story where it’s just not our world, it’s more of a spiritual realm that we’re getting into. So when it comes to environment work and creature work as well, we were really walking an interesting line. This was very interesting for me as a visual effects person, where you’re starting from reality, but you’re twisting it just ever so slightly so it’s no longer exactly reality, it’s something else. Every single animal is based on a real animal, but it’s slightly bigger, it’s got these weird bigger-than-life qualities, there’s something strange about the fur markings or the scars. I’m not going to say they’re anthropomorphized, but they all have a quality that brings them to be almost like spirits of the forest, spirits of nature.
The same goes with the environment. We started with those pristine locations from New Zealand, and as we progress into our story, the world becomes weirder and weirder. And we got to those places like the tree bridge crossing which is, yes, based on location, but it’s augmented in a way where it’s a little bit unsettling. There’s something not quite exactly right about that place. And we end up in this huge swamp with dead trees poking out of gray water, and clouds shifting in the sky with rays of sun, and it’s no longer New Zealand or Alaska, we are somewhere else entirely.
b&a: I love that thinking about that now that I’ve seen the film, especially towards the end, and I won’t try and spoil it, but especially towards the end there’s something a little bit more ethereal about those environments, and I think you did a great job of mixing reality with something else in there.
François Dumoulin: Yeah, again, we can’t really discuss it without spoiling the movie, but towards the end, there’s this thing we called ‘The Eye of God,’ which is a really psychedelic moment, it’s like a mind trip. And I think we pushed things quite far in terms of getting the audience completely elsewhere. ’ And I’m very surprised nobody gave us any notes about this thing, and it pretty much stayed per the initial idea and into the final product, so I’m very happy about that.
b&a: How did you orchestrate the plane crash? We see the plane flying over the landscape quite a bit. We see the characters inside the plane. How were you doing those shots, and then how did you actually orchestrate that crash into the lake?
François Dumoulin: Well, once again, I was dealing with a director who was really keen on getting everything in-camera. I think if he could have crashed a real Cessna, he would have done it! Obviously, we had to find ways of making this crash safely, but David did push to get a real Cessna down in New Zealand and fly it over the mountain and get Brad inside of the cockpit. It’s a very small plane inside of that cockpit, you can only fit three or four people, I think. In the front seat, Brad was sitting next to the pilot. The pilot was dressed the same way as Brad so that we could do over-the-shoulders of the pilot and pretend that it was Brad. And David was operating the camera behind them. That happened within the first week of principal photography. That gave us a fantastic bible of what those shots would look like. Every time you see the Cessna flying over those majestic mountains, it’s all real, it’s a real plane and Brad is actually in that plane, and the director is sitting in the back and capturing footage from the inside. So, there’s a lot of photography that was acquired for real on location.
Obviously, as soon as the heart attack happens and the plane starts to dip down and starts colliding with the trees, this is something we’ve done on a motion base. We brought our Cessna back in Auckland on a sound stage, and we mounted it on a gimbal that allowed us to move and shake the plane in a pretty cool and realistic way. And that was shot in front of a bluescreen with old-school tricks like moving a light source on a crane to pretend that the world is moving instead of just moving a light source.
Based on the flight path that David had in mind, I also put together a little previs with the team from Day For Nite in LA. Those guys were really instrumental in just jamming with us and putting together quick iterations of ideas. I had them extracting geographical data from Google Earth for the flight path and getting a model of the Cessna plane and put together a little techvis for us of how different camera angles would work and match with the photography that David acquired. Because, what we didn’t want to do was have those extraordinary shots inside of the real plane and cut to CG shots that would look completely CG just because the cinematography would be different, in terms of choice of lenses and position of the camera. So we really wanted to respect the way David was framing the shots from inside the plane and the way the shots of the plane were acquired by a helicopter flying along it. That drove a lot of artistic decisions like, where should we put our camera during the crash, what kind of lens should we use.
And so, we did this series of shots of Brad inside of a gimbal on a bluescreen during the crash performing the heart attack up to the point where he’s about to crash. Everything that is outside of the windows of the plane had to be background plates that were acquired. I did a second pass of aerial photography, but this time without the Cessna plane. I had a six-camera array mounted on a chopper, and we basically followed the same flight path to acquire a spherical map of the background that we could use in comp for everything that is outside of the window.
Now, as we get closer to the lake where the plane is going to eventually crash, we’re starting to collide with trees. So for that, we found a really amazing location in New Zealand for the lake. Obviously, we couldn’t bring a plane and crash it, but we used that location as the foundation for a CG environment, which was handled by Rodeo FX in Montreal.
I went there with a small team to do an extensive survey, which was both LiDAR scanning and texture acquisition, but also shooting background plates of angles that we thought would be good to support the plane coming through the trees. And for that, again, having the previs from the team at Day For Nite was really helpful to set up the different camera angles that David would sign off on.
So we did that, we shot a couple of background plates, we surveyed that location, we handed that over to Rodeo FX, and Rodeo basically recreated that location in CG with trees that could be procedurally animated and interact with a CG version of the Cessna plane. The plane itself was also scanned and heavily referenced, and Rodeo did a completely seamless match to the real thing. There’s a lot of shots where, for instance, we’re outside of the plane, what you see of the inside of the cockpit is a plate of Brad and the dog on the gimbal, but all of the exterior of the plane is CG so that we can have complete control over the lighting and the interaction with the trees.
The last part of the puzzle of the crash was getting the plane into the water. And again, in early conversation with David, I understood he wanted to have as much interaction in-camera as possible. So that’s yet another rig where we mounted a hollowed-out cockpit of the Cessna on a massive truss rig with a water tank outside on a backlot in Auckland, and we dropped that Cessna shell into the water tank as fast as we could to get the biggest splash we could get. So the couple of shots where you see the Cessna colliding with the surface of the water, that’s a real Cessna cockpit hitting a real water surface. All of the environment around the backlot had to be replaced because that was a parking lot and a bluescreen, so again, we used Rodeo’s environment assets from the plane crash to populate those backgrounds.
b&a: Another sequence that I really loved is the river rapids crossing. What’s really great about that sequence is, it’s really extended. It’s quite a long scene. What I was thinking there, you would really want to have a lot of control over the shooting of it. So I assume it’s not necessarily all shot on some location that you could find. What were the ways that you had that control?
François Dumoulin: It’s a sequence I’m really proud of because we went with a methodology that’s a little risky, but I think the payoff was phenomenal. And I think visually it’s stunning, and it’s also fitting into the world of the movie quite perfectly. We had to deal with a set of constraints when it came to the methodology for that sequence. The first constraint was, we found just one location that looked good for our story. We found tons of great-looking forests, fantastic-looking mountains and glaciers, and all kinds of things in New Zealand. For some reason, for a dangerous river with big rock formation and white water churning and the cliffs that we needed to tell that story of the dead tree across serving as a bridge, we only found one really perfect location in New Zealand, and that location was completely impractical. The only way to access that location was a helicopter trip, you land in a field, and then you walk with your equipment on a backpack for about an hour. So you absolutely can’t bring a film crew to a location like that, and you can’t build a bridge and safety riggings and bring your A-list actor and dogs. That would be just complete insanity.
So that plan was canned very early on, and David really didn’t want to do any CG water, so he asked his production designer and his team, ‘Can we build the entire thing on a backlot?’ And by the entire thing, he meant, ‘Including the water.’ At some point, there was a plan to build a section of that river and have a massive water system to pump white water and do all of those things mostly in-camera and limiting the visual effects contribution.
But the cost for that set was just completely prohibitive. That’s when the producer said, ‘Look, we understand, David, you don’t want to do a big CG environment, but we have to find a solution, and VFX is going to help us find that solution.’ That’s when I came in and said, ‘Well, I hear that we don’t want to create a CG world with CG water, and I agree with you. CG water is something we can nail these days, but it’s time-consuming, it’s expensive, and we’re going to be spending hours and hours questioning why that particular swirl looks weird, and we have bigger fish to fry. That shouldn’t be our focus in post. So I need to find a solution where the water is practical.’
So then I came back to that initial idea of the location, and I took a trip in the chopper with the production designer and the director of photography, and we spent some time there and we really analyzed it, and the thinking was: we can’t build the entire thing on a backlot, we can’t shoot here, but we can maybe create an hybrid with a section of that real river, we’re going to scan it using a LiDAR scan, and we’re going to use that 3D data to print rocks on our backlot. And we’re going to have rocks on our backlot that are going to be a 1-to-1 match to a real location. So basically, we’re going to have a blend between a real location and the backlot. We’re not going to get into post to start extrapolating and asking ourselves, ‘How does that blend with the background and what’s the distance to the water, etc.’
View this post on Instagram
The production designer and I would sit on a rock and say, ‘Okay, the tree is going to rest exactly on that rock, and the other end of the tree is going to be there,’ and we marked all of those positions on the location. And so when we did the scan, the 3D acquisition of that location, we knew exactly which portion we would extract and reconstruct on the backlot. The director of photography was also with us. We would look at what would be the best time of day if we were to shoot at that location, what was the best position of the sun to get a nice backlighting on our tree bridge, and we said, ‘Well, if we are going to build a matching set of rocks on the backlot, we need to make sure that the position of the sun lines up.’ We really made everything work with the philosophy of: let’s try to do something that can really line up with that location.
We then built that set on the backlot, it was really great looking, and we ended up doing a lot of very complex camera moves with the crane or with drones that prevented a simple A-over-B solution where we would just shoot the backlot plate and then the background plate. It was more involved than that. But at least we had a plan where the design was there, and the idea of acquiring backgrounds based on an exact main unit camera angle was really solving the part of the equation where we needed to do CG water. We really scratched that off: water is all going to be based on photography, always.
During main photography, what I had was what we call a Total Station from Leica. It’s a 3D scanner that is used in construction where you don’t get a detailed point cloud, but you get really fast iterations of positions of objects in space. And what that gave me was for every key camera angle and key camera position, I would know the relation between the lens, Brad, the bridge, the rocks, and I could then triangulate those positions back into the real river location and say, ‘Well, if our camera was there on the backlot, it would sit here in the 3D space of the location.’ We would then do a selection of key takes, key shots, right after main unit photography. We’d do a very quick match-move, import that match-move into the 3D scan of the location and determine the camera position at the location. We knew we could not bring heavy equipment at the location, but we could bring a drone, so that’s what we ended up doing. We brought in a heavy lifting drone, which we were able to mount our Alexa camera on with the same set of lenses, and we were able, with the point cloud acquisition from the set, to send the drone to the exact position. We were not doing exactly ‘motion control’ background plates, but we were acquiring plates from the key camera position, and all of that data was then the foundation for all of the visual effects work.
At the end of the day, we still had Framestore building a version of that environment in CG, so there’s a handful of shots, the wide shots primarily, where there’s a connecting tissue between the backlot plate and the location plate where we needed CG rocks and where we needed CG trees to make that blend happen in a completely seamless way. Of course, when you work on a backlot, you’re dealing with a real sun and clouds, which is why it looks so real compared to studio lighting, but it is also changing constantly, so for the lighting conditions when we acquired the main plate at the backlot and when we went back to the location, sometimes we would not get exactly the same sun or the light would get more diffused, so that’s where also having a full CG version of the world with Framestore allowed us to fine tune the lighting. But what was great at the end of the day is that the water itself, the river, was always in the shadow because it’s kind of this little gorge where you don’t really get any sun in there.
I knew that lighting would never be an issue when it comes to using photography for the water, and that worked really well. In most of those final shots, yes, there are CG assets for the rock and the trees just to blend things together, but the core elements and the design foundation is really all based on a real location and real photography.
b&a: I was going to say it just feels so natural, and I think you mentioned up front that that can be the risk sometimes of filming a partial set with bluescreen. I mean, sometimes things like that even get filmed indoors and sometimes the lighting just doesn’t match. But all those things you did seemed to keep it looking natural, which is amazing.
François Dumoulin: Yes. I have a background as an artist, I used to do comp, and many times I handled plates where what I was asked to do, the intention was completely the opposite of the way it was shot. They might want you to put a background matte painting with the sun on the left, and the character has the sun on the right, stupid things like that. So I think in my position of supervisor now, my role is really to make sure that people on set have a really good understanding of how the visual effects workflow functions and when things will break. The best way to achieve that is to make sure that people on set have a good understanding of what the final product is going to look like. That’s really my role.
For example, given the fast pace of production, an assistant director is constantly juggling a massive schedule and looking for ways to maximize the cast’s time. Because of that, there can sometimes be a temptation to treat a heavy CG character scene as a secondary setup….Well, it’s my role to fight that and say, ‘No, you have to consider that shot just like any other shot of the movie. It’s a drama shot. I want the cast to be there, I want the director to be there, I want the director of photography to understand where the bear is going to be sitting so that he can maybe decide that it’s best to shoot in the morning. I want the set decorators to be there and understand that the bear is going to be crawling over that grass, so maybe it’s more interesting if we dress that foreground with more vegetation.’
In order to achieve something like the CG bear, you can’t just do a tennis ball on a stand, you have to have something on set that is a representation of a performance. I know you’re a big fan of puppets and stuffies, Ian, I am too! So I insisted on getting someone with us. Every time there was an animal involved, I really insisted that we get a performance in front of the camera, not just for Brad to react to, but for all the departments to react to.
It’s something that I really learned while doing Prehistoric Planet, where we used a lot of puppets. I actually used the same team. I called my friend Brian Fisher at Practic Creations in the UK who did all the puppets for us for Prehistoric Planet, and he built all of the animals on Heart of the Beast in the same fashion. It’s all very, very simple, very cost-effective, it’s all very lightweight, it fits in a suitcase, but Brian manipulates those puppets in a way that, instinctively, everybody on set gets it right away. At first, the first few takes where Brian was playing with his puppet bear in front of Brad, people were kind of laughing at us, but after two or three takes, they just get it. They understand there’s a bear in front of us, and it’d show them, oh, we need to adjust the frame to it, the camera needs to react to it, we need to pull focus with it, the lighting needs to be adjusted considering what the scale of this thing is, what the eyeline is, etc.
I’m always a big advocate for finding solution in order for everyone on set to understand what the visual effects goal is. For a lot of people on set, visual effects is just a black box, they don’t understand what we do, they don’t understand why we constantly need to shoot Macbeth charts and gray balls. Those things are like a mystery to them. But the minute you start doing things that are more tactile, like a puppet, or when we were shooting the tree bridge sequence, for instance, I had acquired a Gaussian splat of that location, I had it on my phone, and I was able to show people, ‘Look, this is our world, and when you look in that direction, this is what you’re going to be seeing.’ So every time I can find a simple solution that allows every department on set to understand what the final product is going to be, it always plays in favor of the visual effects work at the end of the day.
b&a: One of the most complex scenes must have been the wolf attack. I mean, you’ve got Odin there, but you’ve also got—is it four or five wolves—plus Brad Pitt’s character. How did you plan for that, and what was your methodology for shooting and executing that sequence?
François Dumoulin: You are asking about the wolf count which is just something that makes me laugh because it was this nightmare of a sequence that really kept me up at night for a while. It was really hard to execute. We initially conceptualized a massive pack of wolves, but narrowing the focus to five or six specific animals ultimately made the choreography much tighter and the danger feel more immediate
When I joined the show, everybody was afraid of that sequence for a lot of reasons. Obviously you’re not going to shoot with real wolves. You can shoot with Husky dogs, but they’re not going to perform like wolves. So we knew right away that there would be some really challenging interaction with an A-list actor and some form of digital animals. And that was making everybody nervous. So the first thing that I did was, it’s pretty much Framestore’s motto, when you get into a creature-heavy sequence, you gather a lot of reference. So I spent hours and hours and hours watching documentaries, finding videos online of packs of wolves attacking other wolves, attacking other creatures, attacking humans. I’ve looked at some really gruesome footage, and I started to understand a few rules that we discussed early on with the stunts team, because obviously a lot of that sequence is performed by a group of stuntmen that perform as wolves.
We also knew that the dog would not be able to do all of the action, we’re not going to have the real dog being dragged and attacked by stunt people, that would have been a disaster. Anyway, we knew there would be a CG dog involved, so basically all of the creatures that you see in the scene, both the wolves and the dog, needed to be CG for the majority of that scene. I had to really work in tandem with the stunt department to build the choreography of that sequence and make sure that all of the behavior from the stunt guys would align with the body language and what the rig of a wolf body would do.
That in itself was very challenging. We had those stunt people perform on all fours, which was really not comfortable for them. We spent days and days and days rehearsing that scene and narrowing down the choreography with stunts and doubles for Brad, up to a point where we felt that the choreography was working.
That’s when we brought David in, and we started doing what we called ‘stuntvis,’ where David is operating a camera or just his iPhone and trying to figure out key camera angles. In parallel, I also had my team at Day For Nite doing a previs version of that with the final look of the set, because the sets were built in parallel. We had to rehearse in a gym, basically, while the sets were being constructed. The previs was working in parallel of the sets being constructed and the stunts being blocked with the stunts. The previs was merging things and updating things as we were progressing and figuring out the key beats of the choreography, which eventually allowed us to break down the shot in methodology where we would say, ‘Okay, maybe we can get a reference of the real dog for that shot.’ Or, ‘This is completely digital, there’s no way we can shoot that. This can work with an interaction on Brad where we have like a fake wolf head that’s going to bite on his arm or on his belly, and then we’re going to attach a wire to the other side of his vest so that we can push him or pull him.’
Every single shot had to find its own methodology, and that was a lot of a dance between the stunt department, the previs team pulling together the shots, and what David really wanted to capture. The way David shoots action is very particular. When I was blocking the previs, my instinct was to do shots that were very illustrative, such as wide shots where you can really understand where the wolves are coming from and how much space that’s giving James and Odin, and really doing things that were almost too choreographed and too clear. Well, David came in and really destroyed all of the set of rules that I established [laughs], because that’s the opposite of the way he captures action.
What David really likes to do is, well, first of all, he operates the camera himself. And the key camera is always David’s camera. He calls it ‘the emotion camera.’ So typically, on any given setup, there’s going to be three cameras. One is probably going to be mounted on a crane, and that’s going to be the DP doing a beautiful sweeping move showing the environment and the characters. A second camera is going to be medium, getting more into the action. And the third camera, the emotion camera, that’s going to be David right in the middle of the shots, right in the middle of the action, the lens as close as he can to the punches, and really getting into the fight. And that’s how he likes to operate. He wants to be surrounded by this chaos. So what that meant for us is that we needed to push the choreography with the stunts to a level of fluidity where instead of doing just little chunks, we could give him not a full master of the fight, but beats that were long enough so that he could really sit in the middle of that action and really capture it from the inside, which in itself was extremely challenging.
View this post on Instagram
b&a: Because you had performers in there, I’m guessing wearing gray suits or green or something like that, did you need to do a lot of clean passes, or did you just not worry about that and know that Framestore would have to reconstruct a lot of the environment and maybe parts of Brad’s body?
François Dumoulin: Well, again, in a very similar approach to Prehistoric Planet, we would always shoot our master first. And whether we were using a puppet for the wolf motion or having a stunt performer, every time, basically the philosophy was, if it’s a wolf or an animal on its own, it’s a puppet, and if there’s an interaction with the cast, it’s going to be a stunt performer. And so we would always block and capture our hero take with either the puppet or the stunts as our master, and then we would come back and shoot a clean plate of the same thing with the camera operator basically trying to do a poor man’s motion control, replicating the same move, replicating the same rack focus and everything.
Now, obviously you want to use your master take with the puppet or the stuntman, because in 99% of the time, the camera work is just going to be better than the clean plate. It’s funny, but as good as your camera operator is, there’s always something about reacting to someone in front of the lens that you can’t quite get right when you’re shooting a clean plate. Never really works. So I knew that there would be a lot of reconstruction involved.
So, we shot clean plates, we also shot a lot of tiles, we shot a lot of different focus planes. To make our life even more complicated, the lighting was made from gobos, and those gobos, like moonlight coming through the leaves, would be animated by the wind. There’s also low-hanging fog. So every time we shot that thing, it would be different. The clean plate, the light, and the atmospherics were ever so slightly different from the main hero take.
I’m not sure exactly how much of the clean plates Framestore ended up using versus brute force reconstruction. What I do know is that we ended up doing a CG version of that environment just to get all of the interaction correct, because even though it’s not in your face, but if you don’t get foot contact correct, and if you don’t get the little layer of FX simulation of leaves and debris being scattered around when a character is dragged on the floor, all of those small things make the characters and the animal really sit within the environment. If you don’t have them, everything appears to be floating. So there was quite a lot of CG reconstruction for the environment.
b&a: Finally, I really loved those flashback sequences to wartime with James and Odin. What I liked about them was that they felt very close-up on the characters, they’re not really wide shots of the war. But what were some of the practical and digital visual effects things you needed to do for those?
François Dumoulin: I really love the way the flashbacks were approached and treated. We were involved with that. Obviously, there’s some set extension and some war effects, like muzzle flashes, augmenting explosions, adding dirt, etc. But we were also in visual effects responsible for the visual treatment of that sequence, which in itself was really an interesting exploration.
David really approached the flashback as something that was more of a mental image, so it’s not really descriptive, it’s more impressionistic. You’re into glimpses of that past, and it’s interesting because David wanted a lot of texture for those flashbacks, and he ended up capturing those scenes with a lot of different film stocks, actually. Our movie is entirely shot in digital with the Alexa with a beautiful film treatment, but the flashbacks made use of actual film cameras. We shot on 35mm, 16mm, we even used an 8mm hand-crank camera and a VHS camera! [laughs]. Yes, we did! We did use a VHS camera. So those flashbacks really came with their distinctive style.
The visual effects work, as I mentioned, there’s a little bit of set extension when we see those glimpses of war, whether it’s Afghanistan or Iraq, we don’t really know. So we don’t really want to be descriptive with those things, it’s there, but it’s a background thing, it’s not meant to take you away from the emotion. There was a little bit of matte painting work involved. All of those things are blended into the fabric of that sepia treatment, where we played a lot with things like multiple exposure, defocus, vignetting of the lens, optical aberration. All of that was the work of a really cool company I’ve worked with in LA called OPSIS. They tackled all of the visual treatment of the flashback. I think they did a fantastic job. Obviously, we’ve all seen a hundred times a certain sepia look, but we wanted to do something slightly more impressionistic and really grounded in optical qualities and film look and old lenses. I think we ended up with a very unique and very distinctive look that really contribute to the emotion of that moment.
b&a: Well, I just wanted to say I loved the film, and I really appreciate you going into all that detail about your methodology, not just how you accomplished the effects, but how you integrated it into the storytelling, which is a really big deal with this film. It doesn’t ever feel like a visual effects film at all, and I think that’s because of the way you approached the visual effects work/
François Dumoulin: Fantastic. Thank you for having me.
