Wētā FX breaks down their delivery of Cokey the bear for ‘Cocaine Bear’.
Tasked with overseeing the creation of a photoreal coked-up CG bear in Elizabeth Banks’ wild and wacky Cocaine Bear, Wētā FX visual effects supervisor Robin Hollander knew that performance and interaction with the real actors in the film would be key.
The VFX supervisor liaised with Banks and the other filmmakers to ensure that an on-set bear performer would be there with the actors, along with a bear stuffy. He also devised a clear workflow for the filmmakers to follow in shooting specific clean plates, in order to help with the visual effects process down the line at Wētā FX.
In this in-depth befores & afters interview, Hollander breaks down that plate shooting process, and details the research and build stages for the CG Cokey. Plus, in special stuffy news, Hollander also reveals that Wētā Workshop actually built a bear bust that could snort cocaine via a vacuum pump…but it was, unfortunately, never ultimately tested on set.
b&a: How much time did you spend looking at bear reference in preparation for this movie?
Robin Hollander: Yeah, it was definitely a lot. One of our assistants had the gruelling task of going through hours of footage and trying to compartmentalize it, so that we could find reference of different bears, walking bears, talking, sleeping, climbing, etc.
Elizabeth and her husband Max would frequently send me Instagram links and say, ‘Hey, here’s a bear doing something…’. Through an initial exploration both on our end and Liz’s end, we tried to develop the character first and foremost. Who is Cokey? What isn’t Cokey? What could she do? What do we definitely not want her to do?
We narrowed it down and we quite quickly ended up with sun bears. Sun bears are incredibly goofy looking, although they’re quite ferocious when they have to be and they just look a bit cooked at all times. Out of all the exploration, we came to the conclusion of, ‘Let’s model Cokey on the sun bear as much as we can.’ It meant that we could focus our attention in reference seeking around the personality that Cokey brings, rather than just going through every black bear or grizzly bear video that we had.
The scene at the gazebo where Eddie and Cokey wrestle, or dance–that actually came about from one of these reference gathering quests. We found this clip of a bunch of guys in the States and then later in Russia actually wrestling bears. We showed it to Liz and said, ‘Wouldn’t it be funny if Cokey wrestles one of the guys?’ The beat was in the script, but it wasn’t quite the way that you see it in the final movie. So it was really nice to see that that planted a seed in Liz’s mind and I think speaks to the relationship that we managed to build as well throughout the movie.
b&a: What departures from normal bear actions did you need to take, given how Cokey was coked up?
Robin Hollander: Initially we got a good instruction from Lord and Miller that they didn’t want this bear to be goofy at any point. In all the reference case studies we presented for how twitched out she could be, early on the brief was like ‘Nah, nah. Too much.’ She’s just a little bit over the edge, but at the end of the day she can handle her drugs pretty well.
Some of the references that we looked at for animation were pretty grim to watch, to be honest, particularly the ones of animals having seizures or issues with their brain. The scene where it’s most evident is when Cokey arrives in the gazebo, which is basically after she’s likely done a lot of coke, but she’s also just been in that car crash and she has a bit of a slurred, almost drunk, movement. That is probably the biggest departure, I guess, from normal bear behavior that we have in the movie, short of the scene later in the gazebo where she actually ingests it and she becomes a little bit frisky and enamored with Eddie.
In exploring how coked out she would be, it really helped us flesh out the animation rig that we had for Cokey. In order to match some of the reference clips, we had to go in and update the rig, and in some cases, break the rig. She had to be able to do things to camera that basically would’ve broken her jaw if she had done that in real life. Even though we didn’t use all of that later on, it allowed us to come up with a rig that in the end was really versatile and allowed us to hit all the subtleties that we were asked to later on.
b&a: In terms of an on-set performance, how was that handled on the film?
Robin Hollander: We pitched to Liz early on that we felt that she needed to have a performer on set because it makes everyone’s life easier. It means that she has someone to direct, she has someone that her actors can act against and riff off, and it sets a certain cadence to the scene as well, so your DP can frame for it. It just makes life easier for everyone all around.
We were very clear early on that we wouldn’t want to call this a motion capture performer, because there’s no real easy way for us to actually capture the motion of an actor playing a quadruped. It’s not possible for them to walk like a quadruped would. The only time we used motion capture was for previs because at that point, it was a little bit more like moving chess pieces around, so we weren’t too bothered by having all the mechanics link up one to one.
We also pitched Alan Henry, who is a local stunt performer. He interviewed with Liz and she really liked what he brought to the table. What was interesting in that process was that she interviewed a few other people, and I think one of them potentially was a Cirque du Soleil performer. She felt that was too much on the performative side, trying too much to be an actor rather than someone who’s like, ‘Oh, you want me to be a bear? Yep, I can grunt and I’m physical and I can get thrown around. No worries.’
It was really nice having Alan here in the pre-production stage, while we were trying to flesh out who Cokey would be and who Cokey wouldn’t be, before he even was sent on set. We could show Alan concept art, we could show him all the studies that we were doing in 3D. So, before he even got on a plane, he had in his mind what he was going to look like on the screen, which I think really brought an aspect to his performance that would’ve been really hard to capture otherwise.
The other thing that we pushed for was to have a representation of the actual bear on set, so we got Wētā Workshop involved as well. They built us a life-size bust of Cokey. We like having a pretty good lighting reference on top of the gray balls, chrome balls, and color charts. It helps us figure out how the fur reacts to the situation. It helps your DP frame, it helps your DP light, and it gives you a physical component to act against.
Actually, quite often between takes, the Cokey stuffy would be resting on a C-stand, and the actors would go up and touch her nose and interact with her. It felt to me like it was what actors do normally in between takes, just getting to know each other and riffing off of each other. You’re getting a little bit hyped up about being able to stare this beast right in the eyes.
Another fun fact with the stuffy was something that we built in but unfortunately didn’t quite test out. As part of a throwaway comment to Wētā Workshop, we said, ‘Wouldn’t it be great if the bust could actually snort cocaine?’ So they actually did that. They built two little holes into her nose and there was a vacuum attachment on the inside. In theory, you could have attached a vacuum and actually snorted up some coke, but unfortunately we never got around to testing that.
b&a: Oh, you never tried it!? I would love to have seen the b-roll.
Robin Hollander: That’s the level of detail it had.
b&a: I love it. I’ve talked to a few Wētā supes over the years about integrating a CG character into the final shot. And I think even I gloss over that sometimes, in terms of clean plating and just the right approach. Can you break that down here–what worked for Cocaine Bear in terms of shooting reference, shooting cleans and just what you got out of a plate so that you could integrate the CG bear as best as possible?
Robin Hollander: We were incredibly fortunate that the whole on-set team and our lovely first AD, Nick Thomas, were incredibly gracious in giving us all the time for the plates. Being on set, time is of the essence. Standing around costs money. So for us to go in and just shoot all the passes that we would love to have in post is often not possible because you have to keep moving. But it all worked out really well and I think we got all of that time.
Our pre-production period was fairly substantial, so we could really put a lot of thought into what we wanted to capture and how we wanted to capture it. In the first meeting we had with Liz, she gave us a few pointers as to what was crucial to her. One was the personality of the bear. She didn’t want the bear to be too bloodthirsty and she wanted it to have subtlety. The key one, too, was that it had to be photo-real, which we obviously took on board.
But I remember Liz turned to me and said, ‘Well, other than making a photo-real bear, what are the challenges? What do you think is the hardest thing to do?’ I rolled back into my experience as a compositor and my reply was pretty quick: ‘Adding anything into a forest with lots of foliage, leaves, grass and twigs, it’s pretty hard to do because it’s almost impossible to key it. It’s almost impossible to roto it. If you do elements, it’s kind of hard as well.’
Identifying that that was going to be the challenge really helped us, and we put together a bit of a package. Step one was the bear performer, Allan. People were asking, ‘Do you want him to wear green, or do you want him to wear blue?’ We said, ‘Neither. We want black.’ Because, if he’s coming from behind the bush, it’s going to be really hard to paint him entirely out of the scene. If something gets left behind of Alan’s, it’s going to be black and we are going to add a black bear to it, so it’s going to be pretty forgiving. If he’s wearing a bright blue suit, which maybe would key nicely against a green foliage, you have to remove him everywhere and it’s going to be a nightmare.
We then asked them, ‘Look, for every take, please shoot a performer plate, then shoot a clean plate.’ And then we had what we called the utility plates, which are your balls, your charts, and your stuffy.
Then we added a plate using two-by-two meter blue Styrofoam boards. I said, ‘While your A-cam is still loosely in position–it doesn’t matter if the operator’s deviated off a tiny bit– just shoot me a take of one of the data wranglers holding this bluescreen board walking behind the bushes, mapping out the path that Alan took.’
This is because once you receive a plate back and you just see out-of-focus bushes stacked off into distance, it’s really hard to know where this performer actually was. Having a bluescreen pass, so to speak, as a walk through gives you an understanding of what is foreground, what is background. It’s also a lot easier to just take a still frame from that, key it off, track it back in. The camera might be off ever so slightly, but your lighting conditions are correct. It is loosely the right angle and perspective, and it is generally a lot more successful than what you maybe would traditionally do, which is stick a couple of bushes in front of a bluescreen on stage after you’ve wrapped shooting.
In the end, there are not that many shots in the movie where Cokey actually weaves through the bushes, but for the plates where we needed it we had those passes, which proved to be incredibly useful. I think that’s something I would like to try and incorporate on future shoots, shooting a per shot element as a very quick loose pass – something that doesn’t hold up the crew too much but actually buys you a lot in the back-end.
b&a: Obviously, Wētā has done furry creatures before over many years, and I think you’ve even done a bear for one of the Wolverine films. But with Cokey, what did you find, once you got into look-dev and putting her into shots, were the bigger challenges of a furry creature on this film?
Robin Hollander: The biggest issue that I think you always find with any big heavy creature is making it look big and heavy. So for a bear, especially as they walk, there’s a lot of jiggle that’s triggered by broad stroke movements. With Cokey, we also had to be very expressive and hit certain performative actions in a very close-up framing of her face. With both of those issues, the low frequency jiggle and the high frequency close-up detail, we found the nature of black fur made it quite easy to over-absorb all of that performance. So quite often what we would see in the animation puppet, or just as a skin render, gave us the performance that we needed, but by the time we actually lit and rendered our creature solve, it was like, ‘Oh, we kind of lost a lot of it,’ because the fur and underlying fascia and fat layer just kind of dulls everything out.
While we were tied to the lighting conditions on set, we often eked out a bit more shape with little helper lights – we did a lot of that for the eyes to make sure we got a good read on them. We also added lots of little helper lights to make sure that her snout felt really nice and wet, and we also tried to shape any blood and gore quite a bit.
If you looked at just the raw renders coming out of our renderer, quite often it looked over- lit because we had all of these utility light passes, which we’d then selectively introduce in the areas where we wanted to show a little bit of spec or add shape to the performance. Then once we’d gone through that whole cycle, we’d quite often have to go back to either animation or the creatures department, or sometimes the models department, to sculpt more definition under the fur in the fat layer.
If there wasn’t enough jiggle in there, rather than dialing up our jiggle settings in our creature-solver, which would’ve made it look like a piece of rubber, we would just go in and actually sculpt in the footfalls and make sure that those bits that we wanted to jiggle, through the context of lighting and rendering, would be amplified. If you looked at the animation, you’d say, ‘Well, this looks broken,’ but by the time it comes through with all the fur and through all the tissue and the lighting, it held up really well.
In the night time scene around the waterfall, that issue was amplified a little bit more because now you have a black bear in a night time environment, which made it harder to get a good read on it. But, I feel the recipe that we ended up creating was really successful, as we got all the shapes and all the performances to read really well.
b&a: Her fur is sometimes full of dirt or blood and then white specks, i.e. cocaine. How did you tackle that? Was that just another kind of texturing challenge or fur challenge? What were the specific things you had to do there to make that really stand out?
Robin Hollander: Most of the coke was an FX component that was simulated back on top. With the fur count that we have on Cokey, it becomes pretty heavy to sim on top of something incredibly complex and dense, such as millions of strands of fur. The cocaine simulations on the fur were pretty much limited to the snout and then her wider face shape. The rest was handled as a lookdev shader. We had a textural component that would frost the tips a little, then we would complete the picture with compositing by tracking on, for lack of better word, some extra little bits and bobs.
In terms of making the coke read, the biggest challenge was the coke on her gums because it had to feel like it was wet and it was clumping up. We were playing around with that and at one point it suddenly looked like rancid yogurt! We wanted it tp still look like fresh cocaine, but just distinguishable from the stuff that was in her fur. Most of that was handled in lookdev in the shaders that we had, and again, we gave it that final polish in compositing.
For the base look of Cokey, we ended up having four different variants. Stage one was her in most fluffy form–she’s just been to the hairdresser and she looks really good.
Stage two was a little bit of clumping and a little bit of blood in her face.
Stage three is post tree attack, when she’s absolutely drenched in blood and post ambulance crash as well–lots of clumping of fur, lots of spec, lots of dark-matted blood, lots of twigs and leaves.
Stage four was the wet look – it was like stage one, nice and fluffy, but absolutely saturated as that’s what we needed for the waterfall scene at the end. It actually really helped us in that environment because initially Cokey wasn’t going to get wet throughout the entire movie, but when the beat was added where she falls off the waterfall, we had a really good opportunity to just soak her in water. This then made it so much easier to light her in that environment and get all the shape out of her that we needed to.
b&a: Was there a ‘trickiest shot’ in the film from your point of view?
Robin Hollander: Oddly enough, the hiker beat at the beginning where Cokey rips Elsa through the bushes, that was probably one of the hardest shots in terms of actually extracting the bush and adding Cokey back in, because she had all of this movement and you could see Alan standing there in his black suit nonetheless. Just extracting all the foreground stuff and adding Cokey was tricky.
In the tree attack, we had some really tricky shots, just in trying to place Cokey in there for the scale and for the angles that we needed. The ambulance attack had some really tricky stitch comps. The beat when Ranger Liz, for instance, flies out the back on the gurney was stitched together from two plates. One was shot on the stage and one was shot out in the field, so that was a challenge. The end of that scene where Beth, the EMT driver, goes flying through the windscreen and there’s a digi-double takeover, that was also a challenge.
If I had to pick one, I would say it was probably one of the shots at the gazebo of Cokey dancing with Eddie. The way we shot that is we had Alan wrapped in this big foam bandage around his neck to match the circumference that Cokey would’ve had, which meant that the actor Alden Ehrenreich on set could actually hug Alan. Even though he was a lot smaller than Cokey, he had the right kind of dimensions. All the contacts that you get from his hand onto Alan’s black spandex suit, and the color casting and all of that, just really transferred really nicely into comp. We couldn’t quite believe how easily it came together in the end, given how complex it could have been.