Stake Land: An Interview with SINNERS VFX Supervisor Michael Ralla and VFX Producer James Alexander

On twinning technology, the proper uses of AI, and not making shit up

before Michael Ralla and James Alexander got their hands on it

After Michael Ralla and James Alexander got their hands on it

There are over 1,000 visual effects shots in the horror blockbuster Sinners, in which Michael B. Jordan plays twin brother juke joint owners put under siege by a pack of bluegrass-loving demons. But until the vampire-combustion fireworks display at its close, the effects are so seamless it would be difficult to isolate exactly what images were worked over by the VFX team led by Supervisor Michael Ralla and Producer James Alexander. They worked closely with director Ryan Coogler, DP Autumn Durald Arkapaw, and Production Designer Hannah Beachler to make sure the CGI maintained the unified vision of the crew, and took painstaking care to preserve every intricacy of Jordan’s swaggering performance of the twins Smoke and Stack. I spoke with Ralla and Alexander about the beautiful imperfections of film, the proper uses of AI, and how they make sure not to “make shit up.”

The interview below has been edited for clarity.

R. Emmet Sweeney: I watched Sinners again last night. And the one thing you notice about the VFX is that you don’t notice them.

James Alexander: Highest compliment you could pay us, Robert. Thank you.

RES: Could you talk about the pre-production process and how you’re working with all the different departments? Together with makeup, costume, etc., to make all these puzzle pieces fit together.

JA: Yeah, Sinners is a movie where all of the visual effects need to support the craft in front of the camera. And that’s true across all the departments, really starting with cinematography and the large format film, 65 mill, that the film was shot on, two different formats, understanding that and making sure that the digital work was able to be integrated with it seamlessly. Going back to pre-production that you asked about, Robert, it’s really starting from that very first stage, reading the script, understanding what the visual effects needs are, Ryan’s vision for the film, and then understanding how we would need to shoot things to achieve them most successfully.

RES: You mention that you shot in two film formats. What are the challenges of working in two separate film formats as a VFX artist? As opposed to an all-digital workflow.

Michael Ralla: That was something that we learned about as we were going through this project. I gotta be honest, at first I wasn’t thrilled about that. I was like, “Oh my God, that sounds painful”. I’ve always been a huge fan of digital cameras, they’ve become so good in terms of sensitivity and dynamic range and noise floor and spatial resolution that I was like, “Why? Why would anyone still shoot on film? Doesn’t make any sense to me”. And then James and I were invited by DP Autumn Durald Arkapaw and director Ryan Coogler to this test screening in Playa Vista at IMAX headquarters. And we saw some test footage that cinematographer Autumn had shot. This is a private theater that’s not accessible to the public. The filmmakers go to approve their films. And we did go there to approve the final print of Sinners. It was somewhere out in Lancaster, east of LA. And it was the widest image I’ve ever seen. Also known as The Hateful Eight, 2.76 Ultra Panavision format. Five perforation holes on each side of the film strip. The film’s going through vertically and we’re using a 1.33 anamorphic lens.

And then all of a sudden it switched without any warning to full 15-perf 65mm IMAX, which is this massive negative. But it is an incredibly tall 1.43 aspect ratio. So all of sudden the screen exploded to double the height. Yet the magnification level was just insane because the projection screen is as tall as a six-story tall building. And so I was scared. I walked out and felt I didn’t really know what to expect and what that would mean. There’s only been a handful of feature films made in that format. Chris Nolan is spearheading that. Nope, The Hateful Eight. There’s very few films in general. Now Dune 3 is being shot in that format as well. The Odyssey, obviously, but you can count them on both of your hands.

RES: Are you learning a new workflow?

MR: That’s exactly what I’m getting at. There are very few people who actually know how to deal with this. And as we were getting into it, we got on a call with John Dykstra, who did the VFX production on Hateful Eight. That was very informative. But as we were discovering more and more, we learned that essentially time stood still and we were gonna fall back into a workflow from 25 years ago. With old DPX files and without CDLs (color decision lists) or any advantages of the effects and production technology that we’ve had over the last years. That was one side of the chain. And we’re dealing with massive 8K scans that also take a long time to get digitized because one frame in IMAX took 30 seconds at first. So we were trying to get turnovers out and with digital it would be almost instantaneous, but with a film scan, it takes two weeks.

As we were going into VFX production, we learned that working with scanned film has become a lost art. Because over the last 10 years, most people have never touched a scanned film negative. But we also understood really quickly that the beauty of film is the cumulative sum of all of its imperfections. That’s what makes it look great. The imperfect gate weave and the halation and the grain and then the lenses that are in front of it that have a really heavy optical footprint with really heavy aberrations. All of that we hadn’t seen in a long time. But it was clear we would need all of that on our perfect digital renders because they normally don’t have any of that and they would stand out like a rose in a cotton field. We had to learn and we had to research and we had to develop tools to actually mimic all of that so that our VFX work is truly seamless.

RES: Was it difficult to find labs to do the processing?

MR: Well, there’s one lab left in the United States, and that’s FotoKem in Burbank. Now we’ve become really good friends, and we’re still talking to them. It’s something that I deeply enjoyed, even though I didn’t at first. It’s crazy to say as a German engineer, but there is something outside the frame that film captures that digital doesn’t and I can’t quantify it.

The Four Troublesome Heads (1898)

RES: Absolutely, yeah. The resolution is just as high now, but it’s different qualitatively somehow. Now, I know you guys have talked about this a lot already, but the biggest challenge was the twinning effects, right? Could you talk about your research into that process?

JA: Twinning as a visual effect goes almost to the beginning of cinema. In doing our research for this film, it was really a history lesson. Georges Méliès is a pioneer of visual effects. And he created that iconic image of the spaceship crashing into the moon. And one of his early films, 1898, was called The Four Troublesome Heads. And it’s a short film in which a magician takes off his head and puts it on a table. and then the heads are looking around on the table and he has multiple heads all looking at him. So creative people and visual effects pioneers understood that that was a technique that could be used with film going back more than 125 years. There’s been a progression ever since then throughout cinema as the technologies become more sophisticated and we’ve been able to repeat camera moves and have cameras moving and have multiple components within a shot to create twinning effects.

Cronenberg’s Dead Ringers (1988) is an iconic example of where we first really saw dynamic camera moves and twinning effects. And then you look to tentpole examples of the last couple of decades like The Social Network (2010) with the Winklevii twins. And Duncan Jones’s Moon (2009), where Sam Rockwell fights himself, and Tom Hardy’s movie Legend (2015) about the Kray twins. There’s so many examples, and then more historical stuff. The Parent Trap (1998) was another one that we looked at. But there’s many, many examples of twinning through cinema history.

RES: There is the recent Robert Pattinson one, Mickey 17 (2025).

JA: Yeah, absolutely. So that was in production when we first started. But it’s a Warner Brothers picture, so we got a glimpse behind the scenes. And just to finish the thought about the research into the history of twinning effects, it was really looking back and thinking: All right, well, here’s the most simple approach that we could use, the most simple methodology with a locked off camera and a split screen. And then we can track how it’s become more and more complex with moving cameras and interaction between the twins. And we could start to assemble our toolbox from which we could then draw and say, “Well, here are the variables within this shot. What tools do we have at our disposal? What methodologies can we use?” And really the bedrock of that decision process was always deferring to the photography, always making sure that we were honoring Michael B. Jordan’s performance, because the success of any twin shot fundamentally relies on that. Michael’s performance is everything.

But as we got into pre-production, you’re absolutely right. Mickey 17 was a project that was underway. Warner Brothers and the team at Rising Sun in Australia have been working on that particular property and they’ve been doing incredible work with head replacements and so we did end up working with Rising Sun for that work in Sinners and the head replacement shots really come into play when there’s complex interaction between the twins, particularly in the final fight scenes and everything else. And that’s why we developed the Halo rig, which was an advancement in the technology and an approach that hadn’t been used before. I think Michael would probably want to speak to the Halo rig. But again, Michael’s performance was absolutely fundamental to the success of those shots. The head replacements are always purely Michael. And his performance is always exactly recreated in those shots. And Michael can talk a little bit more about the Halo rig and why we use that tool specifically.

MR: Michael’s performance was 100% the cornerstone of the success of the film and specifically the twinning work. And Ryan made that very clear from the first meeting on. It has to be his performance. You cannot change it, you cannot synthesize any of it. Like this is what carries the film. And that’s also the point where we really have to give Michael a lot of credit for the character development he’s done with the twins. They have distinctively different body language and very different mannerisms. They move differently. That’s really what it takes to sell them because If that works, then we’re really quote unquote, just assembling bits and pieces of Michael’s performance. There was a little bit more to that on the VFX side, to really keep it all authentic and have the right fidelity for that. But as Ryan sometimes put it, it was Michael who got us all jobs and who kept us all employed. And so we were trying to absolutely honor that.

The Halo rig was born out of necessity for what I quite frankly already saw coming when I only heard the log line for the story, before I even read the script. So he’s playing twins, there’s vampires in the movie, I’m pretty sure they’re gonna be fighting against each other. And I remember, before we had even been sent the script, I told James, we’re gonna have to find some way to get Michael’s head on what will likely be a stunt double for really intricate fighting. You can’t just shoot two passes for that kind of stuff because it gets too complicated. We did shoot two passes for the initial twin establishing shot, which is the wrap-around twin cigarette handover at the beginning when we reveal them. That was really difficult to shoot. And even though we did a test six weeks before principal photography, I was still dreading the day that we were going to shoot it because I knew it would be really hard. And that’s also what Michael and Ryan are saying. It’s probably the hardest shot of the film, but for a good reason, because the way we did it was basically the way you shouldn’t do it.

And I said that to James, we’re going to do this exactly how our peers would not expect us to do it. This would be the one shot where you use a head replacement. Because there’s a cigarette being handed over. And initially, Ryan talked about doing it one or two times. Turns out he wanted seven times, including the light-up moment. So that upped the stakes a little bit more. And then, of course, Ryan Coogler being Ryan Coogler, he wanted to shoot it in one take. It didn’t make our life easy, which was the mantra of the film. No one made each other’s life easy just for the sake of making it easy because we all wanted to make a good film. So that was when we really tried to push things. Also because that was going to be a tentpole shot that would cement the twins for the rest of the story. James lovingly calls it “the footprint in the snow”, which is the big AT-AT Walkers in The Empire Strikes Back (1980), where it was really difficult to show how that foot would interact with the snow. And so they only did it once, showed it once, and then never worried about it again, which wasn’t quite the case for our twins. But that was along the lines. What we also pursued, and this goes back to ILM and ‘90s filmmaking, a Jurassic Park approach, a dinosaur film that only had 63 digital dinosaurs. What I’m trying to get at is we try to make every single shot count and make sure that we’re not doing it just for the sake of showing off.

But back to the halo rig. You cannot shoot fight scenes as double passes. There are too many moving pieces. There’s a camera that’s moving, but there’s also two actors that are in each other’s hair. The variance is just too big. Ryan is a sports guy. He likes to build momentum on set and get everyone into the groove. So whatever we were doing needed to fit into that. That was one parameter. We couldn’t just stop the shoot and send Michael away to go into a captured booth or next door and redo the scene in neutral lighting or come back three months later and do whatever’s missing.

RES: You’re on the set for the whole shoot?

MR: Yes.That’s the way to do it. So capturing Michael’s head performance for those stunt fights at the end of the film, it’s a lot of shots. It had to happen in the moment in the right lighting on location. So that Ryan and the team stay in the groove, so that Michael has optimal settings to deliver the best possible performance. And while he’s still in the same lighting, while he’s still emotionally connected to what was going on, while everyone is still in the groove of what this shot is. But we also noticed that it had to be a full 360 degree head replacement, not just the typical VFX hockey mask approach, which is just replacing the face. Michael’s an actor who’s also a director and a director who’s directed himself. He knows what his ears look like from behind. He knows what his neck looks like from behind. He knows his hairline. We couldn’t get away with the typical VFX approach. It had to be 360 degrees full head. Again, faithful to his original performance. Then all of that had to hold up to IMAX quality standards, which was another level. And once we threw those parameters together, what we ended up with was the Halo rig, which is 10 or 12 cameras on a carbon fiber ring, body worn. So that Michael could still move his head freely. It’s not a camera that’s attached to the head, it’s on his shoulders and he can move through the scene. Exactly the same way he did before and redeliver his performance as the other twin.

RES: Yeah, it’s very kinetic and natural looking. I watched the VFX reaction show Corridor Crew. In their section on Sinners they were speculating that you used the Halo rig footage as reference material for a deepfake. Were they on the right track?

MR: The speculations. They’re all over the internet and we love them because it’s fascinating to see what people think. It’s funny because I left a comment on the video to the Corridor Crew guys and was like, “Hey, let us come in and we’ll show you how we really did it!” They’re close, but not exactly. The key word here is that Rising Sun had to revamp their entire workflow that was used on Mickey 17 because this was IMAX and this had to be a full 360 degree head, not just the hockey mask that was done on Mickey 17, which was new for them. So what they did is called a GaussianAvatar pipeline. And what that is is that they’ve developed a tool set that allows you to regenerate all the camera perspectives that you didn’t have based on the 12 cameras that we did have. And then that allows you to translate Michael’s performance onto the double’s body. There’s a little bit more involved. It goes on top of a very good digi-double. We tested deepfakes because that’s what my eight-year-old son was suggesting at the beginning of the film when I told him about what we needed to do. I can do that with my phone! It was just not good enough. It didn’t hold up. And of course, that’s what everyone thinks these days, but that’s not quite what it is. Keyword: GaussianAvatars. The issue with a lot of deepfakes is that you end up with the face of the person who you’re trying to double, but it’s still the performance of the stunt double. And that wasn’t going to work for us.

RES: Generally speaking, how do you feel about AI as a tool in the VFX space?

MR: I’m playing with it and I think it’s fascinating, really interesting what you can do. I think we’re in a little bit of a bubble, especially when it comes to images and videos and all that, things are really overhyped. There’s a lot of great stuff undoubtedly and the possibilities are endless. But it’s also a little bit of a regurgitator. It’s been trained on a lot of data and somehow it’s always falling back and doing a sophisticated recombination of what’s already been out there. What was interesting is that Ryan explicitly did not want, for instance, AI-generated concepts. Generated AI, he was like, “I don’t want to look at the images with the two noses and the five eyes and the seven fingers and all that”. He would always point out something that I wasn’t looking for, because I’m looking at the bigger picture. And he would point at one specific thing and go like, “No, I don’t want a guy with four hands”. I would respond, “Oh, but the overall bigger picture is… -Yeah, no, sorry, I can’t get past the four hands. Can we just please hire a real human who’s going to draw a concept?” The movie has a lot of heart and soul because he hired people who really injected their creativity into the process. He wanted to make it better instead of regurgitate what’s been there before. I think you can feel that.

There’s a huge AI wave. Everything looks kind of cool, but then also a little bit forgettable at times. And the big thing is this movie was shot for the big screen. This was supposed to be an event. That’s why it’s IMAX. It stands for “Image Maximum”, which I only recently learned. Currently, we’re not quite there yet with AI. And I don’t know if we have to or need to with Gen AI. There’s a lot of really cool stuff where you can use it behind the scenes to automate processes and manage data. We already know AI is enormously good with organizing existing data to project certain outcomes or just for production management, cost management, all those unsexy uses. AI is enormously helpful for what we’re doing. Because you can have a coordinator sort through 10,000 on set photography stills for reference. Or you can have an AI crawl through all of them and re-sort them instead of having somebody go through manually. To me that is what AI is ready to do. That’s what we want to use it for, not necessarily to make movies or songs or music.

JA: Michael has a reputation for capturing or asking our data wranglers to capture more data on set than any other supervisor they’ve ever worked with. That’s going to be reference photography, it’s going to be panoramas of environments, it’s going to be lighting references, it’s going to be witness camera data, all kinds of things, terabytes and terabytes and terabytes. And what Michael’s explaining is where we really see efficiencies that we can make using AI in future productions, sorting all that data and using AI tools. AI is great at taking data sets and putting things into buckets, following clear instructions. And that data management that I touched upon is one use case for that. The other thing in terms of visual effects tracking is understanding how edits change. And typically we’ll have a sequence and we’ll have a number of shots within it. And from a production point of view, I’ll track an average shot cost per sequence. So when you have... edit A and edit B, you might have 20 shots in sequence A at an average shot cost of $5,000. If you then get your next edit iteration through and there’s 30 shots in that sequence and shots have changed throughout the edit, AI will be a really useful tool. And there’s existing programs that do this already, but the technology is really good at diagnosing those kinds of differences. And giving production efficiencies in that regard. So you can quickly process those numbers, quickly make those comparisons. And that information then goes to the rest of the filmmaking team. So everyone can make informed decisions on creative choices. And I think those efficiencies are going to be incredibly valuable to filmmaking going forward. And it will also allow us to focus on the fun stuff like burning vampires rather than chasing numbers and working endlessly in spreadsheets.

MR: Yeah exactly. You very often tell an AI, give me a picture with this and that, but it doesn’t necessarily understand the intent. When I say intent, it’s the story that we’re trying to tell with it. And that’s an inherently human aspect. And on Sinners, we really did focus on what I call preservation of creative intent from end to end, from pre-production all the way to final delivery. And as an example, I want to mention the digital train station that we built. Production could not find a complete train station with an actual functional train in Louisiana, let alone period accurate. We used one that we could find, which was basically only a building and a strip of dirt in front of it. And it was clear the rest would have to be digital. And what then normally happens is you shoot and then the on-set crew goes away and then VFX is left to their own devices on figuring it out. And Autumn Durald Arkapaw, the cinematographer, famously says, “VFX, why are you making shit up?”

Why? I’m like, “Well, because we have to fill the frame with something that isn’t there. So if you don’t want us to make something up, we have to shoot a reference”. So now she’s shooting a reference for everything. And you can hear her say that on the ShotDeck interviews, which is great. Because that makes her a fantastic collaborator for VFX. So when it came to the train station, there was a building, there was a whole city that had to be figured out, and the train. So in post, we started doing concepts, but then we sent those to Hannah Beachler, the production designer, and we got her guidance and her references and her whole thought process on what that city should look like. And she used us as a digital construction department. And what that meant is that what we were putting in front of the digital camera as a digital build had the same creative intent and had the same design language as what she did on set. And, we’re always talking about seamless visual effects work and how important good compositing and matching the lighting and all that is, and the film artifacts that we talked about earlier and lens characteristics. None of that can fully succeed if what’s in front of the lens, in front of the digital lens, has a completely different design language than the piece that we shot practically. So wrapping all those collaborators back into all that digital work as it’s happening is really important because Ryan hired them all for good reason, because he wanted their input. He wanted the specifics of the work that they do. And as I always say, he’s not just casting the people in front of the camera, he’s also casting the ones behind the camera and with just the same thought and thoughtfulness overall. That was the line that I gave to James when I asked him to join me on this film. Ryan’s not making films with a bunch of guys, he’s making films with his friends. And it’s an honor to be invited into that. And I wanna extend that invitation to you, so let’s please make this happen.

So we went back to Hannah at the train station. We did the same for every full CG shot, which there actually are quite a few in this film, even though you would not expect it, and asked Autumn to guide us when it came to framing and lighting. Because again, we wanted her eye and creative approval. Not only approval, but active guidance on how these shots should be shot, even though they’re now digital, was invaluable in terms of preservation of creative intent and making it fully seamless. So she was roped into post-production all the way until the end. We kept sending her images, which is unusual and many DPs don’t necessarily want to do it, but she was so invested and wanted to make it happen. I wanted to make sure that it’s good because she knew if we do our job well, and she can help us with that, she’s gonna get the credit for that. And it’s funny because I see these beautiful shot compilations on YouTube and mosaics of the cinematography. I’m looking at some of those shots and I’m like, this is an all CG-shot, but it’s awesome that it’s in the selection.

RES: How do you guys react when you see action sequences praised for not using VFX (when a lot was used). Do you take it as a compliment, or do you feel like the VFX is an overlooked and undervalued part of the industry?

JA: On this show in particular, the visual effects had to be invisible. Apart from when we set fire to Remmick at the end, no matter how realistic that shot would be, people would understand that we weren’t actually able to burn a vampire. But in every other instance, Michael and I, I remember vividly having the conversation that we need to get to a point where people are saying there’s no visual effects in this film. And here we are. And it’s amazing because this is a movie where the visual effects serve the story. The visual effects are in support of Ryan’s vision and they support all of the other departments whose beautiful work is in front of the camera. That being said, we love talking to people like you, Robert, because we’re really proud of the work.

And the fact that it’s invisible hopefully means that we did a good job. But I think visual effects play just as important a role in this kind of scenario as they do when spaceships and aliens and explosions and all the other beautiful, more obvious visual effects stuff is happening in a feature. I think it’s just as valid, just as important. And we’re really proud of the work and we’re really pleased to talk about it as well.

RES: Is there anything you think you haven’t been asked or hasn’t been focused on a lot in the film? Because I know people focus on the twinning a lot, but there’s a lot of other lovely effects. The vampire eyes are certainly creepy.

JA: The eyes are a great thing to flag, Robert. They were created by the special effects makeup department. And Ryan, very early in the production, mentioned a phenomenon called tapetum lucidum to us. And that’s where if you shine a flashlight into the night and you see eyes glinting back at you, that’s the phenomenon you’re seeing. It’s to do with how light is bounced back from the the cones and rods within different animals eyes and depending on the animal and the configuration of those cones and rods, different colors get bounced back at you. So it’s a real world phenomena and the special effects makeup team were able to come up with this new technology of contact lenses that created this effect. So we actually filmed with these contact lenses in some shots, but they had their limitations, these practical contact lenses. The actors couldn’t see too well in them. So if it was a dynamic shot and there was stunt work involved or a lot of action and movement, it wasn’t really safe for the actors to wear them. But we were able to film them. There’s a great shot of Remmick outside the juke joint and his head moves and the light catches the eyes and the red in his eyes flares up. And that was something that were we to create that effect completely digitally, the flaring of the eyes would be something that would just be too obvious were you to flare it up like that using digital effects purely. But given that we had a real world example, it was great whenever we needed to put the eye effect in, we had some beautiful real world reference that meant that we could integrate it seamlessly.

The other thing I wanted to mention real quick that we don’t talk too much about was in the train scene, the special effects department ran a smoke dolly along the platform. So we were able to film a real platform, but we weren’t able to put a train on it. So the train in the film is digital. We’re always looking for practical interaction whenever we’ve got a real element within a shot. And so the special effects team ran a dolly alongside at the correct position for where smoke and steam would be emitted from a period correct steam train. And so they were able to run that through the shot. So we got practical smoke and steam interacting with all of the passengers who were on the platform. And that also then gave us a place where we knew that the digital train would have to be precisely placed within the shot. to match the emission point of the real smoke dolly that was moving along the platform. So that’s a cool example of collaboration between the departments and also that combination of digital and practical effects that makes the final shot successful.

RES: I was going to ask if the cotton field is a similar combination of digital and practical.

JA: We had 50 real plants. They were created by the Greens department. They were geraniums with the flowers removed and cotton balls stuck onto them. And so there’s a couple of shots where you see Sammy picking cotton and Cornbread picking cotton with his wife and the plants that they interact with are real and the plants immediately surrounding them in the foreground are real. Everything else is a digital extension. So we built a Houdini system, or our visual effects partners did, that was able to create up to three million individual digital plants that then filled those environments. And they had control that meant that we could control the density of the plants, the age of the plants, and therefore the color of the plants, whether the cotton had been picked or not, we were able to look at the plate and see how the wind was moving through the crops and then replicate that within the Houdini system that created the digital plants. So it was a really sophisticated tool that was built especially for the show. And yeah, in some of those shots there are 3 million digital cotton plants that extend out to the horizon.

RES: Remarkable.

JA: Well, another thing that we haven’t spoken about very much is in the final scene of the 1992 moment when they’re in the blues club. The scarring on Sammy’s face there, he gets slashed by Remmick when they’re fighting outside the Juke. We had the great pleasure of working with the legend Buddy Guy who plays old Sammy and we shot with him twice and on one occasion we were able to have practical makeup on his face and the second time we weren’t. So a lot of those shots in the blues bar, the makeup is digital. So again, it’s an instance where we had great reference because we did shoot it practically. But also, because of time constraints and out of respect for Mr. Guy, we put it on digitally in some shots, which is something we haven’t talked about very much and no one has brought up. Again, which is great.

RES: I never would have guessed that.

MR: It illustrates the general idea. We would actually quite often start with a completely blank, empty screen. Do full CG shots on the show. But we would never start without a reference or an element or something to match to. And I think that’s a really big difference and that’s when the “we don’t make shit up” aspect comes into play again. But what that often meant is it would make things difficult. We had a stunt man being set on fire even though we knew we would likely not use that. But we needed it to understand what the fire in that moment looked like and the exposure ratio and all that stuff. And the same for the fire tornado rigs and the same for the roof burn in the surreal montage, which eventually got fully replaced because Ryan wanted to put the camera, push the camera through the roof. There’s a lot more CG work in a film than you think, but it was never done without a reference or something to match to. Preservation of creative intent.