Alex Garland’s Ex Machina is slowly emerging as a sleeper sci-fi hit, and for good reason. It’s a deft, thought-provoking piece of filmmaking that asks tough questions about technology, gender, and — like some of the very best science fiction — the inherent deficiencies in human nature itself. At the center of it all is the relationship between two characters: a shy programmer named Caleb (Domhnall Gleeson) and a robot named Ava (Alicia Vikander), who may represent the first example of truly sentient artificial intelligence.
With so much of the film riding on the dynamic between the two, the visual effects team had to make sure that Ava was immediately believable as a machine without letting design or technique get in the way of Vikander’s nuanced performance. The result is nothing short of a marvel. On the eve of the film’s expansion into even more US theaters, I spoke with visual effects supervisor Andrew Whitehurst about how they were able to tackle the challenge.
Bryan Bishop: Let’s talk about making Ava. Other than the restrictions with budget and time, what were the biggest challenges?
Andrew Whitehurst: I think there’s a sort of film historical precedent issue, which is we didn’t want Ava to look obviously like a robot from any other film that people would be used to seeing. The C-3POs, and Maria from Metropolis; that kind of thing. I made a rule which I applied to myself as much as anybody else who was on the team working on it, which was no one is allowed to look at robots. We did get a whole bunch of reference images of sculptures, by Constantin Brâncuși, who did these sort of very organic, yet almost mechanical, modernist sculptures. And things like Formula 1 car racing suspension. Anything where strength and weight was a big issue in the design of them. And we used that, plus human anatomy, as a kind of starting point when we were beginning to physically build her.
When Alex came to [visual effects company] Double Negative, he’d already been working with a comic book artist named Jock who’s done awesome work. I’ve always loved his stuff, so it was great seeing fresh pictures from him. And they’d come to us with some designs and some ideas, which we then worked up as we were discussing the practicalities of what could and couldn’t be built from the painted concepts. You get to a point with concept paintings when you have to say alright, this is as far as we can push a painting, and now we actually need to start building something tangible. It was [being built] in the computer, but it has form and solidity. Because with paintings, you can hide a lot in the brushwork. You can do a little brush stroke that suggests something, but it’s not discrete, and when you’re building something for real it has to be entirely discrete.
"A lot of times when you see robots... you can cheat like hell."
Overall, I think the approach to Ava was that iterative design approach, but she had to be mechanically plausible as well. A lot of times when you see robots, particularly in movies and you’re doing it with CG, you can cheat like hell. You can have things that will intersect ... you can get away with murder. I was absolutely determined that we wouldn’t do that. If it’s a film that has a slightly more fantastic aspect to it, it’s fine. But Ava as a character and as an entity on screen had to be entirely plausible to me to work. So I wanted to make sure when we were designing her that everything was functioning, plausible, and everything was in the right place.
The validation of that was that a large chunk of Ava was 3D printed for a scene that takes place in the laboratory where she’s constructed, and when they 3D printed it, it all slotted together, and it all articulated properly. That’s something that most movie viewers, they don’t consciously notice it, but I think if we had cheated something about it wouldn’t have felt quite right. And I think that would have undermined what we were setting out to achieve.
It's a difficult balance to strike, because the audience has to understand she’s a robot, but for the movie to work that idea then needs to fall away, in the same way it does for Caleb.
Exactly. We also put a lot of effort into things like the muscles contracting properly, and the various pipes and wiring having just a tiny amount of jiggle. And it’s something that you really do not notice. But I remember when we were looking at shots, for whatever reason when we put a shot through to render overnight that secondary animation hadn’t rendered properly, so it was missing. And everything suddenly felt very stiff. And you kick the shot off again, this time with that animation integrated into it, and it works again. It’s not something that you can necessarily put your finger on as being wrong, but if it’s missing then you suddenly feel that something’s strange.
So much of visual effects is psychology more than anything else. You’re looking at what are the cues that we as humans latch onto when we look at something that makes it feel right. And it’s often those really kind of subtle things that help ultimately push it that last few percent to really sell the illusion.
How many VFX shots were in the movie overall?
Overall I think it was about 800, 350 or so of which were robot shots. ... But I think the thing that is notable — I mean, that's not a particularly high shot count by modern visual effects standards, but this film has many fewer shots in it. The average shot is probably about eight seconds long, which when you consider most action films, you're talking about shots that are often less than a second. The number of shots may be lower but the actual number of frames is pretty high just because each of those shots plays out for quite a while.
I’m curious about how you shot the Ava scenes. It’s my understanding that you didn’t go the greenscreen route, or have Alicia Vikander’s face covered in crazy motion capture dots or anything.
No. It was the first film I’ve ever done where we did not put a single greenscreen up. The key thing about the film fundamentally is that it is a series of conversations. And for those conversations to work and be engaging, you have to have two actors that are talking to each other, and then the audience can get involved in that conversation. And if that doesn’t work, it doesn’t matter what the visual effects do. You’re screwed, because there is no drama there. So we knew we had to let Alex and Rob [Hardy, the cinematographer] shoot it as they would shoot any dialogue scene with two actors sitting there, and that’s what was on set.
Alicia was wearing a costume that was mostly made of that grey mesh you see in her shoulders and upper torso area — she had a kind of catsuit that was made out of that kind of material. But the scenes had to be shot as if they were a regular dialogue scene, mostly because you needed to have that interaction between the actors. So you couldn’t use motion capture, for example. But the other reason was we had a six-week shoot, so we were doing between 15 and 25 set-ups a day which is a hell of a lot to shoot for a movie. That’s really motoring. And if you’re [constantly] saying, "Oh we need to put greenscreens up," or, "Oh we need to do a motion control or motion capture shoot," there just isn’t the time. You have to be fast, and you have to be very flexible.
So the approach that we came up with was "shoot the scene however you want, take as long as you want." When [Garland was] finished doing a set-up, we asked the actors to step out, and then we shot a clean pass where was asked the camera operator to mimic as closely as she could the movement that she’d done when the actors were actually there, and that gave us a clean plate. And then when we got into post-production we could track both versions of the shot — the one where she’s in there, and the one where she isn’t — and then paint her out and restore the background behind her, which was a very painstaking process. And then we could begin to body-track her performance, so we could capture as closely as we possibly could exactly what she was doing on set. Then we used that animation data to drive our robot. So the physical movement of her is all Alicia, and the face, the hands, the feet are photographic in 99 percent of the shots.
It meant a lot of work for us in terms of painting clean plates. But I think what we gained out of it was vastly more important, which is you got that sense of the drama played out, and we were able to shoot at speed which was absolutely crucial.
"The face, the hands, the feet are photographic in 99 percent of the shots."
During a lot of the movie Ava is very still and static, which I imagine was easier for your team versus the moments where she’s more active?
No, quite the opposite! The hardest stuff to track is when someone is not moving. Because nobody is not moving, they’re moving but really, really subtlely. The easiest thing to track is when she’s running down a corridor or something like that. That’s much easier to do. Probably the single hardest shot — there’s a shot in the film where she’s holding out a drawing for Caleb to look at. It’s a tracking shot, and it’s 1,600 frames long. She’s literally just sitting there holding a drawing out and talking. If you just held your hand out and looked at it against something that isn’t moving, you’ll be aware of the tiny, subtle movements you’ll be doing even if you’re trying to hold your hand still. We were then trying to exactly copy that, and it was brutally hard. The hardest job on the entire show, I think, was body tracking Alicia, and particularly in those shots where she wasn’t moving much.
You’ve worked on some very big movies in the past. Was the idea of working on something smaller and more intimate like Ex Machina particularly intriguing?
Yeah, absolutely. I'd just worked on Skyfall before I started doing Ex Machina, which was obviously a considerably larger enterprise. There's pluses and minuses to working on bigger and smaller shows. I don't think you can necessarily say that one's better than the other, but they have different qualities. Certainly one of the big appeals of working on something like Ex Machina was that it was a small team of people, all of whom knew that we had to work together very closely, because it's the only way that you can make something work on a low budget, is if everybody agrees what the end goal needs to be and then everybody goes and works like hell to go and achieve that. I think with a small show, with a fewer number of people, you're a bit lighter on your feet, so you can make bigger decisions quicker, and you can change things and react to things a little bit faster.
It was a short shoot. We only shot for six weeks, I think, whereas for a Bond film it's a six-, seven-month shoot. So everything had an energy to it and a pace to it, which was a really lovely way of working. And the other plus was that I genuinely could not have asked for a better group of people to work with. Both internally at DNA [Films, the production company] and Alex, and Rob Hardy the DP, and Sammy [Sheldon] who did the costumes, and Mark [Day] who edited it. It's just a really awesome bunch of people.
Ex Machina is now playing. Images courtesy of A24.