Yeah! And that’s one thing I did want to ask you about: the role of special effects in moviemaking. Because like you said, the idea is that most people shouldn’t even be noticing it.
Exactly! There’s a book about this called ‘’The Invisible Art’’, which is about Matte Painting but it’s still relevant. With any visual effects, even if it’s just a talking creature, the secret is to support the story. Because we’re not an art exhibition! The audience don’t come to the cinema to see just one beautiful visual effects shot. They come to the cinema to be entertained while eating popcorn and drinking a soft drink.
Our objective is to support that by enhancing the storytelling. If someone comes out of the movie and asks : ‘’So what did you guys do?’’ that’s always the best reaction for me. Because it doesn’t matter, we did our jobs and got you to focus on what the director had to say. I think that’s the secret to good visual effects.
How would you compare doing something like a Spaceship effect to a character or a digi-double? How differently would you approach those things?
Originally, Framestore was very much a niche, go-to house for talking creatures. Because we were really good at the animation side of things and at nailing the look of those characters. I suppose Gravity was our first show where we only had environment work and with physically active rendering, that’s become a lot more straightforward.
Getting a digi-double right is all about matching to the shots of the real performer. Again, you want to be in a spot where your audience doesn’t notice the difference. In the case of someone like Spider-Man, a human obviously can’t do some of those acrobatics, but your audience believes in the character, so our job is to be convincing and not break that immersion. It’s been one of our biggest challenges for the longest time. For that reason, digi-doubles used to be really small and it’s only very recently that we’ve got a bit braver and started using them in medium close-up shots.
As for spaceships, they’re all about getting the level of detail right. If you do a vehicle or something with a hard surface as CGI, then that’ll be because it’s cheaper than doing it practically, especially if you might have to destroy it multiple times.
So to answer the question, they’re very different beasts, but they’re equally challenging as well.
On that note, what would you say was the single biggest challenge with Captain Marvel? Is there an effect that you’re particularly proud of because of how tricky it was to pin down?
Probably the environment. There was an aspect of the art direction that we needed to get right, which was that everything was supposed to be covered in a thick layer of alien fog. Finding out how to make that look different, rather than just like dry ice or smoke, was really hard from a creative standpoint.
Meanwhile, there were also challenges associated with the quarry. It was quite big originally, but when they put the rough cut together they realised it didn’t quite have the scale they wanted. So we were tasked with splitting the set into about 4 pieces- those being the entrance, two middle sections, and then the temple at the end- and making it all look spread out. We had to fill the gaps with fog and render it in a realistic way. This had to look convincing from all angles, be it a wide-shot or a medium shot.
On a technical level though, the biggest challenge was probably the shot of Starforce underwater, where they torpedo towards the camera. Firstly it’s a fully CG shot, so nothing there is practical and there’s no plate for Brie or anyone else. Secondly, we needed to slow down time. It’s not 100% a bullet time effect, but it almost is. We slow down as the camera travels around Brie and then they shoot off again in real time.
I think it lasts for about 500 frames and the difficulty there is to figure out how to simulate things like bubbles, hair movement and all of the particulars of underwater physics in an accurate way, both in slow motion and in real time. Doing one or the other is relatively easy, but switching between the two is quite hard because you have to change your simulation behaviour. Normally what you would do is render the whole thing in slow motion and then speed it up when you need to. But this was already 500 frames long and if you slow that down by 500%, then suddenly you’re working on a 250,000 frame shot.
Oh that sounds like such a headache. Especially since it’s completely CGI.
And you would throw away 90% of that work when you speed it up. So that wasn’t an option for us. It might sound trivial to most people but it was a huge technical feat.
And that’s something that no audience member is ever going to conceivably think about. But I guess that’s indicative of what you were talking about earlier: that notion that if you’ve done your job right, then no one will even notice.
I would never have considered the challenges associated with slowing down and speeding up a completely CG shot. In fact, I would have naively assumed it’d be easier. But, if anything, you’re saying that it’s more difficult.
Yeah! If you were doing it for real then you could just shoot the whole thing at 500 fps and retime it. But to do the same thing digitally would require you to generate all the bubbles, all the hair physics and that’s too much f data. In fact, we probably wouldn’t even have enough storage to do it.
I have a quote here from Chris Townsend (Captain Marvel’s overall VFX supervisor), where he says that, for this project, he wanted things to look ‘’cool, photographic and physically accurate’’. In that order. What did that mean for you guys? How did that factor into your work?
That was said to us when we were doing the battle before Brie gets captured. Chris said: ‘’First of all it has to look cool. Then it has to look photographic. Finally it has to be physically accurate. And in that order.’’
So the first part is just about making sure that the audience jump up from their seats and go like ‘’Yeah!’
Obviously these days it’s also important to be photographically seamless and in keeping with the rest of the footage. So basically – even if it’s a full CG shot – you still need to consider things like the camera characteristic, the lens variations and the colour behaviours. That way the audience still believe it’s a photographic image.
Then, if you can, you need to base it in some kind of physical reality. Clearly we don’t know how someone would react to superpowers or any of these things, but we still want to have logic. So if this guy gets hit over here, then he needs to land over there, you know? You need to figure out the trajectories so that it doesn’t look weird. Even if these things aren’t really possible, they should adhere to some semblance of physics .
That makes sense. It’s about how believable it looks, rather than air-tight scientific accuracy.
Yeah, it needs to look right to the audience. A few years ago you would always have these shots in visual effects movies where you’d think: ‘’Yeah, I just don’t buy it’’. But now I think visual effects have gotten so good that we’re starting to buy more and more. Even if it is something as far-fetched as the Hulk destroying a building. As long as the relative physics are still correct in their own world.
What Chris was saying in that quote is essentially what I said earlier. Your audience just need to believe it and those three factors are the key to achieving that.
Just to return to something you mentioned earlier, you said Framestore were responsible for the sequence in which Carol goes through the different colour variations on her suit. How do you pull of something like that?
Luckily, with the quality of the rendering available to us now it’s not that difficult. If you give me a good reference for a real material, like a costume or a suit, then we can recreate it fairly easily. In the past you would have done this with lots of roto and colour grading. These days however it’s not that tricky.
In the end we curved the gloves, her hair and her face and then we removed the entire body to replace it with a CG suit. The thing is, you don’t want your directors to notice something like that, so we tracked her body language in order for the CG suit to stick to her head and her arms. Once we had matched that part up, we could just change the colour.
Well it’s very seamless.
I think we’re confident enough these days to pull that off. I mean, there are actually shots in that sequence where she’s wearing her real costume, but the quality of the work we’re doing is so strong that you won’t be able to tell the difference.
Sometimes we test it on clients to make sure and we did that here. We sent Chris a shot where we had replaced her real black and green suit with an identical CG version just to see if he noticed. His reaction was: ‘’What are we looking at?’’, which is all I needed.
We do that sometimes. When we replace something with a CG version that looks exactly the same, we’ll send it back to the client. If the reaction is ‘’Why are you sending me this?’’ then we know it works. You can’t do it too often, but every now and then we try to sneak it in [Laughs].
Looking into the future, what kind of projects have you got in the pipeline?
We are doing quite a few shows across London, L.A and Montreal. For example, I have just joined The Aeronauts for Amazon. We’re working on Mulan, Avengers: End Game, Spider-Man: Far From Home, Artemis Fowl. We have a bunch of projects in the making.
You indirectly tackled this in the first question, but do you have any advice for someone who is hoping to break into the VFX industry?
I think understanding artistic fundamentals is key. For instance, when we started the office here we had to train a lot of people up and take some risks in terms of who we hired, but the quality we always searched for was a good artistic knowledge. You know, a traditional artist can always translate those skills over. It’s like I said, new software is just another brush.
So my first bit of advice would be: don’t ever underestimate the importance of knowing the basics! Even art history can be useful, because sometimes we get asked to [take inspiration] from Leonardo Da Vinci or whoever. To the younger generation, I would also stress that it can be hard to get into this industry. So I would encourage you to attend visual effects events or conventions, and get yourself out there. You don’t always have to strive to work to ILM or Framestore as your first employer, there are so many small visual effects houses you can try. Internships are good too.
Once you’re in, there will be a lot of hard work and I must admit it’s not a 9-5 office job. We’re all here because we’re passionate about what we do. If you’re not, then you will have a really hard time.
Many thanks to Christian Kaestner for taking the time for this interview.