Behind the design and the tools utilized by the studio on the film.
If there’s one trend stirring up the feature animated film industry right now, it’s stylization. DreamWorks Animation has certainly always brought a level of stylization into their films–think Madagascar and the much more recent The Bad Guys.
Now, with Joel Crawford’s Puss in Boots: The Last Wish, the studio has combined some of the stunning photoreal work they’ve accomplished in animation with more illustrative qualities.
At the VIEW Conference in Italy, befores & afters caught up with The Last Wish visual effects supervisor Mark Edwards to break down these particular stylized approaches, and how they were intertwined in the existing pipeline for animation development.
You’ll also learn the amazing term: ‘crap encapsulating objects’ and how it’s now one of DreamWorks’ go-to tools.
b&a: I’m one of those people who loves the ‘art of’ books that come out for animated films, and I have to say, until recently, the concept art in the books sometimes look nothing like the film. It often looks more stylized. Obviously they are realized for inspiration and mood and color scripts, but I always wanted the film to look like the art. Now it seems that’s happening more, which I love. I feel like this new Puss in Boots film jumped on that stylized look a lot more.
Mark Edwards: I think so. And it’s an interesting challenge anyway, because even the art team is a bunch of different artists and every artist has their own style. As much as you say, ‘Hey, here’s the look of the film,’ some of them were more inspirational than final, and some of them are really close to it, because you could find the elements to say, ‘Yes, this totally fits with our look. Let’s try and match that exactly.’
b&a: On some other films, I feel like I may have seen that some of the background paintings are ‘painterly’–partly because they are paintings!–and then there’d be the CG characters over the top. On this film, I really felt like there was much integration between both background and foreground painterly looks. How hard was that to do?
Mark Edwards: It was a challenge. We didn’t want to be so strict on layout that we couldn’t do dynamic moves, for instance. I mean, you can do a painted watercolor film. There’s that great The Old Man and the Sea animated short by Aleksandr Petrov which was painted on glass for each frame. But it’s noisy, that’s what you get. There is not a lot of temporal coherence. It’s an interesting style, but we need to stay with these characters, and so we can’t have stuff shifting on the edges and all this kind of noise all the time.
We wanted it to feel like these were painted assets, but we didn’t want to feel like it was just a Photoshop filter applied to it. Then with all those rules in place, there were all these challenges of, how do we deal with cameras being anywhere? That was where making sure a lot of our tools were more procedural, than it was layering that effect. That meant every shot we could actually have some control downstream to go, ‘Okay. Yeah, the CEO on this table, it’s just too big for this camera. Let’s just scale it down in lighting.’
b&a: You just said the word CEO. What does CEO mean at DreamWorks Animation?
Mark Edwards: The name is based off of ‘CIA’, which is an old system that we’ve used forever, and it stands for ‘crap in the air’. So for all of our particulates and things that are just floating, that’s a system called CIA. Our head of look, Baptiste Van Opstal, came up with ‘CEO’, which is ‘crap encapsulating objects.’ And it is a similar particle-based system that has some of the same toolset.
b&a: I love it. But what actually is it?
Mark Edwards: Well, at its core, it’s basically a particle set and a shader hook. So in fact, it’s mostly what’s in the particles. We store all of this data so at render time we know exactly what all the material properties are, what it should be looking up in terms of color. A lot of those controls can be adjusted at render time. The toolset itself is all in Houdini to actually generate that point cloud.
b&a: In terms of props and backgrounds, what was your methodology in modeling things, but then still giving them a stylized look?
Mark Edwards: That was interesting, because we started at an even more stylized place. I think we’ve always thought about that with our design of assets. So if you look at the Madagascar world, even their character design, everything, you have certain shapes and a design aesthetic, and we always have that wonkiness–nothing is ‘straight’.
We started there for this film, but as we did more and more tests, and had the rest of the style come together, we realized we didn’t go far enough. And I think, once you aesthetically break away from things needing to feel absolutely physical, there is huge freedom. For example, we had a lantern model where we basically have breaks in the lines, that was something I think modeling would never try. And we said, ‘You know what? It’s okay, because everything fits around it.’ It was really fun to push that far and build up from there.
b&a: In your presentation you mentioned something about a ‘stamp map’–how did that work?
Mark Edwards: That was actually prototyped in look of picture test. Baptiste had done certain prototypes and we were working together daily on what things would work and what wouldn’t. The stamp map was a way of building a point cloud around an object at whatever density you want. It has all the same kind of scale orientation controls as a regular projection. Then you can project an image or a set of images onto the surface.
It’s temporally coherent, because it’s in reference space. For certain things like normals, we did the work to make sure it could form, so we could put it on characters, which is critical. It effectively allowed us to layer in another set of stylization such as brush strokes or really abstract maps for reflections, or whatever we needed, into any kind of shader component.
We really liked that idea, because we had looked at, how much do we want to build into the material? And we did do material work, some toon map setups and things for some stylization. But when you build it all into your shading model, you’re a little bit locked to that. But here, we could say, ‘Hey, let’s plug it into just the refraction.’ We might want reflections to be normal, but then everything refracted is going to be completely warped.
All this was procedural, so if the stamps were too big or too prominent, we had all the controls downstream to dial it. And that happened all the time. His guitar had some really pushed normal stamp maps, and you would get all that breakup. And that was fine at a distance, but then we just dialed it back when we used close-ups.
b&a: Let’s jump to the characters, and the main character, in particular. So, DreamWorks obviously has a CG model of this character, and in the past there’s been a way that it’s modeled, rigged, textured, lit, rendered, comped, etc. If you are generally going for a more stylized look, which you did, what was different here with Puss?
Mark Edwards: We looked at things like facial features, eye size and silhouette of the fur. The shoulders were another thing–really trying to make sure it felt like a cat. The old model was great, and it was appealing, but it had some things that weren’t quite as streamlined as we thought.
Then the big thing was the costume, and the costume elements. Because we can do so much more with our character FX team now, it was a no-brainer to say, ‘He should have a cape. Capes are cool, and we can use that silhouette everywhere.’ It was a real heroic thing. And because of the stylization, we made sure that it was a little looser in areas and abstract so that we could scale it wherever we wanted.
b&a: Is it possible in the past your CFX team endeavored to make it as realistic as possible, and now they use the same tools to go for a different kind of look?
Mark Edwards: Yes, and that’s something other films have done a lot as well. I mean, the Boss Baby franchise is very silhouette based. And so, all of the clothing for CFX there, you’d start with more of a physical sim, and go from there. I mean, you could do drawings to get that silhouette edge, and all of that right on top. We’re really always working hand in hand with the animators. They’ll do draw-overs and ideas, and CFX tries to kind of get that same look.
b&a: Then there’s fur, where again, for so many years it feels like we’ve been aiming for photorealistic fur. How do you add in something that brings in that more styllized look to fur?
Mark Edwards: That was one of the biggest challenges for us for sure. Partially because some of the other techniques that we were developing were great for hard surfaces, but didn’t make any sense on fur. There was even a point where we thought, ‘Are we going to do fur at all? Should it be some other geometry?’ In video games, say, because of rendering restrictions, they do a lot of painted cards with transparency. That’s a very certain look and a style.
However, here it wasn’t about losing detail and richness, it was about controlling it. I pushed to make sure we actually had a groom–we still needed the fur to ‘stand on end’ and things like that when Puss is in danger. So we said, ‘Let’s add stylized elements to it.’ We were also able to paint in some very particular hair pieces on character like mustaches and stubble for beards, which worked, too.
b&a: I think the guard hairs also felt more illustrative this time around.
Mark Edwards: Yes, and those guard hairs are a natural thing in animals. It’s kind of this extra set of hair. The original Puss in Boots had those, too. We thickened them up. We gave them a transparency breakup. And we could completely art direct just where they were, both in the silhouette and just on his face. They became like accent lines. We had considered doing them as ‘line work’, like we had done for The Bad Guys. We did a little bit of that testing, but really with the soft fur look, it just didn’t make sense, and none of the rest of the world needed it or had it.
b&a: What about the animation of the characters? I definitely noticed it was ‘stepped’ where it seemed to work, but then more smooth elsewhere. Where did you come to in terms of doing it on 1s or 2s?
Mark Edwards: Most of the film is on ones, so it’s pretty smooth. But whenever we get into a heightened action sequence, we transfer into what feels best for those shots with the animation. And so, it really was just any of these big epic action sequences going to 2s or 4s or whatever we needed. We called it stepped animation, because it wasn’t just going to 2s, it was whatever worked for the shot, which that was a fun nightmare of a challenge. It’s all about making sure that our animation and CFX pipeline and downstream rendering had the right data to handle that.
b&a: There was a really great thing you showed with the wolf fight and the smooshing and multi-limbs in betweens.
Mark Edwards: It’s a technique that’s super common in 2D, and we probably only used it in 10 shots. But when we needed it, we had it. Other films have used it a lot more. But it was great, because it was so integrated that animators could just choose to use it when they needed, which is always the best if they just have a tool, and they can say, ‘Hey, I’ll just plug it in and use it.’
b&a: Rim lighting, too, was different here, how?
Mark Edwards: I was really excited to work early on with Nate Wragg, the production designer. We both loved the graphic feeling you can get from that rim light. And because we’re physical, but not totally realistic, it again, gave us that freedom to just use it because it looked good. We would use that all over the place to just pop characters and get that graphic edge. The lighting rigs were really flexible.
b&a: You showed in your presentation where there were rendered paint strokes for some of the atmosphere or the background, which then gave you an extra textural look. Had you done it before, or was that an experiment here?
Mark Edwards: That was an experiment that worked. We had some other early tests with the giant in the village, and were playing around with some other concepts of light beams with his eyes, and eventually that was too distracting. But they did the same thing where they tried just to get, again, a little texture, because if you’re doing a volumetric renderer, it’s going to smooth step through it unless you add noise, but we didn’t want procedural noise. So we really wanted that kind of same stamp map, brushstroke feeling. It was essentially a 3D version of paint strokes that was rendered as a separate pass and then composited in.
b&a: I am convinced audiences are obsessed with the more 2D looking dust movements or explosions, but which are still FX sims, and they look amazing. And they’re in a few films recently. But you had a very particular approach here as well, right?
Mark Edwards: Yes, we knew that we needed some physicality of certain elements, and there’s some nice things that you do get from that. But then we wanted to make sure that we had the graphic feeling fit the rest of the world. And so, every single FX mmoment was a challenge of, how do we break that down and figure out which elements make sense. And so, with dust or explosions, in particular, we had the volumetric renders but also surface renders–we could call it our ‘scalloping’.
We always had these passes where we could get more of a 2D graphic hard edge, and we would layer that in. There’s always a little bit of tooth to it, and they would layer all that stuff together.
That approach was not completely what worked with some of the fire, though. So then we had to develop some of that a little more. A lot of things would be augmented just with hand drawn effects. And Michael Losure, one of our effects sups, had developed a tool that let the artists effectively draw on Houdini these lines that then they could use as 3D elements. It was like a 2D to 3D element.
You could do things like hand it off to lighting for reflections, but then it was controlled just as a 2D drawing. The big end battle with the wolf had a lot of late requests from the directors to get some dust and sliding and all these impacts, which we could do all-2D.
