The CG side of the animated ‘Mulan’

Looking back on the mix of 2D and CG in the 1998 film.

Walt Disney Pictures recently released its live-action Mulan, an adaptation of the 1998 animated feature. Back around the time of the animated film’s production, Disney had been slowly integrating more and more CG/3D techniques into its traditional hand-drawn approach.

Those CG/3D—and two-and-a-half-D—techniques would ultimately become crucial for helping to realize several of Mulan’s most challenging sequences, including a crowd of Hun horse riders, hordes of people engulfing Mulan’s approach to the Emperor’s palace, and in many other areas of the film (possibly more than audiences were and are aware of).

One of the artists at Walt Disney Animation Studios who was part of the digital production team responsible for this work was artistic supervisor Eric Guaglione, now an animation supervisor at Framestore. With the release of the new film, befores & afters decided to go retro with Guaglione to discuss the integration of CG and hand-drawn elements in 1998’s Mulan.

The state of play with CG and hand-drawn animation at Disney back then: Disney’s prior films had used CG, but in many cases I think the application was, ‘Oh, this is going to be incredibly difficult or laborious to do, so can we get a computer to do it?’ That was the mindset—it’s a computer doing it, not an animator. Which of course was not really true. If you look at the ballroom sequence in Beauty and the Beast or the wildebeest stampeded in The Lion King, there was much more to it in terms of creative input.

By the time Mulan was in production, I think there was a greater acceptance for how CG could not only just help achieve work really difficult and laborious to do, but also changing the thinking to, how do we actually use CG in a way that ultimately has an overall effect on the movie itself in terms of the look and feel?

I had done traditional animation before doing CG animation—I was a well aware of considerations like, when do you animate something on twos versus when you do you animate on ones? It’s not always about it being cheaper to draw every other frame than it is to draw every frame. Instead, there’s a certain feel that you get from that as well.

So, one of the things that we immediately said for Mulan was, we’re going to animate things in CG in the same way that we would animate them with a pencil, for instance, we will animate things on twos if they should be on twos, or ones if they need to be on ones, and we’ll jump between them if need be. That actually made it stylistically more fitting with those elements that were drawn around it. At the same time, it was also beneficial for the 2D animators that had to draw in conjunction with our CG animation, because we were applying the same methodology—the same thought process that a 2D animator would apply.

If you look now back at something like the ballroom sequence in Beauty and the Beast, it’s beautiful, but it’s very obviously CG. By the time we got to Mulan, we did think about, how do we better integrate everything that we do? So it was about approaching everything aesthetic, like I said, on ones or twos just like the 2D animation, but also like, how we actually do our tone mattes so that the tones look more hand-drawn? It was a very conscious effort on our part to say, let’s try to make everything look drawn that you really cannot tell what’s CG.

Crafting crowds: The crowd software we made was 100% proprietary, it was called Attila. We started with the base work that had been done on the wildebeest sequence in The Lion King, and then we developed on top of that. One of the things we knew the Hun charge would require that the wildebeest sequence did not is that we would need to do a lot of mix and match—putting this type of horse with that type of rider wearing this kind of hat, but not that kind of jacket, but this kind of weapon and so forth and so forth.

We needed to have something that would allow not only the ability to mix things up but also edit those things that we mixed up. Of course, being a Disney film they were very particular about everything. The directors might say, ‘See that guy over there, I want to change his hat to this one over here…’. So we were thinking, our approach will need to be incredibly flexible and powerful to be able to change details easily.

The flocking aspect of the system, ie. moving something along paths, wasn’t something terribly new. We had to of course make it work over the terrain but the whole idea of being able to really edit to the most minute detail on any single character was something that was new.

One part of the crowds tool was editability. Then, at one point we met with the directors and they said, ‘We have a problem.’ The problem was that Mulan was supposed to be going to the Emperor’s palace and all the people of China are watching her there. The problem is Disney couldn’t afford to animate all these individual shots with bespoke hand-drawn characters. So, how were we going to populate these sequences with the number of people that we wanted to do? It was mind-blowing how difficult that was going to be, if done the traditional way. The directors said, ‘Do you have any ideas?’

We looked at Disney’s previous films and the one that had been released before Mulan was The Hunchback of Notre Dame, which had crowd work done with all-CG characters. They had them mulling about, or they were cheering. That was the right idea, of course, to use these generic people that you could place into these sequences. But at the time that this challenge was brought up, we were already a little too deep into the Hun charge and other things, so we just didn’t have the resources to build all the characters and do all the various cycles.

Cards as the answer: So I said, is there a way that we can use our crowd system for the Hun charge using cards instead of CG elements? Plus, I thought it would ultimately look a little bit better if they were hand-drawn characters, because it would just fit better within the sequences, as opposed to having them as all-CG characters.

I also happened to know there was a little bit of a thing going on with a story change which meant some animators were idle, and could handle the amount of drawn animation required using this card approach. We took a couple of cycles of hand-drawn characters, put them on the cards and tried out various aspects like how far could you take a camera and go up over them and see that all of a sudden they’re flat? Or, how many people did you need to have to make a crowd look fully unique?

We soon realized, too, that we were still going to have to have animators controlling what these crowd scenes looked like, so then we had to work out, how do we make it as easy as possible for the animators to be able to say, ‘Hey, I want a bunch of people here, but not so many people here’.

I asked our programmers, is there was a way that we could just create a simple paint tool and just paint that information. Like, ‘I want some over here and I don’t want as many over here.’ And we’d have a transparency for less or opacity for more. So the artist could just take the Layout image and paint over it and use that as a projection map of where to populate it with people, run it through the software, and all of a sudden you have your crowd.

Going two-and-a-half-D: We found right away that we had a very intuitive artist-friendly way of populating shots. It was amazing how high you can actually go on a card before you realize that it’s not 3D, and it doesn’t really destroy the illusion. It was something like 60 degrees; I was really surprised.

And then in film like Mulan, where everything else is hand-drawn, we were really dealing with a lot of ‘flat’ art. Chinese art is by its nature flat, it is all about planes. So it actually made it feel very right to make everything feel as graphic as possible, to have some degree of flatness to it.

Things did get trickier though, when it was like, ‘Oh, well, we need to have 10,000 people out here, and then at frame 36 they all need to start jumping up and clapping.’ At that point we had to figure out ways to do transitions in the movements and the behaviors that they were performing.

But it all happened very quickly. When the creative problem was brought to us, I think we did R&D for three weeks or something, and proved out that it was going to work. You go from rapid prototype to scratching your head and saying, ‘How are we going to do this for production? How are we going to actually make that work?’ But we did it fairly quickly because of the work already invested in Attila.

Once we started seeing the benefit of how much you could create animated artwork on cards, we started doing it with all kinds of element. For instance, when they’re shooting flaming arrows and there are these beautiful curly smoke trails, those are actually all CG, using 2D artwork on top of these flat tubes.

Moving the camera—Faux Plane: Just prior to joining Disney, I was working on Star Trek. One of the projects I was doing was the opening title sequence for Star Trek Voyager at Santa Barbara Studios. There was a moment in which there’s this icy landscape in the foreground and the ship flies overhead. When the client looked at it, they said, ‘Oh, that’s really cool. Is there a way that you can make that planet roll?’ But it was like, oh crap, it’s a matte painting, how do you make a painting roll? It’s flat. We came up with the idea of effectively using two-and-a-half-D geometry that we projected the painting onto. Of course now it seems like a really obvious thing to do, but back then it wasn’t something that people had thought of.

So when I joined Disney, I was thinking of the same thing, how do we do that, but not quite the same way? We were going for something more graphic, we were trying to create something that had partial depth to it that would make it feel like a classic multi-plane shot that you see in the golden era of Disney animated films—achieved by introducing 2D planes with added dimension to each.

It was achieved through writing proprietary software. We started with the opening title sequence where the camera’s pulling back and you see the Great Wall of China. We did an early test of that sequence to just prove out how that could look. Everyone went crazy for it; it was just so perfect for this Chinese style of art. It’s something that feels flat and yet it doesn’t, it was something much richer than we could ever do with traditional multi-planes because it actually had depth to it.

Then we started using this technique in other sequences. For example, when Mulan first enters the Emperor’s palace. There were a couple of shots where she goes in through the main gate and the camera’s swooping through. The entire environment was done using CG FauxPlane, all the crowds are done through CG. The fireworks are CG. The only thing that’s hand-drawn was Mulan.

The other thing to note is that when I was thinking about Faux Plane, I was talking to the team who were working on Tarzan. They made a tool called Deep Canvas. That was about painting brush strokes into 3D geometric space, with those strokes making the geo feel more painterly, which was a really nice effect.

We were looking for something that had more of that Chinese style and a little bit more flatness to it, so we went in the direction of doing something that had more of a 2D look. Then after Tarzan, some films did use the Faux Plane technique and it likely continued to evolve and eventually become full CG. That two-and-a-half-D methodology still exists today, though, including at Disney and Pixar. They still do dimensional matte paintings, for example. We were doing it on Mulan for the foreground through to the background, and now it tends to be used as something in the deep background.

Get bonus VFX material by becoming a befores & afters Patreon member