Now available as part of the Monsters University Blu-ray, Saschka Unseld‘s The Blue Umbrella is undoubtedly set for repeat home viewing and some very intense scrutiny. Thankfully, not only is it one of the most memorable short films yet from Pixar, and definitely one of the most distinctive, it’s also one of the most aesthetically rich and sophisticated. This film bears repeat viewing well.
I spoke to Unseld in some detail about the film, what makes it work, and what the technical and creative challenges were. From what he told me, it became very clear that this film took the team at Pixar way outside of their comfort zone.
So, here’s what Unseld told me, illustrated along the way with a clip, an excerpt from his original pitch, and more images from the finished film.
It started when I took a photo of a broken umbrella on the side of the street. I just thought “There must be an interesting story there because it looks so sad.” It was from there that I built up what became the story of the film.
Half accidentally, I always seemed to have in the back of my head how the film was going to look though, even though I wasn’t necessarily aware of it. When I pitched the story of the film to John Lasseter I told the whole story verbally with no visual references at all.
When I was done with the pitch, I also showed a test film I had done for something else. A couple of years before I had the idea of doing a music video where the city sings the music. All I had done was go out on the street with my phone and filmed faces I saw in the city then loaded that up in my computer and morph-animated them so that they sing the song.
I said “So when I said the city comes to life and is smiling and happy, here is what I meant by that” and then I showed this test. I wasn’t thinking “This is exactly how it needs to look” or “It needs to look photoreal” or anything like this. All I had at that point was the story and I knew we would have to work to find the right way to show it visually.
That test starts off with five seconds of nothing happening, it’s just me filming the sidewalk. You’d think “Why is he showing us this clip?” and then, exactly the same as in the film, where the first character blinks, I had that same character come to life. And everybody reacted to this magic of something inanimate suddenly coming to life. It’s so fascinating, I think, because before that moment you think it’s real and something couldn’t come to life in a film that’s real.
After I had shown that test I started to think “What if we did the film like this, to keep this magic to it?” There was this confluence of a test I had done, just because I know how to do technical stuff and I enjoy it, and this story I felt strongly about. There was a perfect match between the look of the test and the look needed for the story.
The biggest technical challenge was that, outside of Pixar, in the world of visual FX, there is tons of stuff that looks real – almost every visual effect is supposed to look real – but in the confines of animated films, that’s normally not so. The whole production pipeline, how our system was set up and how everyone was used to working, just wasn’t targeted on this realistic visual look. Everybody here is normally working towards something stylised.
It was a big challenge for us to adjust how we make things, how we use the tools for this new kind of style. We had to get people to think differently. They were used to thinking a bit more cute, a bit more cartoony and that’s not the right direction. Even unconsciously people started to model things the way they were used to, so that the edges were a bit too round, and then the images didn’t feel real anymore.
A good example is the second character who comes to life, the blue mailbox. We gave photos of a real mailbox to a modeller to build it in the computer and we noticed that this model, somehow, didn’t look real but looked slightly cartoony. When we looked more into the detail it was exactly this problem. There is a certain physicality to real objects – the metal has a certain thickness and the machines can only bend it in a certain way because of the attributes of that metal, how easily it bends, when it would break, all of those things. If these details are not correct, the model you create does not look real.
It’s almost as if you imagine a human face with wrinkles just painted all over it with no rhyme or reason. The observer would be able to tell there’s something wrong. And that’s just like cracks in asphalt, for example. They’re not just random, there’s a reason for why they’ve appeared in a certain place. Then there’s grass in the cracks, and people would walk over the asphalt tons of times so it would have a certain kind of worn nature and look to it.
Or, another example, is that trash doesn’t just lie randomly all over the street. People have probably kicked it so it’s closer to the kerb, maybe; then certain things would get washed away, other things would stick to a certain place. All of these details make the difference between something that looks stylised and cartoony and something that seems like a real city that people live in.
I do think a lot of this comes down to the roundness or the sharpness of edges. CG always tends to round stuff slightly. A lot of getting this right came down to me telling the team “Look at the materiality of the object, try to look at the way it was manufactured.”
If something would be cut off in the manufacturing process we should see a really sharp edge that you could almost cut yourself on; if it was bent in a machine, the radius of that bend probably has something to do with the material we’re portraying. In CG everything is surface and nothing really has thickness but depending on the material you’re representing you should show a sharp bend or the opposite.
And after this, we had to think about how something ages. We thought there would be multiple layers of paint on some of these objects, and that if there are creases in the object, those would get filled in with paint. We needed to pay close attention to how things are treated in the physical world.
To make these city characters move, to blink and react, their surfaces had to deform. That’s not something that’s inherent to these surfaces and the materials we’re representing. Stone doesn’t really stretch or deform. Metal might, a little, maybe. The tricky thing was finding out how to stay true to the characteristics of the material and to not have stone suddenly feel like rubber. My solution was to approach this as though we were doing puppet animation. We actually ended up animating some of these things on twos or on threes to give them more of a jitteriness or harshness in their movement, something more cohesive with the qualities of the material.
The second thing we needed to do was remove motion blur. Most of the city characters have no motion blur in their movement. The blur would have made them look too soft and too smooth again. There’s motion blur from the movement of the camera, but the characters are treated as though they were animated by taking a picture, changing the model, then taking another picture. We really stuck to this puppet animation style to keep the material looking true even when moving.
The choice of lens was definitely a big part of the look. After I showed the test and everybody responded so well, the first thought was that it worked because everything looked real. You could tell that I had just been out there with my phone shooting something and it had this home video look that you might perceive as being more “real.”
One aspect of this is that the camera was never perfectly steady, that there was a slight shake, but I don’t think that was all of it and we certainly didn’t want to shake our virtual camera like crazy and do the fake home video look that way because that quickly gets distracting.
So what we did was try to use a lot of long lenses and place the camera somewhere that a person could really be. We never have a free-floating, in-the-sky camera – except for the dreamlike moment where the umbrella floats in the sky. For the rest of it, we’re down on the ground, the camera is always somewhere a person could stand but never too close to the umbrellas. We designed the shots as if the cameraman was really in the city and, as if by accident, they witness something going on.
We wouldn’t normally be standing next to a couple as they’re falling in love, we couldn’t be standing between them, so the way we shot that moment was with a camera on the other side of the street with a really long lens. You get the feeling of something happening over there, where we can see what is happening but it’s their own private space. We don’t invade their privacy by standing next to them with our camera.
We tried, as much as possible, to not cheat anything, to have the same restrictions you would have if you were in the city. You couldn’t just remove half of a building to get a clear line of sight. Even though, of course, in live action you often do that, you build a set and you remove pieces. But I wanted the audience to believe in the physical reality of my set, and the more I would cheat this, I thought, the more it would feel like a virtual set. Everything here is virtual and giving the audience believability of this world was essential to me.
I think the only way we cheated on this a little was in how crowded the amount of umbrellas would be directly in front of the camera. But in respect of buildings, or power poles or lampposts, we always kept them. Sometimes, this gave us nice surprises, with something in front of the camera being a nice thing.
The final decision to never show a human face came down to both a technical point and a creative point. The creative side to it was that I wanted to tell the story of the umbrellas and I liked the idea that the story of their two owners, who ultimately meet as well, was something that happened on the side. Especially during the main part of the film I didn’t want the audience to be distracted and thinking about what was happening with the humans underneath. That would break the trust of the audience, their belief in the umbrellas being alive, when you remind them that somebody is holding them, have them wonder what that person is doing.
Technically, that came down to a question of deciding where to invest the amount of work the crew could put in. Focusing more on the people and showing the faces of the owners would have cost an incredible amount of work to make it look right. The question is, do I want to spend half of the work of my team on that, or do I want them to use that time on things that are more important to the core story of the umbrellas?
The way we did rain was completely built for this film. There’s rain in lots of other Pixar films but typically it’s just in one sequence and to a certain extent, a background character. In The Blue Umbrella it’s really like the third character. The film is about the rain to a certain extent, it’s a love declaration to the rain
I feel like I went to every department and told then “You think you have to do this much, but I want you to do way more.” With the FX department I said “You know it’s going to rain but it’s not just going to rain, I want the rain to feel differently in the different sequences.” I wanted the rain when they’re falling in love to feel super smooth, with long exposure times, something really soft and nice. Then, in the car crash, I wanted it to be really harsh and uncomfortable and brutal.
There were different rains that the FX department provided depending on the sequence. Some of it is much longer with longer exposure times, with a super-soft haze, and other times you could almost see the full outline of a single raindrop, and there’s no motion blur at all.
It wasn’t just a job of adjusting the amount of motion blur. There’s such a massive amount of blur that a raindrop would have, that for the romantic moments we had drops that were stretched: the physical object was stretched before we motion blurred it more, and that’s what gave us the right look. We were then able to keep a certain amount of sparkliness in the rain even though we motion blurred it like crazy. The process became more of a painterly thing – what feels right for the scene versus what would be technically correct. It is a magical world that we dive into.
The rain had some of its own light sources, particularly in the areas where it is sharp. Then, the rain in the background has extra filters on it to make it more sparkly. We partly achieved this some of by moving an invisible object around in the front of the shot, something so out of focus that we can’t see it but the lights behind pulsate and feel like the city is alive in the background.
We had thought initially that maybe we would shoot the film in live action, like the test film, but the amount of control we wanted over every element was something we could get much more easily in CG. The rain is a good example of this. To have the rain look specifically the way we wanted it to required having the controls to change the specularity of an object in a way you can’t do in live action at all.
But there were lots of reasons to not make the film in live action. You would need to find the perfect street corner with exactly the faces you need, you’d need to shut down a big part of a city, you’d need hundreds of people with umbrellas, you need hundreds of people in cars and then you need to make it rain on the whole set. And then you’d need to track the faces of the umbrellas onto the live action umbrellas, which will need to be moving in sync for the performance you want. It would have been a massive nightmare. You wouldn’t be able to control the details to the level that we wanted.
The most important thing to realise about real umbrellas is that they don’t have faces. All of the characters in our city have faces naturally. We saw characters on the streets of San Francisco or New York that, when we put them in our film, it felt like we’d just forced a face in there. You can’t push a face forcefully into an object, it would not look real anymore. Any way that the umbrellas might naturally have had a faces in them – through patterns or the folds – just looked disturbing, and especially if we animated them.
I didn’t want to break the reality of the umbrellas themselves so the right thing to do, I think, was to go in the complete opposite direction, as far away as you could get from photoreal to as stylised as you could possibly get. We made sure there was still just enough detail in the faces so that we could get subtle emotions and animations. We started with two dots and a line and worked out how much we needed to add to this as detail for the faces to express the more subtle emotions through our animation.
Thanks again to Sashka for taking the time to talk to me. You can check out his Tumblr, Rainy City Tales, for more production art and designs from the development of The Blue Umbrella.
If you want to read his pitch in full, It’s available from Disney’s “For Your Consideration” website.