When I was first contacted by Director Adam Hashimi and Producer Christian Zethner, both from Bacon CPH, I wanted to convince them to do this in CGI. There were some initial ideas about doing a football player in an animal suit on a green screen, but I wanted the squirrel to look real and I wanted it to be a dynamic mix of squirrel and soccer player moves; from the very fast move all down to the squirrel breathing. And this could of course only be achieved with CGI. On top of this, I also needed to convince them that we could pull it off even though we hadn't done something like it before. I wrote a treatment on how we would approach it, explaining our ideas of how to solve it, about detailed animation, about extensive use of squirrel and soccer references and about how we would solve the zooming and moving camera. But even after this, I can't thank Bacon and the AD agency enough for trusting us from beginning to end, even though they couldn't judge it before very close to final deadline. Their patience and trust is an essential part of the outcome and quality of this spot.

Fur, Shading, Rendering by Peter Szewczyk and Peter Szewczyk

As with any photo-real work, the first step was to get plenty of reference material. Often the only reference one needs is only clear after the work has had an initial first pass. For instance, when collecting references, I did not think to look for shots specifically showing off the furs specular, until I had come to that part of the process. In short, the more reference the better. Also I should note that in this case that since many of the aspects of the squirrel change from species to species we had to confirm some of the reference with the client.The next step was to get something that looked good using only texture maps and no fur. The texture maps would drive many of the fur attributes, so seams had to be clean and in general the assets should work quite well (as, say, for a far-away shot) with only texture maps. It is very important to mention that whenever possible the construction of the texture maps should come directly from reference. Something drawn from scratch almost never looks as good as something cut out from reference.

Once a solid texture map was in place I went ahead and created fur for the body and the tail. These two pieces were separate since, unlike the fur on the rest of the body, the tail is very long and fluffy. Once the fur was in place I started the daunting task of creating custom maps for just about every attribute.The color of the fur was driven by the texture maps but a precise level of hue and saturation variation was added on top. Direction was obviously painted as well, with particular care being taken to eliminate UV seams. Since clumping doesn't work in Renderman for Maya, a lot of experimentation was done to the length variation, and fur density to get agreeable results. One of the turning points for the asset's look was when transparency was added to the hair tips. This way we could control how plush or rough we wanted the fur to be.

In terms of lighting I tried first an array of deep shadows, but this proved expensive and unless blur was turned up high it made the fur look dirty. I settled on a solution of global illumination; granted, only the geometry of the asset would occlude (and not the individual hairs themselves) it looked good enough for our purposes. There was some experimentation on what the global illumination map would be and I settled on one that I painted myself incorporating some green bounce for the grass. I also had a few extra lights acting as rims on the body and the nut. And some reflection-only objects were sometimes used to get better specular in the eyes.Every few days a turntable would be created against the background that we would ultimately be using. It is also important to have at least one pose that matches an onset asset. At one point the client was concerned that we were no where near the photo reference until we showed him a side by side comparison. When animation was applied to the asset I always chose to work on the frame that would show the asset at it's worst. There’s no point in deceiving yourself by picking an easy frame!

Finally, in terms of rendering, there was some experimentation with the number of hairs we could use. Ultimately, even on the asset's close-ups, there were never more than a million hairs. To get the motion blur looking good pixel samples needed to be relatively high but I found that since it was mostly fur that was being rendered, the shading sample could stay at 1. Only for troublesome areas were shading samples turned down to 0.1. Some experimenting was done to find the optimal filtering/sample for the global illumination. Frame renders were always somewhere around 30 minutes.The passes we rendered included a rim light pass, a primary pass, a bounce pass and an ID pass. Inside of all those passes were secondary outputs that included solid color, specular and occlusion.

Once we were able to nail exactly what of the asset's references the client was looking for we moved through the project pretty swiftly. In my opinion Renderman for Maya was essential on the project and no other software could have handled fur and motion blur as well.

Production : BaconCPH
Agency : &Co
Producer : Christan Zethner
Director : Adam Hashimi
Creatives : Thomas Hoffmann, Martin Storgaard

Post Production : Ducking A/S
VFX Supervisor : Jonas Drehn
Model / Rigging : Grant Walker
Animation : Søren Cornelius Larsen, Jakob Welner
Fur / Lighting : Peter Szewczyk
Lead Compositing : Per Mørk-Jensen
Compositing : Emma Lide

Duckling Studio : www.duckling.dk