Issue link: https://digital.copcomm.com/i/1547068
we were sitting in the editing room aer the shoot and David and the editor were like, "Ah, it's a shame we didn't get the dog to do this." Then I could go to Framestore and say, "I know it's very, very challenging because it's not an action shot, but we're trying to do an emoting shot of the dog. Can we try and do that as a CG shot?" Those were probably the most challenging shots that we had on the show. Because there's no doubt Framestore can create a photoreal dog that completely looks like the real thing. But getting that CG dog to perform an emote exactly like Uber, our real dog, was very, very challenging. Really, the biggest challenge on the show was to get the emotion of the CG dog to match the emotion of the real dog. Because there's a handful of sequences where we're cutting back to back from, let's say a medium shot of the dog—and it's a real one— to an extreme close-up of a dog that is completely digital. That's the most chal- lenging thing you can do. There's no place to hide, really. It's very bru- tal, but I think they did a fantastic job. Tell us about the process of creating the full CG dog asset. How did you bring as much realism as possible to the facial features, especially the eyes? François Dumoulin: Yes, the eyes are always the most tricky thing. Towards the end of principal photography, I organized a day of shooting in front of a blue screen with the dog trainers and our hero dog, Uber, with multiple cameras. We had the dog perform a wide range of actions and emotions. We captured a lot of material, walk- ing cycles, running cycles, getting up, laying down, barking, all kinds of things. I gave all of that material to Framestore along with multiple photo scans of the dog. We did extensive scanning of all of our dogs and a lot of reference material with David that we called the "Dogology," which was kind of our visual bible of Uber performance. We gave that to Framestore and they came back with a split screen where on one side, you have the real dog doing a complex action with both body motion and facial expressions. On the other side of this screen, you have a CG version of that same action. Framestore went on to build the assets with their amazing crea- ture pipeline. They developed this fantastic tool kit that is called FAT, which stands for Framestore Anatomy Toolkit. It's a big step up in their creature effects pipeline, where they're able to retarget muscle rigs and fat, skin, and fur simulation very quickly from one rig to an- other, and simulate multiple things on top of each other. You need less artist time to get a great looking first pass, you can do a lot of things that are brute simulation passes, very, very fast. But building an asset at this level, that can hold up in an extreme close up, that is a work of art. So they built the dog based on the scan. Then they did a very de- tailed matchmove of that real dog performance and they sent us this test. I had the director and the studio executives look at it on the big screen and asked them, "Okay, which one is the real dog and which one's the CG dog?" And that opened up the trust to do more complex work. So the foundation with Framestore was to create a replica of the dog that was completely seamless, getting these as- sets to a level where you basically can't tell the difference. It's great when you have a reference and you're just matching it because you