There’s just something particularly endearing about realistic-looking animals that talk. Even more so when you add espionage into the mix.
In 2001, audiences got a behind-the-scenes look at some of the scheming that takes place between cats and dogs, as these domesticated animals scratch and bite their way to the top of the household pecking order in the feature Cats & Dogs. The film introduced the diabolical white Persian cat named Mr. Tinkles and his scheming operatives, as well as a cast of canines who use extensive intelligence to outsmart the felines. Recently, moviegoers were reintroduced to this ongoing animal war in
Cats & Dogs: The Revenge of Kitty Galore. For this sequel—which (surprisingly) has these two rival species collaborating against rogue cat agent Kitty Galore—director Brad Peyton ramped up the action, while the visual effects artists at Tippett Studio dialed up the realism of the talking characters. Sometimes this required face replacements, while other times it meant an all-CG character.
Here, Tippett co-VFX supervisor Scott Liedtka, along with graphics software engineer Michael Farnsworth, animation supervisor Will Groebe, and compositing supervisor Colin Epstein, detail the studio’s work on this fun-loving family film.
What was the scope of Tippett’s work on the movie?
Cats & Dogs: The Revenge of Kitty Galore, Tippett Studio (TS) finaled close to 400 shots in about 18 months. Blair Clark, VFX supervisor at Tippett, was on set supervising the shooting of principal photography for several months. He also supervised an all-day element shoot on the Bay to photograph sky and water backgrounds for the Ferry Fight sequence.
Pre-production for this show lasted several months as all the characters and props were designed and built. There were several characters that had to match animal actors on set, but there were also a number of characters that would always be CG in the movie. So they had to be designed first. The evil henchmen were particularly fun to build because they had a lot of gadgets, like the scuba suits and an underwater sub right out of Thunderball.
This production changed a lot while we were working on it. The main actors were recast, and the ending of the film was completely re-envisioned. Because of all the changes, a big part of the job was juggling the work around so that the crew was always working on something useful. TS also did a lot of work putting backgrounds behind greenscreen foregrounds that were shot on a stage. So, we built a lot of props and characters, and then created engaging character performances. There is a lot of over-the-top animal acting in this movie.
How many characters was the studio responsible for?
(Liedtka) TS built 10 different CG characters, including the evil Kitty Galore, Seamus the pigeon, Scrumptious the mouse, Puppy, Paws, cat assassins Angus and Duncan MacDougall, Catherine, Diggs, and Butch (for face replacement shots only). Seamus was the pigeon that everyone was trying to get and was also the comic relief. He was a major character for us, in addition to Kitty Galore, with about 150 shots. The Ferry Fight sequence was big for us, with a lot of characters and a lot of action that our animators laid out before it was shot. The MacDougall twins appear in this sequence in tricked-out wetsuits as they try to steal Seamus from Butch, Diggs, and Catherine.
You have done some animals before. How did this work differ?
Liedtka) With every show, we try to make our characters better, both better looking and easier to work with. Before this movie started, TS wrote a new fur system from the ground up. We were still using the fur system that had been developed for the original
Cats & Dogs, but as a program, it was getting harder and harder to support and add features to. So the fur on all the animals in the sequel was created using a brand-new system. That took a few months for the R&D department to get ready, and then a few months more for the art department to adapt to the new tools. It was a good thing that we got the new fur going because we couldn’t have done Kitty without it.
Aesthetically, we worked on a few things. Our animation is always strong, but to support them, we worked really hard to get clean models and rigs. We also put in extra time on our eye textures and reflections. We used more image-based lighting on this show than in shows past, although those processes continue to improve. Production-wise, we relied more heavily on Shotgun’s software for production tracking, but as previously said, we also spent a lot of time juggling things and trying to shield the artists from all the stopping and starting that normally results from radically changing schedules.
What was the key to having digital animals stand up well alongside real animals?
(Liedtka) Matching a digital animal to a real animal is difficult. It is much easier to create a believable character that is always CG in a movie than matching to a real animal. The problem is that even though shot to shot, frame to frame, the image of the real animal changes, its essence stays the same. Kitty stays Kitty even when she completely changes appearance; somehow you know it is the same cat. So how do you capture the essence of the animal? We started with really good reference. We used a rig (developed by Sony Pictures Imageworks for
Beowulf) to simultaneously capture reference photos from six cameras. Because an animal won’t hold still for several exposures, shooting from different angles catches different views of the animal in the same pose. This reference was matched to ensure that all the dimensions were exact. For the face, there was a lot of iterating on the smallest details. One of our art directors, Mark Dubeau, calls it ‘a war of millimeters,’ but the fact is we just kept working at all the little details, especially around the eyes and mouth. Even after the model and the look match, it’s tough to sell the character if its animation is a lot different than the real animal. Often times we are asked to deliver a specific performance that the real animal would never do, and that takes away from believing our CG creation is ‘real,’ but that’s the process.
What modeling tools did you use?
(Liedtka) At Tippett Studio, we use a lot of Maya throughout the pipeline, and our modelers use that to lay out the basic mesh of production models. We are also using both Pixologic’s ZBrush and Autodesk’s Mudbox for sculpting. The sculpting tools are really useful for mocking up prototypes before shaping them into Maya production models and for developing facial expressions. The other thing ZBrush and Mudbox were used for was sculpting the wrinkles for Kitty. These sculpts were then converted to the displacement maps so the shader could organically blend between them.
Did you use any new techniques?
(Liedtka) Kitty required improving a number of techniques that we had been developing over a number of shows. (See “Fur Less” in the October 2010 issue of CGW to learn about the unique issues that Tippett faced in creating the hairless cat Kitty Galore.) On this show, we did more face replacements than we had done on previous shows, so we also fine-tuned those techniques. Kitty was originally planned to have 20 or so face replacement shots, but they were omitted. We did a fair number of face replacements on Catherine, Diggs, and Butch, though. We wrote a completely new fur system. (See “Hair-raiser” in the December 2009 issue of CGW for details on the new fur system.) We used new production-tracking and scheduling techniques. We took advantage of studio-wide improvements to Tippett’s infrastructure, where we have initiatives going to simplify and modularize the way production artists and software interact with the pipeline. We integrated the playback tool RV and created pull-down menus so artists could see their latest work in the cut. RV was also really helpful in creating quick playlists—for example, to throw together a playlist of all the dailies sent for a client review.
So the new fur system worked well?
(Liedtka) The new fur system took a few months to develop to the point where it could be used by the art department, and a few months more to get all the bugs worked out and the new techniques mastered, so there were many challenges along the way to get the new system running well. Whenever you introduce a new system alongside an existing one, the fear is that artists will slip back to the old way of working because they can get more done faster the old way. With the new fur system, there were bumps in the road, but as a system, it was universally hailed as a big improvement.
Was there anything especially unique or compelling about the modeling process?
(Liedtka) To me, there were a few interesting things about the modeling process beyond the fact that our crew is just so amazingly talented. One new, cool thing was the use of simultaneous reference stills to better capture the basic forms. Also, we modeled five cats for this show, and they were all based on a super-clean master cat. Great care was taken to lay out the loops so that the muscles (and wrinkles for Kitty) would flow naturally, especially when the cat began to move. We started with a model that we had used before, but reworked it extensively, and it really came out great. The last thing of note is how much modeling work goes into the facial animation rig. We used sculpts for the basic expressions, and then the modelers painstakingly cut up the sculpts and turned them into component blendshapes. The advantage to this approach is that it works well to keep the character on model. The downside is that it is a ton of work; that’s where a lot of the modeling time for a character is spent. And we had 10 different characters to do!
Was the animation keyframed?
(Liedtka) All the animation from TS was keyframed in Maya. We are using some motion capture on some current shows, but on
Cats& Dogs, everything was keyframed.
Did you use any references? If so, how did that help?
(Liedtka) Actually, we used a lot of reference. We looked at the witness-cam footage as the actors performed their lines. We also looked at previous performances of the actors in their live-action movies. There were some inspirational pieces that we looked at from classic movies, as well. One difficulty was that when the performer changed, even though the lines were the same, the read was completely different. We basically had to start over on more than a few shots when parts were recast. Reference is always good for those subtle nuances to help sell the character.
What was the key to pulling off the animation?
On Cats & Dogs: The Revenge of Kitty Galore, we animated a variety of characters, each with its own unique characteristics and stylizations. The main villainess, Kitty Galore, was a fun combination of realistic animal movements and over-the-top acting. As far as the realistic movements, we knew we had to match our CG cat to the actual cat that was shot in some scenes, so we had to be able to animate our cat moving and acting like the real cat. We studied at a lot of video reference of the actual cat shot on set. She was a very hyperactive cat; not at all like the relaxed housecat many people enjoy in their homes, but a more agitated cat with fast tail movements and a constant shifting of attention. We had to make sure the tail on our cat was wagging at the same pace as the real cat, and that our cat had the same erratic head and body movements.
For the more over-the-top acting moments of Kitty, we were lucky enough to have the voice of Bette Midler for inspiration, as well as drawing upon other classic feminine villains from film and animation, such as Gloria Swanson's Norma Desmond from Sunset Boulevard, and Cruella Deville from the original 101 Dalmations. Kitty has a white mouse as a pet, much like Blofeld's white cat in the Bond films, with whom she has a sort of love/hate relationship. That was another fun aspect of her character to play with. In one scene she will be stroking her mouse lovingly, and in another toss the mouse in anger.
What were some of the big issues you had to contend with in terms of animation?
(Groebe) One of the more difficult things to accomplish on this show was that it was the largest show we had worked on to date. We had to grow our animation staff to over 30 animators to finish the nearly 400 character animation shots. We streamlined our hiring process and training methods, and adjusted our staffing to incorporate sequence leads who managed their own teams for each sequence.
What other software did you use for this film?
(Farnsworth) RenderMan 14 is our primary rendering package. Our typical hardware setup consisted of 64-bit four-core Dell machines with 8GB of RAM running Fedora Linux. In addition, ZBrush, Mudbox, and Maya were our main software packages, again with most time spent in Fedora Linux. There were some exceptions: We did use some paint and texture facilities on Mac OS X; and sculpting in Mudbox and ZBrush were usually done on Windows XP.
Compositing required particular attention. Please describe that work.
(Epstein) For the compositing team, the film was quite the hodgepodge of challenging tasks. In some shots, we were replacing greenscreens with ocean views. Others had us combining multiple takes of animal performances into one image. Sometimes we were placing full-CG animals into shots, sometimes we were doing face replacements on real animals. Several times, we were doing all this in one shot. We also had some entirely-CG environments to refine. Along with all that, some of our CG characters, especially Kitty Galore, had several subtle compositing touches included in refining their look in every shot that they were in. This ever-changing set of compositing requirements kept the team on its toes throughout our time on the show.
What was some of the more complex scenes in terms of compositing?
(Epstein) The Ferry Fight sequence, in which our heroes battle cat assassins known as the MacDougall Twins on an Alcatraz-bound ferry, was probably our most complex chunk of compositing work. The entire sequence takes place out in the San Francisco Bay, but it was shot on a stage ferry set in front of greenscreens. As soon as these plates were turned over, we were placing skies into those backgrounds so that the TDs had something complete to light to. We took a chartered boat into the San Francisco Bay and shot several background sky and water plates from the boat, looking in specific compass directions. Using a chart with these directions marked on a boat silhouette, we matched camera angles shot on stage to the backgrounds shot from the real boat. This allowed us to have a geography to the sequence that helped tie the shots together, rather than have somewhat random seascapes in our background. Plus, we got real landmarks, like Alcatraz Island or the Golden Gate Bridge, which helped set the scene.
Added to that plate work was the number of shots that were built out of multiple takes. If there was more than one real animal in the shot, they were usually filmed separately, with their trainers getting individual performances out of each one. It was our job to combine these passes seamlessly. This involved a lot of diligent work from our roto/paint department, like removing trainers or their shadows from many, many different elements. The comp team was under pressure to get at least rough versions of these assembled plates out to the rest of the crew as soon as possible, because match-move and animation needed to know where all the characters were. Unlike most films, whereby the compositing tasks tend to be weighted far more to the end of the production, we were in there from day one, providing the foundation for everyone else’s work.
How did the look development affect compositing?
(Epstein) The Ferry Fight sequence also had some major look-development tasks for the compositing team. The first was the MacDougall Twins themselves. They wore diving suits with round, glass helmets. It took a lot of work to get those helmets to look physically real. We didn't want the twins to look like they just had circles placed around their heads; we need different levels of reflection, refraction, and surface detail. My fellow comp supervisor, Chris Morley, did a bang-up job working with the art and TD departments to come up with a standard set of renders delivered with every MacDougall shot that let us dial those helmets in. We got rendered maps that we used to add details, like dried water spots on the glass and subtle grunge at the seams of different sections. The renders of the twins’ heads inside the helmets, as well as the background behind them, were actually refracted through the helmet geometry before they came to us. So were all the rendered mask layers that we’d normally use to dial in the look of the cats’ faces and eyes. All these elements, combined with a standard comp’ing stack that Chris developed, let the comp’ers control all the layers to meet the needs for each shot and make the MacDougalls look their best.
The twins also factored into the other major look challenge in the Ferry Fight sequence. The first time we see the twins, it’s in a long, underwater shot that reveals them coming out of the deep murk on their own diving vehicle and sneaking up to the ferry. This was an entirely-CG shot, comprised of dozens of layers. Our main comp’er on the shot, Jordan Schilling, put in a lot of time and ingenuity on that one. He developed comp treatments for each layer, and worked directly with other departments to break everything up into the layers he’d need to massage the shot to look as convincing as possible. Since the shot was meant to take place in the heavily trafficked bay and not some pristine lagoon, every element needed a certain amount of murk and grunge, from FX detritus in the water, to a CG school of fish, to the twins themselves, and to the hull and props of the ferry. The comp work ran the gamut from creating most of the background environment, to treating rendered FX scuba bubbles, to have the silvery, reflective look of the real thing. It took a lot of time and countless iterations, but it all came together beautifully.
Did comp have a big challenge when it came to Kitty Galore?
(Epstein) Perhaps the greatest challenge for every department working on the film was realizing Kitty Galore. While some shots had a real cat in them, and we did a few face-replacement shots, the majority of Kitty’s screen time featured our fully animated character. Creating a convincing hairless Sphynx cat poses so many challenges, with no place to hide your mistakes. Comp worked very closely with art, TD, and our show supervisors to take the amazing renders of Kitty and make them look even better. Compositors made subtle but key adjustments to so many factors in Kitty. These included things like the amount of light scattered in her skin, how dark her many wrinkles got, the color of her few tufts of fur, and the visibility of the peach fuzz that covered her body. The standard compositing script for her eyes alone was pretty complicated. Once again, Chris Morley played a huge part in refining these treatments, and really helped Kitty look her best. Because Kitty was seen in so many different lighting setups—ranging from diffuse daylight to a nighttime carnival, and often in extreme close-up—the compositing team had to constantly react to each shot’s new challenges to fit her seamlessly into the shot. In the end, I think our work on Kitty turned out beautifully, if one can say that about a hairless cat.
What unique challenges did this project bring?
Liedtka) Creating Kitty Galore presented several unique creative challenges: matching her wrinkles and her velvety skin. Continuous editorial changes with actors recast after we had started production and a new third act written more than halfway through our schedule was another production challenge that we hadn’t seen before.
Is there anything else you want to point out about Kitty Galore?
(Liedtka) Kitty’s monolog introduction shot was so long that we had to cut it in half and put two different animation leads, Brian Mendenhall and Geoff Wheeler, on it. After they were both finished with their half, Geoff had to go back and animate over the transition to piece it back together into one shot.
Are there any details about the film’s ‘other’ characters that you’d like to mention?
(Groebe) One of the other fun things we did was study actual bird anatomy to help in the development of the rig for our pigeon character, Seamus (voiced by Katt Williams). Based on the research we did (from studying an actual pigeon skeleton to handling dead chickens), we were able to create a fully functional, realistic bird wing. On past shows, we had to cheat bird wings to work in CG, but now we have a great wing system.
(Liedtka) The wing rig for Seamus was a technical achievement. His wing rig was rebuilt twice to allow for his wings to fold and unfold realistically, as well as for him to act with them. In reviews, [director] Brad [Peyton] would act out what he wanted Seamus to do, and he would gesture with his hands as Seamus’s wings. It’s always a challenge to balance realistic movement with an engaging, comic performance.
In addition, we wrote a new fur system from the ground up for this show and implemented tangent space vector displacement. And, we all had a lot of fun creating the cast of villains: Kitty and her crew. The MacDougall twins were especially fun with their wetsuits and gadgets. Scrumptious was always cute. Paws turned out a lot easier than we had thought. We were really worried about sim’ing his long fur, but using fur shells made that come together pretty quickly. I hope people who go to see the movie enjoy them!
See the accompanying videos in the Video section for a look at this technology.
Karen Moltenbrey is the chief editor of Computer Graphics World.