In Part 1 of this series, Ellen Besen sits down with maverick CG director Chris Landreth, creator of Bingo and the new, breakthrough film Ryan, to discuss the current state of CG human characters and realism.
There have been many attempts over the years to get a serious CG conference and exhibition off the ground in London. The fact that none of them has succeeded in the long term is a bit of a puzzle, really. After all, London is one of the global epicenters of excellence when it comes to the digital visual arts, with a goodly percentage of the best vfx work in all fields emanating from the cramped streets of Soho. And beyond the capital, the depth of talent in the U.K. games industry is truly wondrous to behold.
Maybe people are too wedded to their trips to Amsterdam for IBC, and Stateside for NAB, SIGGRAPH and the GDC to want a top-rank show on their doorstep curtailing all that lovely expense-account travel. Maybe no one yet has managed to come up with the right mix of big-name speakers and new, must-see technology to bring the punters in through the door. Maybe everyones just too busy when it comes to it. Whatever, all recent attempts have fallen on some pretty stony ground.
The UK CGI Festival, though, on the evidence of its inaugural event earlier in December, has a better chance than most of escaping the culls and prospering into the future. Yes, the vibes from the Leeds leg of the event in November were disappointing. (Leeds is a couple of hundred miles to the north, and thus to Europeans a long way away). And, yes, exhibitors moaned in London about the ratio of students vs. facility heads with purchase orders in their hands, but when dont they? And, yes, the London exhibition was fairly small and insubstantial. But attendance was fairly good, the conference sessions had a discernable number of warm bodies in them (a couple were genuinely packed) and there was some fascinating stuff to offer.
One of the more interesting things to note was how often the subject of crowd replication and behavior came up, with most of the speakers from the vfx end of the spectrum mentioning it at least somewhere in their presentation. As The Mills Jordi Bares put it, You get one film going out with crowds in and suddenly everyone wants crowds. Naturally enough, however, it was very much to the forefront when Stephen Regelous, writer of the Massive software that steered so many orcs to gory deaths in The Lord of the Rings trilogy, hosted a masterclass in Artificial Life for Animation.
Regelous who committed the unfortunate though very Kiwi sin of having an attractive woman on his stand with the company name emblazoned across her chest is an affable chap with an impressive showreel. Hes also entertainingly clumsy at using his own software at times, but no amount of Ooops, I didnt mean to do that moments can hide the power of Massive. Beyond showcasing the software itself and making reassuring noises that the motion tree design tool is nowhere near as complex as it looks once you start tackling it, Regelous also showed the first entrants in the ready to run agents library introduced at SIGGRAPH: Loco Guy, Stadium Guy and Ambient Man. The promise is that more are on the way for 2005, as will be the realtime animation tool Massive Live.
In a complimentary vain, Natural Motion ceo, Torsten Reil, demo-ed his companys highly impressive endorphin package and talked the audience through its development. The software originated at Oxford University, where researchers were attempting to get stick figures to walk using a combination of neural networks and physics simulation. Fast forward a few years, add in some biomechanics and other things and youve got endorphin, which is the sort of software that has animators shaking their heads in disbelief when they see it run in front of them.
Natural Motion grandly calls what it does Dynamic Motion Synthesis, but its essentially a mix of biomechanical data coupled with adaptive AI behaviours. Push a 3D character down the 3D stairs, and the animator can set him up to first try to keep his balance, then curl to protect his head and ribs, then relax at the end as if losing consciousness. The result is very convincing animation in very quick time all ready for export to Maya, XSI or wherever. And with the introduction of background dynamic retargeting, more behaviors and the company working on what it calls signature extraction being able to distill the essence of a movement out of MoCap data and use it as a dynamic asset the futures looking bright. Soon, concluded Reil, an animator will only need a few seed assets to create all their animation needs.
Back to the crowds, though. Part of Regelous message is that crowd software is about more than just large armies going at it. It was thus interesting to hear The Mills Bares talk through a couple of commercials the post giant has done using Massive; namely PlayStation Mountain (people clambering over each other) and Nike The Other Game (stadium crowd replication), both the sort of highly entertaining and technically demanding projects that London specialises in.
Over at DreamWorks/PDI, meanwhile, effects supervisor Mark Wendell explained how they used SOFTIMAGE Behavior to build and animate the crowd scenes in Shark Tale. We would have liked to have used Massive actually, he said, but it wasnt really ready when we started development on the movie.
Other tid-bits grabbed from the presentation: Shark Tale had to rotate fish fins 90 degrees to give the illusion of bipedal movement and all the innate expressiveness that goes with it to the audience; characters were rigged especially to have extreme deformations without breaking; the movie would have taken one person 474 years to complete; occupied more than 30TB of disc space and gobbled up more than six million CPU render hours. Oh, yes, and 27 babies were born to DreamWorks employees during production.
As this shows, it can be a bit frustrating getting the one-hour overview of a production at a conference, leaving you as likely to come out with an idea of the reproductive rates of animators as about anything specific and useful. So the best presentations tend to be themed and focussed on one topic, such as Scott Easleys excellent masterclass on the making of EAs Medal of Honor: Pacific Assault. Easley was lead animator on the game and is passionate about detail. Its all about getting it right as far as hes concerned, so a group was sent on a three-week field trip to the Pacific theater of war where they found amongst other things that sand behaves differently from how it does on Venice Beach and that Japanese soldiers moved in a fundamentally different and more efficient way from their American counterparts. So they engaged Japanese actors to get the MoCap right, researchers to explain the culture of the Empire at the time, military advisers to make the animators realise what being in the middle of a war is like, took pictures of the sand and generally immersed themselves in their work.
Animators should take ownership of their work, says Easley. But not to an excessive extent. Addressing the inevitable question from his audience on long working hours at EA, he replied, The Medal of Honor team does not work those hours. At six we kick them out. If you have an efficient pipeline you dont have to work those hours.
So there you go
Efficient pipelines are becoming more and more important as the 3D industry continues its drive toward real world complexity. Take Danielle Feinberg, lead lighting artist at Pixar. On Monsters, Inc. they had 59 lighting set-ups. On Finding Nemo there were 50. On The Incredibles there were a whopping 179 with 17 people doing master lighting set-ups alone. That sort of work needs a fat and very well-designed pipe to stop clogging up.
Feinberg nicely illustrated Pixars steep learning curve when it came to doing The Incredibles with the salient fact that the company had only ever done cloth once before, Boos t-shirt in Monsters, Inc, and now it was faced by a whole cast wearing the stuff. So they tackled that, they tackled hair (A physics project surrounding the head) and they tackled people, and sometimes they got it right and sometimes they got it wrong.
Right was the subsurface scattering routine they wrote to get human skin looking good. They found out that they were going to have to turn it off at certain points on a body, otherwise Elastigirls elongated limbs were going to glow rather strangely. So, they then hit upon the bright idea of turning it off under the characters clothes, thus saving render time. Bonus. Wrong was spawning all the background characters from one model, Universal Man. Not only did Universal Woman never quite establish her own identity as a result, but also as Universal Man was carrying around all the fixes for all his spawned progeny, Pixar found it could only render 10 of them at a time.
Elsewhere, Jim Radford, head of creative CG at The Moving Picture Co. (MPC), detailed how the company transformed the faces of actors into exact replicas of WWII leaders for Tiger Aspects Virtual History The Secret Plot to Kill Hitler (good tracking, having people that looked mostly there already and using filming techniques of the time seemed to be the keys). In its recreation of 1940s footage and blurring of the boundaries between history and drama, it was a highly controversial project expect to see more work like it soon. And Cinesite visual effects supervisor Matt Johnson, whod spent much of the previous year laboring away on Jerry Bruckheimers King Arthur, happily quoted from a Variety review that stated it was `great to see a summer blockbuster with no visual effects. Considering that one scene, the ice battle, was essentially 275 vfx shots cut back to back, Johnson thoroughly deserved to look pleased.
The busiest session of the two days, though, and one that illustrates the preponderance of students in the audience, was the Showreel Surgery run by Shelley Page, the European rep of DreamWorks Animation and one of the top talent scouts in the industry. DreamWorks gets an estimated 300 reels a week, so when Page pleads for animators to keep them short (three minutes max) you can understand where shes coming from. Other advice included: make it visually exciting, avoid repetition, clearly indicate what you did, list software used, include contact details (youd be amazed how many people forget that little detail apparently) and always check studio Websites for any submission requirements. And if youre determined to try getting into character animation, whatever you do, no robots or sliding feet.
What we really like to see are problem-solvers, she said. People who have taken off the shelf software and wrangled it and tortured it to make it do what they want it to.
EAs Easley then chimed in with guidance for the games side of the industry. Show me a walk, show me a run, show me a jump, he said. Hit it, then quit it.
Andy Stout is a U.K.-based freelance journalist, who previously worked in Quantels marcomms department. Before that he spent a decade writing about 3D and vfx for numerous U.K. magazines.