| |
Interview with Graham Clark
Omation Studios' CG Supervisor talks about the technical challenges of Barnyard The Movie, the production process, the studio and XSI.
August, 1st, 2006, by Raffael Dickreuter
 |  | Graham Clark, CG Supervisor on Barnyard at Omation Studios.
| | |  |
| |
| | | |
|
How did you get started in CG and why?
1986, I tried CG, at Pratt Institute in Brooklyn where I was studying painting, but at the time it was faster to figure 3d co-ordinates for meshes out on graph paper and input that data than to make anything in 3D interactively (fun) so I went back to my painting degree and having a great time around NYC. Though I did use AT&Ts TIPs 2D software there to prototype my paintings. 1996, a decade later and it was all different. After my Film degree from Ryerson University where I did some minor Pascal graphic programming with a Targa and a short and unpleasant try at 3D Studio version 2 at home, I went to Sheridan College where Avrim Katzman helped me get going with Softimage 3D in 96. I then completely fell in love with the potential of CG. Later I chaired at an art college, a few CG departments, 3D etc and founded the first Motion Graphics and Game DEV undergraduate and graduate programs. 2006, currently wrapping on Barnyard as CG sup and trying to find time to finish my masters from Rochester Institute.
You have worked in France, Canada and USA. What's the biggest difference when comparing these three countries (related to do 3d work)?
Not much really, just scale. And it’s more production staff heavy in the US. At first I admittedly really didn’t see the need for coordinators, there were lots of different types on Barnyard but the departmental coordinators that I most often had to deal with directly on issues: Pete Busch, Mike Bodkin, Joy Carmeci, Rosie Server, Julie Peng, made it so much easier to concentrate on our work and kept us on track. and of course my own dept. co-coordinator for Shot Finaling and Hair/Fur, Philip Hillenbrand who kept us on track and up to date, making it easier to concentrate on Look DEV and make systems to help others stay organized in the pipe and solve issues.
Tell us about your role and responsibilities on Barnyard
I started as a TD, became Hair/Fur Supervisor, then Shot Finaling Supervisor, then CG Supervisor. It was a very interesting transition moving through the production developing pipe, tools, and look. As a TD, we were initially seen as providers of ‘band-aids’, there was no pipeline system till late in production, lots of people new to XSI, and not enough time or support or consideration of what was the real required deliverable as a healthy asset to the next departments people.
While hoping for pipeline system proposals to be approved, I tasked myself with showing people XSI solutions, setting up workgroups for each department along with basic glue tools here and there to help people on an as needed basis. Typical tools for topology modification, synoptics, layout tools etc. I had a lot of freedom, to solve things, under my supervisor Jason Barlow. As TD at Omation, I also had the fortunate experience of working as part of an RnD-TD collaboration with Helge Mathee and Adam Ferrall. It was a very obvious and sensible experiment that was very successful. I'd TD front end some tool needs and DEV and the RnD guys would do the real heavy DEV.
I learned an incredible amount from these guys. Thanks, our office was so much fun.
As Hair Fur Sup it was also a lot of work in new areas at first. Hair is currently the biggest hurdle it seems according to CG film colleagues, DVD commentary, and articles on most of the CG features coming out recently.
The main issues I was initially concerned with were:
1. Getting hair assets in production, no one wanted hair on anything except I think the director.
The character asset topology was not designed for hair, nor were the rigs designed to support hair rigs or hair tool attachments, like sliding hair rigs to prevent the current hair systems from ‘krinkling”. The basis for the sliding rig came from a concept Frederic Bonometti showed me when we worked on Doogal (Magic Roundabout) in France. We had to create a lot of crazy hair rigs to account for hair arriving late in production. I had to develop some ‘wrapper’ or ‘inverse rigs’ on top of baked meshes as the cleanup stage reshaped the moving surfaces following rough animation which often prevented us from using rough animation skeletons or their deformed surfaces to work off. To allow for non-linearity and updates to topology earlier in the pipe, layered rigs required some complexity even on the most simple things like tail tufts to allow for full range of motion from the simulation and animation that created some crazy twisting and up vector issues, while needing to maintain hair groom/shape.
Otis Tail Tuft rig. Bending hair by deformer without krinkle is solved with sliding rigs, while the crazy spinning that can arise from the internal hair transforms and upvector issues are solved with layered rigs that also allowed for asset updates earlier in the pipeline.
It was also very difficult to get our furred assets into the existing 'pipe'. Departments were not initially setup to include them. We made a few tools to accommodate people importing them. eg. Adam, Leonard Chang and I made some tools to load referenced characters, and appropriate mesh streams based on shots and versions from our database.
I think being Hair & Fur Sup was the most challenging thing I did in production because it required investigation in modeling, rigging in very unusual ways, rendering and shading issues, dynamics and animation, all of them not out of the box in a way that worked for us and needed custom DEV, and with hair added late in production, no place to easily get assets into the film.
There is so much to investigate creating hair/fur assets. Some of the weirder things you find yourself doing as a hair/fur artist are moments like buying lots of stuffed animals or getting caught in the grocery line starring at the back of someone’s head, or at their children or pets hair. There’s really no point in trying to explain, they are already freaked, an explanation will just seem weirder.
I think this site for odd jobs and crazy careers, and the other jobs listed, says it all.
One of the areas I enjoyed developing with my team was developing the overall pipeline and our internal department pipeline. Internally it was very important based on the long shelf life of film assets, and potential creation artists turnover, to have a consistent system for internal deliverables with just the live ops and mechanisms left to recreate the asset and pass only the most discreet and efficient healthy asset (ie mesh data, deltas, etc. in xml) as our deliverable to the next department. Jim Jiang took over as Hair/Fur sup and did and amazing job, especially with the character Wild Mike.
2. We also had to do lot of look DEV
Rebecca Macdonald/Baldwin began the first Look DEV on some of Barnyards lead characters with hair, working with Phil Cruden the Art Director and Steve Oedekerk the director to get a buy off on hair use on characters in Barnyard. Thank you Becca for being the founder of what was to become the Hair/Fur department. The FX department also required some look DEV for vegetation, trees, grass etc. In conjunction with some of the FX guys like Meng Lu, Elliot Rosenstein developed the render efficient grass meadow look using the Hair instancing Softimage added for us in XSI version 4.
We had some key characters with hair styles, that were not just straight long hair, but rather had hair ‘dos’, and other characters that required hair performance, ie hair moving in a stylistic way, beyond what simulation alone could provide.
It quickly became very apparent based on Steve Os notes that we’d have to create our own hair dynamics and animation system to achieve the motion he envisioned.
As Shot Finaling Supervisor I had to create another department from scratch that didn’t previously exist, creating an internal pipeline and fit it within the overall CG film pipe we were developing. Tim Smith, our production manager, decided there should be a Shot Finaling department before lighting, and that I should run this department. At first I just wanted to kill him, as I felt the layout department should just do it, but the amount of non-layout things that needed to be fixed or recreated, like more efficient shade networks or lower rez textures etc was extensive, and lighting could not be expected to light and composite and to put their shots together. Our task in Finaling was that we basically had to fix everything before it went to lighting, and wow, I guess I shouldn’t say it, but yes the shots were broken. Anyway Tim was right, sorry and thanks Tim.
One of our roles in Shot Finaling was improving asset health. Beyond the usual making shader networks more efficient and creating LODs etc.we had to develop tools and new asset types such as cards and sub-assets. I asked Ben Barker to investigate cards for all our vegetation which he turned around in a day. Ben Barker prototyped the card system using XSIs Rendertree, and created a system for auto-creation of the cards and their various maps including methods for illumination using any lights and allowance for self occlusion shadowing. Once this method was production proven I asked Matt Lind to develop a custom shader for it so we could further improve on performance and features, such as the need for only one ray depth etc and Matt had some interesting additions such as Photons and controls to alter the light angles, mattes by “clusters” of objects for post work, and surface shading for bent cards etc.
The lightable self shadowing cards in cases improved rendering times by thousands.
As CG Sup my role is to help interpret the Directors vision using the tools and resources at our disposal, to develop the pipeline and look DEV, and address issues while bridging the departments. Andrew Egiziano and Aaron Parry, our producers that arrived later in production to save the day, made it so easy to get a pipe going. Having supportive knowledgeable producers backing your efforts is so helpful. I also ended up directly supervising RnD, who I already listed, and the remaining TDs, Aaron Walsman (rigger and rig pipe maker extraordinaire), Ben Barker, Dyllan Lu, Jentzen Mooney, and Steve Caron.
Explain a bit the process how Omation Studios were setup for this production
Steve Oedekerk lives near here and started Omation Studios in the hills above San Clemente California. We are very thankful to Steve O for that, after a hard work day its beautiful to be able to relax in a small beach town. Omation Studios, at one point we grew over 240 to occupy the building next door as well.
Omation Studios, located in San Clemente, California near San Diego.
Omation is situated close to the beach, how was it working near the ocean?
Awesome to live and work near the ocean, we had a few “board meetings” at lunch hour. I surf or kayak surf at least 3 times a week, mostly surf Kayak because I'm scared of sharks in the ocean, I might also be scared of lake and pool sharks, but I'm not crazy, not scared of bathtub or puddle sharks.
Plus after work there's the great beachfire parties, swimming etc I love it, I've become so OC and I don't care.
Tell us a bit about the production process your department had to handle
As CG Sup I don't have a department, I'm the only one on the crew page with “n/a” for my department :(
Tell us about the extensive use of hair in this production and what the effect of that was regarding technical challenges as well as render times
We had to put hair on something like 120 characters, not including sub assets. Sub-assets were a concept I developed, predicting the Directors need to differentiate duplicate animals, without adding XSI Models to the database. Yes, you can override clusters even though Softimage told me I couldn’t :p The big hairy issue is getting what the film needs compared to what comes out of the box from XSI for hair. We essentially used very little except the rendered hairs, and were actually in development of our own shaders so it would have been all custom hair if time permitted. We had our own groom and dynamics tools and used only our custom dynamics for characters.
We really had to lower the render hair count to get decent render times and preview interaction for the lighters. One of the characters, Biggie Cheese, that Hair/Fur artist Katie Folsom worked on we let go with a high hair count, that therefore looks damn good and I think shows the potential, but we are still very happy with the outcome.
We also used hair for vegetation, trees grass etc, and even RnD’d some particle simulation flow guidance using hair.
What was the solution to get good looking dynamics for the hair of the characters?
Steve the director wanted a certain hair performance requiring custom tools. So after arguing extensively with Helge over a few pints at Oles, our local pub, and scribbling the overview for the tools on napkins, we went back to the studio 3 sheets to the wind. I built the front end and outlined my needs and Helge coded the tool I had envisioned and added so many more great ideas, and 9 hours later we had a wonderful hair dynamics system that allowed us to blend groom, collide, goals etc. It was really awesome, I am so thankful to have worked with Helge. I’m a big believer in tools that allow a mix of simulation and hand keying or goaling. Our hair dynamics tool gave us this as well as a number of other controls most importantly definable by weight maps and by profile curves along length of hair rather than just at base or the whole hair strand. This allowed us to maintain groom mixed with simulation, hand key goals for desired performance notes on characters, and hand guide collisions rather than just accepting the results of the sim or being limited to sliders as our only means of control.
Which is the most complex character also in terms of hair?
In terms of motion, Wild Mike or Mrs Beatty. Mrs Beatty needed to maintain a hair groom that allowed simulation (Very few peoples hair is long straight hair that moves like a shampoo ad). Wild Mike is arms, legs, eyes and no body, just hair, and there were very specific directors requirements for how his hair should move and specifc notes about where the hair was to be at given timeframes in each shot. Thanks to Jim Jiang, who did most of the work on Wild Mike. One of the other issues with Wild Mike was needing to post key the eyes through the hair with consideration of post motion blur etc. so I had the RnD guys write some custom zdepth tools for XSI and Eyeon Fusion.
What were some key elements that were developed from the ground up to suit this production?
Wow so many, we just finished documenting the thousands of tools we created in our reports for Omation and Paramount this week. The biggest thing created was our pipeline backbone software called o_net. O_net was basically our pipeline system, covering so many needs like:
-allowing check in and out of assets,
-creation of metascenes, metamodels, etc.
The RnD guys helped setup something called metescenes for us that allowed a user to load only required elements and their required relations. Metascene creation also allowed for scene rebuilds, a great way to get around the binary dump save XSI does, that can sometimes carry corruptions, slow scenes, and consume disk space through a production pipe.
-issues and solutions databases attached to assets
-propagating changes across LODs
-queing and batch systems
-showed relations across all assets, eg: with a couple of clicks one could go from the conformed EDL for film out, back to a redo of the contributing composite rendering on farm, to list its input image sequences, to the various 3D 2D apps that created them, to the scenes, to the models referenced within, etc, etc.
This saved artists hours of time when redos were requested especially when other artists that created it were no longer available. I could on and on but I think its best for some future review done with the RnD guys. I think the O_net pipeline system was as robust as any commercial solution out there and being ours, customization was immediate. Thanks to the RnD guys, Adam Farrell, Andi Ireland, Michael Beal, Helge Mathee and Jon Keto for being so accommodating and brilliant. We also had number of custom tools created for specific shots and individual departments such as the hair tools mentioned, rig systems, texture and shader propagation systems, and of course our own XML IO apps for storing metadata such as an XSI metascene (an XSI scn ascii file)
Explain one of the most complex shots of the movies and how the problem was solved
It would take so long to answer. But in brief it would be the first and last shots of the film. The last shot the “star shot” was a very simple shot calling for a very simple effect but the notes very particular. It took us a long time working through what Steve O envisioned.All the RND guys and Felix Gebhardt worked on this shot. Felix essentially created his own particle system with states and custom forces and fields to achieve it.
The “hill surfing” sequence at the beginning of the film is one of the most complex by far but since we completed it near the end of production and we were a much better functioning machine by then it went fairly smoothly, and where it didn’t thanks to some of the excellent lighters like Francois Madere, Charlie Winters, Peter Boeykens, and Becca Baldwin for solving the issues and creating some beautifully lit shots. This sequence has thousands of trees, millions of hairs for grass/vegetation, and covers extensive terrain as the characters fly down the hill creating so many issues for the lighters with pass and asset management, and rendering issues like moblur.
How was XSI used in this production?
XSI was used to produce almost all the 3D elements rendered out to image sequences used for compositing. I’ve been an avid fan of Softimage for years and given the requirements of Barnyard I’d recommend XSI if we had to do it all over again.
Which features did you find very useful for this kind of production?
The SDK. NETVIEW interface. Instead of workgroups which can slow XSI and a network down the RnD guys had a setup that cross referenced the Database for comparison by date and ‘pushed’ latest tools out to the user available through a customized XSI interface that I think most users did not realize was not part of the standard XSI by appearance.
Which areas should be improved?
Particles, they are simply way too basic, black boxed, and overly kludgy to navigate.
Operators and scene graph. I love proceduralism and real non-linearity and the op stack is far too linear, things need to be redone all the time instead of being able to change inputs. Gator is a very good step in the right direction but to me is more of a workaround the linearity. This is huge and important. Not being able to properly prototype with branching, to quickly asses how it would turn out by changing your original steps, and not being able to change the inputs with mechanisms to accommodate changes, ie non-static clusters is frustrating.
FxTree, or perhaps AVID should just make or buy a compositor for a few reasons.
The mid level composite market is struggling but we all need a good solid compositor with a strong and well supported SDK.
I think the composite part of the pipeline in CG film is far more the hub for tracking assets in a pipeline, and addressing more quickly and cheaply the directors and others requests. Most of the work for a CG film may be done in a 3D app but from an overall view the hub of film production is around the edit with the director, and the prior step in the pipeline at the composite stage is the faster and cheaper place to change things if setup well. For example post lighting instead of relighting and re-rendering. The composite stage is also to me, more than in the 3D app, the pipeline hub, being the larger collection of more final various input sources 2D 3D and data. Maybe the balance will tip back with the current advances in real time shading. We used the FXtree in Layout and Shot Finaling to comp our rendered passes to evaluate the shots of assets before passing them on.
Softimage, Marc Stevens, Simon Inwood, Jennifer Goldfinch, Christine Charette, Chinny etc have been great, really helpful and informative. I look forward to working with XSI and them again.
How easy was it to find talented artists to fill all the required positions?
It was very difficult, our associate producer Andrew Egiziano was frustrated at times. We had to bring artists from all over the world and with stricter visa requirements due I think to 911 etc, I could not get some really great people from the XSI community that I wanted here. There is simply a real shortage in the US of *production* knowledgeable artists that know XSI, and unfortunately still, a shortage of schools with *production* experienced teachers.
Graham Clark, Amanda Petrey, and Phillip Hillenbrand at Paramount Studios night of the Barnyard Movie wrap party. (and Ben Barker is upper left BG)
Recently many animated features have been released and many more are announced. What do you think about this trend and where do you think will it go?
Sometimes its scary and yet other times good when I overhear people at restaurants or the bars getting very excited about CG films or talking about their kids reciting the popular phrases. I think maybe there is enough space for so many.
Most animated movies these days are about animals, where do you see a lot of potential where animated movies could explore different playgrounds?
I don't think most animated movies these days are about animals, they are about people represented as animals. People relate to animals so well in CG, CG people are still too creepy, cars certainly seem to be ok personified.
Is there anything you would like to say to the rest of the cg community?
Sure, Cheers to all of you and see you on the next one! And to the people I worked with at Omation, Thank You! I know I've not mentioned everyone at all in my responses here.
Everything is about people, not tools etc and we had some great people on Barnyard, hard working, dedicated, investigative, and funny at the toughest of times. I'll miss them and hope to work with them on upcoming projects. Bye Bye, Barnyard Buddies…
BTW I’ve been writing this interview on my deck:
Related Links
Graham Clark
Barnyard - The movie
Omation Studios
Discuss Article at the forum
| |
|
