| |
Interview with Olivier Ozoux
Associate Digital Supervisor on Happy Feet, Olivier Ozoux talks about building a pipeline for an animated feature, animating large crowds of penguins, his past days at Softimage Special Projects and using XSI in production.
December, 5th, 2006, by Raffael Dickreuter, Bradley Gabe
 |  | Olivier Ozoux, Associate Digital Supervisor
at Animal Logic on "Happy Feet".
| | | |
| |
| | | |
|
Tell us a bit on when and why you got started in the cg industry
I ended up in CG a little bit by chance, I was living in Milan (Italy) at the time, and I was fascinated with industrial design, and I wanted to study at the INSTITUTO EUROPEO DI DESIGN. Unfortunately I knew I couldn't afford a 4 year course, but they had just started a 1 year introduction to Computer Graphics . This was in 1988 so we had Amiga's and a few DOS PCs with the little turbo buttons on them. I got my first CG job while I was still in class, doing 2D and 3D graphics for local television.
You've been working at Animal Logic for the past few years. Can you tell us of the responsibilities you had a the company
I worked at Animal Logic for 3 years, as Associate Digital Supervisor on Happy Feet. I came on board to setup the production pipeline on the performance side, and to work with Brett Feeney, the other Digital Supervisor, and help translate George Miller's creative vision into reality. The Digital Supervisor position is also called VFX Supervisor or CG supervisor in other companies.
To what extent did the Animal Logic pipeline suit an animated feature production, the company so far more known for its visual effects work?
This is the part that attracted me to Animal Logic, the fact that this was their first animated feature, and in some sense this was our biggest challenge and our greatest asset. I also really liked the script, and the fact that George Miller was directing.
We had to build a production pipeline from nothing, and even though the technology used is the same, the fundamental difference between Visual Effects work and Animation is the creative process. In VFX you receive plates, and the live-action components channel you toward the solution. Here, we had a blank canvas to work from, and everything could change at any time. This changes the creative process, and that has a huge influence on how the pipeline needed to be put together.
At the same time, not having a set way of working or a long history of making animated features in a certain way allowed us, I think, to make a film that was fundamentally different. George Miller is a live action director, and I think it shows in the approach and cinematic language of the film. We built a pipeline that would fit that way of working.
Tell us about the production process of Happy Feet and your involvement in the production
I was in charge of the performance side of the production, so everything from previz, motion capture, editorial, sound, character rigging, animation, deformation, and some of the crowds. I was working with the supervisors of each of these department, to figure out how to best use the existing tools and technology, and what additional tools were needed to pull off what we were trying to achieve.
I also spent a lot of time trying to figure out how we could give tools that would let George interact with CG almost as if he was directing a live shoot. Using mocap as the basis for a lot of the animation helped, because George could direct the performers on stage, but we also ended up building two tools that allowed him to work with the performance data and the cameras in the same way he could work with sound and images in an editing package. This gave him a virtual sandbox where he could find the right performance, viewpoint and timing, and once we had the shots well defined could add all the subtleties of animation, effects and lighting. I think that's what help make the cinematic experience of Happy Feet different from other animated features.
What do you consider the biggest challenges of Happy Feet you had to solve?
We only really had 2 things to worry about in the film, Penguins, and Antarctica. Thankfully, the various water in all its forms (Ocean, Snow, Ice, Clouds) wasn't my primary responsibility, but I've had many penguin-induced dreams. Merging motion capture and animation was something we worked a lot on, because we wanted to keep the advantage of each approach without imposing the limitations of one on the other, as well as diffusing the usual quasi-religious issues that can arise when you use mocap and animation in the same sentence. From a rigging and skin deformation point of view, the secondary characters, leopard seals, skua birds and the elephant seals, each brought their own challenges.
Can you tell us about the pipeline and how data transfer was achieved between XSI and other packages
We used XSI for the entire character pipeline, including rigging animation and deformations. Surfacing was done in Maya, and Modeling was done in whichever package the artist felt comfortable with, which ended up being Maya for characters and XSI for the environment. We ended up writing 2 custom tools to manage the data in and out of XSI, one to talk with the mocap system and transport the mocap rig and data, and another one that would transport animated and deforming geometry around as a point cache.
Voice actors for "Happy Feet" Robin Williams and Brittany Murphy.
What were the approaches for facial animation, making subtle expressions possible for animators?
The facial rig was deformer driven, rather than using shapes, which allowed us to re-use the same rig elements across all the characters, but also gave us some independence from the geometry, which could still change after some of the animation was started. I think it also gave us a lot of control around the eyes, allowing a lot of secondary motion between the different areas.
George Miller and Daniel Jeannette (our Animation Director) ended up obsessing about the eyes, adding all these micro eye motions and tiny little changes in the expressions. George liked to say that 95% of our visual cortex, the part of the brain that interprets what we are seeing, is dedicated to reading the eyes. I think the results adds a lot of presence to the characters, and makes them really feel alive. There is this brilliant scene towards the end where Mumble is reunited with his father. It's very powerful, yet when you watch it, the characters aren't really moving or talking at all, but whole story is in their eyes.
The film has many scenes with large crowds. Tell us a bit what techniques were used to create these shots.
We ended up using a mix of techniques, mainly because we had so many different type of crowds, and I don't think that any single solution would've been able to cover all of them. We had the standard naturalistic penguin crowds milling about and doing their "March Of The Penguin" impersonation for which we used Massive, then we had underwater swimming and a lot of dancing, for which we wrote our own crowd system.
The challenge with dancing is that you can't use randomness and emergent behavior from a few simple rules, which is the basis of most crowd system. So we took our inspiration in music sequencing programs and their humanize function, which can time shift a signal based on a master beat, so that all the rhythmic cues are preserved, but you still have something that feels like a group of individuals rather than robots perfectly in sync or a big mess that is completely out of sync.
We then spent a lot of time optimizing the back end of the process. As an example, we have the sequence where Mumble and the Amigos arrive in Adelieland. For a very long time, our screening would include a big red arrow with a caption that read "Here are 50,000 Adelie Penguins". When that shot was finally approved, their number had swollen to almost 700,000 penguins. And the other problem that we found out, is that black feet against snow is a lot less forgiving when it comes to foot contact and sliding than armies of sandal-wearing mercenaries raising a lot dust in their path.
What was the approach to create the fluids and oceans in Happy Feet?
We wrote our own water system for the project, and mostly used a procedural rather than simulation based approach for most of the shots, with the larger waves and ocean surface being done at the surface level, but with a lot of additional deformations and wavelets done in shading. A lot of the interaction between the water surface and the iceberg was also done in shading, ripples radiating out of the intersection, and of course the ice and snow becoming wet at the right place and time.
Which area of the production required the largest amount of custom tools being developed?
On a production of this size I don't think that there is a single area or department that didn't get custom tools written for it. We even had to build tools for editorial to help them cope with the amount of data we where sending to them, and to maintain the sync between our live tap-dancing which was captured on stage along with the motion of the dancers, and cope with changes done in both motion editing and editorial, all the way to the final mixdown.
Our main proprietary systems include the Ocean, Feather and Fur, Muscle and Fat, the entire Snow and Ice shading system called Freezer, and the whole renderfarm, asset tracking and remote processing infrastructure, and I'm sure I'm forgetting a few more.
How was XSI used in this production?
As I said, the character rigging and deformation system was all in XSI, along with all of the animation. XSI was also used to model all of the environment and props. We also used XSI for our motion selects, creating a single blended performance by hacking all the different takes together into a single seamless action.
For several years you were Director of Softimage Special Projects. What memories come to your mind when you look back?
As Director of Special Projects for Softimage, I was interacting with every kind of production companies and projects, all around the world. This gave me a unique perspective on the industry, and I was able to learn a lot, but also help setup all kinds of different pipelines and find solutions to very different problems. The Special Project team always had a foot in production and the other in software development, and we contributed to a number of great projects. What I did miss from full-time production, was being able to follow a single project from start to finish, instead of contributing as a consultant only for a few days or weeks.
If you compare creating 3d software for a software company and compare that to developing in house tools at a production company, where is the biggest difference?
The main difference is obviously the scope. Building tools for a production is a by-product and not the end result, so the way you approach a problem is fundamentally different. When you work in a software company, the end result is the tool, and that tool has to be sold, then used by a number of different people, all with overlapping but slightly different and often conflicting needs. When you work on in-house tools you usually try to solve a very specific, production-driven problem, and this allows you to tailor the solution to the problem. Having worked on each side has given me a much better perspective on both processes.
To what extent do you think software companies these days understand (or do not enough) the needs of big productions?
I think that the problem isn't so much that software companies do or don't understand the needs of big production, but also how much they can afford to be responsive to those needs. Large production companies are more interested in a solid platform that can be extended easily, because they have in-house R&D to add project or pipeline specific features but smaller companies are expecting tools that just work out of the box. Ideally software tools should cater to both, but that's hard.
In what areas of XSI and what features did you contribute to the software?
In the Special Projects days, I was also the de-facto program manager for Film VFX and long-form animation, so I was always pushing for scalability, performance, extensibility. When a developer gets really excited about what they put together with a bunch of spheres and cubes, you sometime have to remind them about what happens when you have to deal with dinosaurs, realistic human doubles, or rebuilding half a city. The other aspect that we pushed on in Special Projects is the whole API and Scripting interface, support for linux and python, things like that. On a more personal level, I can be blamed for a lot of little tools and scripts scattered around the system, which are usually called OO something or other.
Which areas of the software do you like the most?
I've always been a generalist, and I believe in the big picture, so I don't think that there is an area that I prefer over another. I've spent more time in the modeling, rigging, and animation side than effects lighting and rendering, but I'm at ease with both.
Which areas do you think should be improved after having used it extensively in production?
We pushed the envelope using the the XSI reference models, external sources and the animation mixer. Thankfully Softimage was pretty responsive, and about 20 or so QFE later we had a working system, but we had to work hard around some of these areas, so it's nice to see them improved in v6. The main area for a production the size of Happy Feet is stability and an API that gives you transparent access to everything the UI let's you do. There is nothing more frustrating than not being able to access some areas of the software, because the developers forget to give you proper access.
Where do you see cg software go in the next few years and which areas are still the most difficult to solve?
I think that we are at a point where most of the basic technological needs of 3D are understood. A lot of them are out of reach, either because they are still expensive from a computing standpoint, or because the technology is too complex for a project of a certain size or schedule. However, the one aspect that has made no progress at all is the way the various software assume people will collaborate on a project. 3D, because of its scope, is always a team sport. However, every single piece of software still treat the problem as if 3D was a solo effort, like word-processing. We need a Massively Multi User Creative Environment, not a series of standalone workstations.
Compare living and working in California with Australia. What is big difference?
Sydney is really a great place to live, and I really enjoyed the 3 years I spent there. The weather is fantastic, the Australians are friendly and easy-going, and I found the lifestyle to be a good mix of Californian and European sensibilities. On a work level, I think you face the same challenges and often work with the same people, regardless of location, be in Sydney, Los Angeles, Toronto or London. One of the things I love about CG is the fact that people working on a project often come from everywhere in the world. On Happy Feet we had people from every single continent, including the Antarctic, if you count our virtual characters.
What's up next for Olivier Ozoux?
Happy Feet turned out to be an amazing experience, and I'm glad that it's doing well in the box office. After spending the last 3 years in Sydney, I'm back in Los Angeles and taking a little bit of a breather as I'm looking for the next project.
Related Links
Olivier Ozoux
Discuss Article at the forum
| |
|
