|
|
| Interview
with the Fishes from SplutterFish CGarchitect.com interviews Scott Kirvan and Steve Blackmon, two of the fishes from SplutterFish, makers of Brazil r/s, to find out just what they went through to get this great product to market, and what they have in store for the future. 
CGA: Could you introduce yourselves? I am Scott Kirvan. Steve Blackmon and I are the co-founders of SplutterFish.
I have been interested in photography and CG since the early Amiga days (that should tell you how old I am). In 1992 after getting my electrical engineering degree, I started working at Impulse, maker of Imagine, one of the first 3D animation apps for the Amiga and PC. I wrote shaders, modelers, procedural effects, a dedicated 3D spline/font editor, etc. It was through Imagine that I met Steve, who at the time was one of the "hot" Imagine artists. Steve's interest in CG programming and art began when he was a kid, and after he left college, he then started his own graphic design company. He joined Blur Studio in 1995 and a year later, helped land me a job there as their lead production programmer. Steve is an expert lighter and over the years he has worked as senior animator, sequence lead animator, and on-set and off-set VFX supervisor, as well a programmer. So, that was the beginning of our partnership. With the intense schedules we had at Blur, we were able to pump out tons of custom code, including a ray tracer which eventually became RayFX in 3dsmax.
In 2000, I left Blur to do freelance VFX/technical consulting for film and television productions. Meanwhile, Steve started playing around with GI and Ghost, the precursor to Brazil. At the time when Ghost really started creating a buzz in the CG community, I had just finished working on a film project. In early 2001, Steve and I felt the time was ripe for us to start SplutterFish to bring this exciting technology to commercial viability. Over the years at Blur, we had talked about writing our dream renderer. The design of Brazil really is the culmination of years of brainstorming, algorithm refinement, and object-oriented core design. These discussions were happening long before Ghost was even started. Brazil r/s is not a technology that was 'slapped' together overnight. I firmly believe that when artists use it they can tell that.  Image Courtesy of SplutterFish - Rendered by Johan Thorngren CGA: Without giving away too much, and to help our readers understand how the various renderers out there differ, how does Brazil compare technically to other GI/Radiosity renderers out there? What techniques or algorithms is Brazil based upon? We don't think of Brazil r/s as a GI renderer. Brazil is a system of rendering tools based on a super-fast, highly-evolved ray-tracing engine. It provides many advanced tools like sub-surface scattering, area lighting, sophisticated global photon mapping/caustics, in addition to GI. With Brazil, the artist now has this arsenal of tools to choose from and can pick those best suited for a particular type of project (indoor, outdoor, highly-detailed scenes, etc). Brazil can render very fast using its advanced features and shaders. Area lights or global photons for example, can speed up render times by several orders of magnitude - hours to minutes in many cases - and is much faster than straight GI and with no or very little sacrifice in image quality. Another thing that sets Brazil apart is its highly-evolved ray-tracing engine. For the last six years, and up to the present, Steve and I have spent due diligence refining the engine--optimizing code, constantly redesigning and rewriting algorithms for speed, adaptability, etc. In my opinion there is no ray-tracer running on any platform that even comes close to it's performance. Another
key characteristic of Brazil is its anti-aliasing. Because both
Steve and I come from a production background, we understand that
this is a serious concern, particularly in animation. Other renderers
cut corners with their AA to enhance rendering speed. These short
cuts make the renderer impractical or unusable for many types of
production work, especially broadcast and film quality animation.
In order to get usuable results, you have to crank the settings
so high to get rid of the artifacts and flickering created by these
other renderer's "fast" algorithms, that the render times
shoot prohibitively through the roof. Using Brazil, you don't hit
that wall. Artists can render complex scenes with reasonable settings,
reasonably fast. Brazil's carefully engineered memory handling and
use of system resources promote a level performance curve. This
means that throwing more and more complexity to the scene at a certain
point levels off, adding very little little additional system load
and/or render-time.  Image Courtesy of SplutterFish - Rendered by Jorge Seva & Sergio Miruri
The SplutterFish team is not as small as one might think. We have four software engineers, three whom work full-time on Brazil. Cuneyt Ozdas, one of the world's top 3ds max software engineers, joined us late last year and has been an invaluable asset to the team. Connie Jacobs, who essentially handles the day-to-day business operations, is also a software engineer who writes our networking code. In addition, we also have a very dedicated group of testers, professionals and non-professional artists, with whom we interact daily. These artists come from a variety of backgrounds and fields so we can keep focused on the needs of all artists. The intense interaction and collaborative effort with these talented people are the reason Brazil is such a superior product. I don't
think we ever had doubts that Brazil 1.0 would get completed, but
I wouldn't say it was always easy. During the past year, there were
the usual mental, emotional and even physical strains inherent when
you're up against a tremendous challenge and time constraints. The
support and overwhelmingly positive response to Brazil from our
pre-release clients really kept us going. 
Image Courtesy of SplutterFish - Rendered by Al Barranco
While the weight of the Brazil 1.0 release has lifted, our work has really only just begun. We dedicated ourselves last year towards building a solid rendering system. Laying a proper foundation in any endeavor is the most time-consuming, challenging and critical stage. If the foundation is flawed, any development thereafter is inherently weak and it's really difficult to overcome that weakness with hack fixes and bandaids, not to mention that adding to a flawed foundation jeopardizes the whole structure. Others have attempted to copy our UI and our workflow. These attempts are purely superficial. Looking like Brazil does not mean it has the rigorous standards and quality of Brazil. With our core architecture in place, we are now able to have fun and build upon it rapidly, and with confidence. Brazil r/s is now solid, reliable and ready for supporting the new features we have in mind for 2.0.
As
mentioned earlier, Blur needed some GI and so this code, the precursor
to Brazil, got posted on Blurbeta. It caught on like wildfire. People
started pumping out beautiful images. We started filling up hard
drives from all the images sent to us. When we decided to develop
the code commercially, we continued this open pattern of communicating
with the 3D community and it really aided the development process.
 Image Courtesy of SplutterFish - Rendered by Al Barranco CGA: What features and technologies does Brazil r/s provide specifically to the architectural CG market? I think a good way to approach this question is to share what our pre-release arch viz clients brought to the development of Brazil. In many ways, their needs far exceed what is required in, say, standard broadcast rendering. There were situations they would run into that would reveal even small weaknesses in our design. For example, I remember one client had a little tiny railing in the background of a building interior and thankfully, it forced us to improve upon our anti-aliasing algorithms even more. Large format work was another thing the arch viz community forced us to address. They render these huge stills, something we production guys don't frequently have to deal with. To accommodate them, we added a new VFB to reduce memory usage and we also contracted Dave Humphreys to write a maxscript to offer at least some limited form of distributed rendering capability. The script Brazil offers presently is far from an optimal solution, but it is useful in some of these large architectural still situations. Of course we have plans to do full on, bucket-based distributed rendering which we know they'll really appreciate. One
of the most appealing aspects of Brazil for architects, is the lighting
features (GI, radiosity, skylight, area lighting, etc). We wrote
Brazil to offer the artist a natural, intuitive way to set up lights
(and cameras) and it gives predictable results. It's quite simple--one
sets a light in a room and the room is lit just the way one expects.
It makes the workflow fun and natural, and in my opinion, the image
quality and beauty of the lighting makes Brazil unbeatable. 
Image Courtesy of SplutterFish - Rendered by Alex McLeod
Architectural artists make up a significant portion of our client-base. Their work is very demanding and requires precision and accuracy--things that you really can't (or don't really want to) fake. As I mentioned earlier, working with architects and visualization professionals during last several months of our pre-release sales program pushed our development efforts farther than we might have gone without their input. It actually forced us to add a very talented arch viz guy to our inside team of testers to ensure that the needs of the community were being met. His invaluable contribution helped us address the compatibility issues that came up with popular arch viz 3rd party plug-ins. He also helped us improve our distributed rendering script geared towards large architecture stills and address the lighting and workflow concerns that arch viz artists have in choosing a renderer.
The
next twelve months are going to be very exciting for SplutterFish.
We have already started working on Brazil r/s 2.0. Presently we
are working on other highly desirable features such as true distributed
rendering, displacement, 3D motion blur, etc. Registered owners
of 1.0 will be able to beta test these as we go along. As we have
happily learned, there is nothing more valuable than to have thousands
of people testing the software -- professionals who are 'hammering'
on it in real-world production situations -- to reveal bugs or flaws
in the workflow. This enables us to catch and fix things early when
they are easier to fix. We also recently announced that we are developing
Joe Alter's “Shave and a Haircut,” the acclaimed hair-creation
package, for 3ds max. We expect that will be available towards the
end of the year. Additionally, we are keeping busy by forging key
partnerships to develop Brazil for other platforms.  Image Courtesy of SplutterFish - Rendered by Calibre Digital Pictures
Steve reads every online paper that is written about CG/3D and keeps up with all the game sites. I enjoy visiting sites that show all sorts of short films, animation, and I visit our clients web sites to see all the cool stuff they're doing with Brazil. http://www.neilblevins.com
|
|
|||
