In this months edition of The Digital Eye, A2M CTO Martin Walker describes how this Montreal-based videogame developer has upped its 3D artistry with a new workflow and pipeline.
When we first approached art asset creation for next-gen games, it didnt initially seem all that different from any other game console evolution -- its always been a question of quantity and precision of assets that need to be produced. A2M has been around for 15 years, so weve adjusted and readjusted our pipelines as well as workflow processes to accommodate every major progression in gaming since 1992. We have released more than 40 titles for Game Boy Advance, Nintendo DS, PlayStation 2, Xbox, Game Cube, PSP, Xbox 360, PlayStation 3, Wii and PC.
When I look at how our pipelines have evolved, however, I think back to what was then a revolutionary platform -- the Sony PlayStation. During development for PlayStation games, we had teams of approximately 15 artists, designers and programmers with three to four technical engineers. When we started developing titles for PS2, our average project requirement went up to 55 artists, designers and programmers, with a technology team of 20 engineers. Now with next-gen platforms, were seeing asset growth on what we consider an order of magnitude, but we dont necessarily have correlating extensions of budgets or timelines. So our greatest challenge now is to create economies of scale and pipeline efficiencies to accommodate project teams that are currently averaging 100-120 artists, designers and programmers, with an additional core of 30 TDs, programmers and engineers. What were faced with is all about complexity of programming for the console and quantity as well as quality of art assets expected to achieve near-life visuals.
Everyone developing games can create beautiful graphics on next-gen consoles. I remember E3 a few years ago when the consoles first started hitting the market and every game was filled with shiny-lit reflective objects (comically, it seemed that every title had metal armor to prove that they could write complex shaders). Not that weve necessarily achieved the ultimate level of graphics possible, but were getting to a point where its harder to differentiate developer styles amongst different games. So while a lot of titles are striving for realism, we at A2M are opting for looks that are more stylized with a visual language that stands out from the rest of the pack.
In order to accomplish filmic visual looks, the art assets are becoming a greater challenge, and so were forced to focus on productivity boosters for our artists via pipeline efficiencies that enable them to do more work in less time. Just to present a tangible example, with PS1 games we averaged between 500 to 1,000 polygons for main characters in each title. With PS2 games, those numbers went up to around 2,500 polygons with several pushing 4,000-5,000. Now, with PS3, were looking at upwards of 10,000 polygons per character. Additionally, with the level of detail on the PS3, we have to generate light maps, normal maps and bump maps, which require very high-resolution models, read millions of polygons, that we then map onto some texture mechanism -- thats why were reaching such high poly counts. As a result, whats happening now in games is that artists who use to be vertex pushers, manipulating vertices one at a time, now face millions of vertices requiring tools to help them do mass manipulations. The market is responding with tools like ZBrush and Mudbox to sculpt characters, but then you still have to create light maps efficiently so that artists can get near realtime results. Again, were up against the constant drive to allow artists to do more in less time with less effort.
Faced with this mandate, how do we boost productivity? With previous generation consoles, all of our focus was on game engine features that made games better and more comprehensive than the competition. We didnt worry about the tool chain as much because our artists could be productive enough to develop data as was required by those platforms. Now, with next-gen, whats key is the tool chain, because its all about what we can provide artists to make their jobs more productive. Yes, were still pushing the game engine, but on the engine side we havent added as many new programmers, while on the tool chain side, weve doubled if not tripled our efforts at boosting productivity.
Meanwhile, a majority of our proprietary productivity tools are either plug-ins or customized add-ons to Autodesk Maya assisted by some highly specialized applications. Theres a lot in Maya that we benefit from, but a big part of our effort goes toward building tools that are tailored to our projects. Some of the custom Maya tools weve written include overriding Mayas renderer with the results of our internal shader engine. Shaders are little programs that represent how images are displayed on screen, so they require both a deep intrinsic artistic sense but also the technical know-how of a code writer. In order to get around this, we have built a customized visual tool where we have removed most of the math and programming involved in building a shader. Its a tool that artists can understand and use to create original shaders within Maya, their everyday tool, making them more independent and less reliant on programmers. Maya offers us the flexibility to build these types of very powerful workflow tools that are also artist-friendly and well integrated into their everyday production pipeline.
We have been using Maya for around seven years when we first started working with PowerAnimator (the precursor to Maya) on our PS1 development pipeline. When we moved to develop for PS2 and Xbox, we refurbished our 3D pipeline and standardized on Maya for all of our art asset creation needs. We also use Autodesk MotionBuider as part of our motion capture pipeline and have an animation rig that weve built inside of Maya that integrates beautifully to give us the kind of fluid animation that we need for next-gen consoles. Maya is very malleable in the sense that we can customize tools to suit our specific needs, and the flexible plug-in architecture and scripting language permits us to embed even more productivity boosters inside of Maya. Ironically, we bought Maya when it was the most expensive option on the market, but chose it because it seemed to be the most extensible and flexible 3D solution and we havent had any regrets about that decision to this day.
Given our size, we can benefit from economies of scale, and building our own tools and technologies into our Maya pipeline really helps us work as efficiently as possible. This combination of off the shelf and custom tools has been a great recipe for us. We can develop a workflow that specifically suits our needs given that we have hundreds of artists and programmers generating thousands of assets on a daily basis.
We are currently working on our most ambitious project to date -- Wet, a new adventure game for Xbox 360 that will be released by Sierra Ent. late in 2008. The technology to create the game has been in development for more than two years now with a team of 30 programmers. The production team is projected to go well over 100 artists, designers and programmers. This game reinforced the fact that with next-gen titles productivity becomes a huge priority for developers, while dealing with multiprocessors on the new consoles a close second. The game employs a very stylized rendering technique, and in early reviews it has been compared to the look of films such as Robert Rodriguez Grindhouse and Sin City. We decided to push our shader tools to the max and let our artists explore a totally unique rendering genre. Taking cues from filmic techniques and applying them to 3D rendering gives the game a very edgy look that we think will make it stand out among other titles.
All this time and energy on behalf of the artists and programmers on Wet creates a glut of digital assets. In preparation for what will be a large demand on storage for our next-gen projects, we changed our whole storage infrastructure from three to 15 terabytes. When we had teams of 15, the amount of data was manageable. With 50, it was scary, but with 150 people creating assets on a daily basis, it can become a logistical nightmare if you havent thought about it throughout the process. We made a name for ourselves with high profile tiles such as Monsters, Inc. for Disney in 2001, Scaler in 2004, the Kim Possible game franchise and three movie tie-in games last year, including Monster House, The Ant Bully and Happy Feet. Coming in 2008 we also have Iron Man along with some other very exciting next-generation titles that will continue to break new ground and push the demands of our pipeline. These demands increase the potential for logistical overlaps and thats why so much emphasis at A2M -- and with next-gen development in particular -- has been put on consistently updating, refining and customizing our pipeline to create a working environment that is both artist friendly and incredibly productive.
Martin Walker is the CTO of A2M, the largest independent videogame developer in Canada. The company employs more than 430 people with offices in Montreal and a small satellite studio in Hong Kong. A2M works in partnership with top tier videogame publishers and licensors such as Sony, Disney, Warner Bros., Universal, Marvel, Ubisoft, Midway, THQ, Take2, Activision, Vivendi, Sega, LucasArts and EA.
