During Microsoft's GameFest 2008 developer event in Seattle, Gamasutra attended a lecture given by Bungie lead effects artist Steve Scott.
Scott's talk spoke in-depth on the intricacies of the effect artist's field - a relatively emerging profession in games - and shared some tales from the development of Bungie's Halo 3.
A Discrete Creative Discipline
"Effects have evolved into a discrete creative discipline," Scott pointed out. "Treat it as such."
Far from being a side duty, effects are now a major part of a game's development, he argued. "It's really critical to have someone managing the effects process from the beginning to the end of the project," Scott continued.
"You may say you have a small team, and say you don't want a dedicated effects artist, but the days when the art director would spend this nights and weekends cranking out effects are kind of over. Even if you have to share an artist, make sure he has time."
This means that effects development must be planned out, and integrated into every step of development. "If you're going to make hundreds of effects for a game, that's a process challenge," the developer said. "It's easy for effects to be relegated down to the end of the chain. Sometimes there are effects that should be planned for and built into the process from the get-go."
Effects artists must have the support of the engineering team as well. "If it's an effects-heavy feature, maybe you should have the artist pool those resources together and move forward," said Scott. Even if you license an engine or effects tool, "integrating those into your game and making sure they support the gameplay requires engineering time."
Effects Development For Halo 3
Scott summed up the scope of Halo 3's development with respect to effects creation. Over three years of production, the game included over 30 weapons, each with its own muzzle flash and projectile effects; 100 material types, which can require different effects depending on the material; 15 vehicles, each with its own destruction effects and material interactions; and over 40 minutes of real-time cutscenes, which use effects.
This was handled by one effects engineer, two full-time effects artists, and "a little help from our friends."
"I'd like to tell you that we had a really beautiful elegant standalone effects tool to create this content, but we did not," Scott admitted. "Effects are created at Bungie using the same tool we used to manage all in-game data, the same tool engineers use. All said and done, it was a pretty universal generic tool...not an art tool."
That tool had been in use at Bungie for years, he explained, "was not exactly WYSIWYG," and was difficult and time-consuming to upgrade. "I would change that data and then see the result on the Xbox dev kit, and that took about 15 seconds every time I wanted to make a change."
Still, using such a universal shared tool had its benefits as well. "It's constantly maintained. When [everyone] is using it, if there's a bad check-in and it goes down, it won't stay down for long," Scott pointed out. "Any time we'd come up with a custom UI element for the effects side, we were able to integrate that into the tool and suddenly that was propagated for any discipline.
The unified tool also allowed effects to access any game data, since all that data is at the same level. "We could use objects as effect scalers very easily," he said. "We know the speed of a Banshee; we could plug that into our effects data."
The highly modular nature of the effects and the game data was part of what made it possible to manage such a huge effects library, and allowed elements like dynamic lights, sounds, and models to be used as components of effects.
Halo 3 Effect Examples
"Going into Halo 3, we had the advantage of an established style and gameplay," Scott pointed out.
"We sort of understood the scope of the effects tasks, but we had to plan ahead for this generation of hardware and reevaluate some of the fundamental ways we approached building an effect."
The artist gave some specific examples of implementation lessons learned. "we found if you just ramp down the alpha at the end of a [fire] particle's lifespan, it just looks bland; it doesn't have the volume or the life we wanted," he said. "We solved this by writing a shader which; we black-point faded. It crushes in the blacks and makes more interesting shapes. It's richer. We ended up using this pretty much everywhere in Halo 3: water, dirt, explosions. But bitmaps had to be authored with this purpose in mind."
Scott summed up the difficulties of balancing budgetary and time concerns by pointing to how the team approached needing muzzle flashes for 30 different weapons.
"The easy thing to do would have been to take a common bitmap and globally tint it. There's a possibility you could get away with that in the last generation but we would not have got the quality we wanted," he said. "The other option we had was to author dozens of unique muzzle flashes, but that would have completely consumed our effects budget."
The end solution fell somewhere in the middle. "We looked backwards," he explained. "When you would create a set of assets, you would have to palette-ize them back to small colors. We'd create a grey muzzle flash texture along with a 1x256 color texture. This gives much more range and color than just tinting the whole thing."
Scott pointed to that solution as example of what can be achieved when an effects artists collaborates with an engineer on the team to find effective solutions to a problem.
"Every game searches for the perfect fireball," claimed Scott. "In the past, we'd used a flipbook texture - but for the memory that eats up, we wanted to come up with something a little more procedural.
He described the heavily-layered process that go into Halo 3's attractive explosion: "We scale and animate the UVs against a tiling plate. We have a poofy cloud texture which gets paired with an alpha texture which doesn't move. We take a section and we map it to the alpha and scale that up. Underneath that we layer, in the next frame, we take another section, map that to the channel and scale that up. Then you repeat that process until you've got layers and layers scaling and blending into each other. It was used for pretty much every explosion that you see in Halo 3."
Still, Scott expressed some reservations over the full effectiveness of the effect. "I think it's pretty good at giving a sense of secondary motion," he said, "but when you look at a cluster of cards that are all using this in a fireball, just between you and me, I don't think it was as successful at giving a cohesive sense of volume to a fireball."
To illustrate the importance of effects to the visual punch of complex scenes, Scott turned to the death sequence for the game's monstrous Scarab enemy, a huge enemy tank walker.
"This was one of the single biggest effects in the game and became a signature element of Halo 3," he said. Driving his central point home, he rolled video of the sequence without effects - an underwhelming affair.
He listed the many components that went into the final effect - fireballs as the first layer, light beams next, then billowing smoke palette-ized for Covenant technology, then a layer of sparks and debris.
"The key thing here - lots and lots of sparks," Scott stressed. He then spoke on a performance saving trick that takes advantage of a player's tendency to fill in missing detail. "Collision detection is very expensive for your particles. One trick we found in the process is that if you have a spark cloud that has maybe 300 sparks, if only 2% of them collide, once you start seeing a couple of sparks bouncing around in the world, your mind fills in the rest of the detail."
Major Engine Challenges
Though Halo 3 came in under its texture budget (to the delight of the engineers), it had a 256 simultaneous effect limit, and that limit was frequently strained (less to the delight of the engineers).
At one point, a demonstration was being given to a gaming magazine, and as a result of hitting that limit, a Scarab simply didn't draw its effects, making the scene look terrible and causing Scott to realize it was time to tackle the budget issue seriously.
Performance takes a big hit once multiple layers of transparency start overlaying on screen, Scott explained. Part of addressing that problem is simply managing content.
"I would go in and pull out some of those lingering smoke clouds," he said. "Performance and gameplay are king. This is actally when we came up with the idea for the screen effect; we set up a tight radial boundary. It helped smooth out and take the edge off of those LODs."
Another major challenge area was Halo 3's oft-trumpted theater, which allows players to record gameplay demos and play them back from any angle in real-time. "We looked at that and said, 'Oh shit, we've got to do this for real,'" Scott recalled.
Allowances for low-speed replays were added into the review process for effects, meaning the team would sometimes add details that were too fast to notice at normal speed, but could be perceived during a slow-motion replay. Despite the added work that caused, it was "absolutely worth" it, due to the resultant gamer reaction.
Lessons Learned In Halo 3
Scott collected his experiences into four main lessons. First off, effects are an organic process. "It's not a terribly structured environment and takes a special personality to want to do it over and over again," he admitted, going back to his initial point about planning.
"Leave room in your schedule to iterate on tools, respond to the effects artists needs, or resolve unexpected problems. Sometimes the specific solutions to problems can open up cool opportunities across the board - everything is interrelated."
Next, he reiterated the importance of having a dedicated effects engineer if possible. "We had a great engineer, but unfortunately he was assigned to other critical systems on Halo 3," he said.
Third, ensure that effects are integrated into the process early - the sooner that happens, the better the end result will be.
Finally, he once again stressed the importance of planning, as well as te benefits of having a solid production pipeline. "With modular systems, and a solid plan moving into it, I was able to do the entire [Scarab] explosion in one day," he concluded.