I just played through Valve’s Portal. Here’s some stuff I noticed.

Austere Design

For a long time, the progression of game graphics has been towards more polygons in view. For decades, our goal was always more detail, more objects, more stuff. Something has changed as of late. We’ve finally reached the point where it is computationally possible to add more detail than is ideal. Just in the last year or so, it has become widely possible to over-detail an environment.

This is a new problem for us. For decades, over-detailing has simply been impossible because of slow hardware. We’re entering a new era, where adding more detail is an actual decision that must not be taken lightly.

Portal showed off this concept perfectly with its extremely austere, but quite effective visual design. The developer commentary indicates that every part of the game went through many, many iterations. What’s interesting is that as the iteration happened, the world became less detailed, not more. And why not? The purpose of this game was not to wow us with graphical details. The purpose of the graphics was twofold:

1. To support the innovative puzzle gameplay

2. To create an impression of a real place; to create suspension of disbelief.

In these two goals, the extremely sparse visual design of Portal succeeded admirably. I wonder what else we could do if we refocus our efforts away from maximizing detail, and into creating environments which support holistic game design goals instead of providing pretty views and ostentatious detail. (Not that there’s anything wrong with pretty views, of course. I’m in the process of executing a pretty hardcore futuristic Hong Kong sunset cityscape right now, actually. You’ll see.)

I see a new visual design challenge in games: attempting perfect execution of fundamentally simple environments.

Say you’re doing a level with an alleyway. Classical design sense says you put as much stuff in the alleyway as possible to make it more interesting. Dumpsters, garbage, boxes, fire escapes, locked doors, lights, broken bottles, cracks in cement, and so on. Of course, if we pushed this as far as possible today, the alleyway would be two feet deep in garbage and there would be so many posters, signs, doors, and windows on the walls that you probably couldn’t even see the bricks. This does not support any kind of gameplay, and it feels ‘gamey’.

So what if, instead of packing detail in, we take a simple, almost completely empty alleyway, and spend all of our graphics computational power on perfectly executing subtle details? In our alley, there is a subtle lumpiness to the concrete, indicating poor quality pouring and multiple patch jobs. There is a tiny plant in a crack, and it waves when the wind gusts through. The wind also forms tiny ripples on the surface of puddles which have formed in the depressions in the lumpy concrete surface. The bricks on either side are marred by layers of sprayed and washed-off graffiti. A brick has a chunk broken off one corner. Torn bits of poster are stuck on here and there, and a few cans float around on the floor. Even the atmosphere moves with the wind: a slight, smoky haze changes in character and speed when the wind gusts through.

You could spend just as much time working on this ’empty’ alleyway as on the junk-filled one I described earlier. The difference is that none of the details in the empty alleyway will jump out at the player. Instead of spending our time adding more stuff, we perfectly execute the stuff that really belongs there. A large part of the design process would be spent thinking through the history of this alleyway, the people who have walked through it, the things it has seen. We build the alleyway to tell all of these stories in its own subtle way. We are willing to accept wildly varying levels of detail, just as occur in real life – a large flat concrete surface may only demand one polygon, while a hanging clothesline may demand thousands of polygons, semi-transparent light-scattering cloth materials, and a complex and expensive physics simulation.

If we can do this successfully, the resulting environments may seem mundane at first glance, but they will achieve an unheard of level of immersion. It’s the difference between adding polygons and creating a strong narrative context for the gameplay.

I’ve digressed somewhat from Portal. Portal removes details but it does not attempt perfect execution of subtlety. I can only imagine what the game would have been like if it had.

And finally, I want to point out this completely awesome and totally hilarious video I found, A Day in the Life of a Turret. It captures the irreverent, absurdist humor of the game perfectly, and adds its own flair.