The Cost of Content

The most expensive part of game development is content production. This is a bit unintuitive: if you look at a game like Kirby’s Epic Yarn, you immediately see the cool graphical style, the fun game mechanics, and the well-polished charm. The clear advancements Starcraft II made over the first game are its increased graphical fidelity and its modified gameplay elements. The levels are just where the game happens; they’re rarely the interesting part. However, I’d bet you that more person-hours were spent making those levels than on any other single aspect of the game. This is a big part of the appeal of procedural content.

Hand-Made Content

The project I’m working on right now, Beneath the Waves, looks like it’ll be my longest-playing game yet. Most of the hours I’m putting into it are being spent on laying out the world and creating graphic art. Since I’m using Flixel, the basics of the game’s programming are done for me. The game loop, tilemaps, collisions, physics, those are all coded for me. Even the customizations I’ve done — adding sloped collisions to tilemaps, optimizing the rendering of sparse tilemaps, and so on — haven’t taken too long to do. The level design, though; that’s taking forever. I have eight regions, plus an implicit tutorial section. Each of these regions is taking me three to four hours to lay out, and that’s only in a two-dimensional game.

A modern 3D game requires exponentially more content. Every object in the game world needs a model, a texture, and a normal map. Anything that moves needs an animation. Levels need to be laid out in astonishing detail. Individual bits of trash are distributed in a pattern that must not look too repetitive or unnatural. Non-player characters and pickups must be placed, in ways that make sense for the game’s fiction, look good, and work for the purposes of gameplay. Each kind of surface needs footstep sounds, and each kind of room needs an environmental audio filter. Compared to that, the programming is easy… especially since most games use an established engine like the Unreal Engine, just as I use Flixel.

The solo game developer is a level designer first, and a programmer second. Or last.

Procedural Content

But procedural content changes things. Procedural content generation is the technique used by Minecraft (and, to a lesser extent, by Oblivion and Just Cause 2). Instead of a person deciding where each rock and leaf is placed, the designer crafts an algorithm that lets the computer design its own world, often with random numbers influencing the results. Using procedural generation, an expansive world can be created in a fraction of the time. There’s still a lot of work involved designing and perfecting the algorithm, but once it’s done, it can produce a theoretically infinite world.

Why not use it on every project? Well, procedurally generated worlds tend to have a certain sameness to them, for one thing. If you’re trying to craft a story or a progression through the game, you’ll need to do a lot of tweaking to get the world to turn out right or insert a small amount of hand-made content into the procedural world. Another objection is that procedural worlds seem sterile or haphazard. Actually, a well-designed procedural system can feel just as authored as a hand-built one, but that takes a lot of work and care.

The Way Forward

We’ll see more procedural content in games in the future. As players demand larger and more detailed worlds, developers will have to employ procedural generation. Most games will still hand-craft environments on the scale of a room or a building, but procedural generation will fill in the very small (books on a shelf) and the very large (the city around the important buildings). However, there will always be a place for the purely hand-crafted. Smaller games excel at attention to detail and an experience that’s been tweaked on every level by a conscious mind.

Making levels takes so long, and sometimes I’d like to just push a button and have the computer make a level for me. Alas, that’s another game entirely.

14 thoughts on “The Cost of Content

  1. Crytek has a procedural city generator. Each building has an interior, furniture, different decorations. All of which can be blown up, because it’s Crytek. I wonder if they’ll actually put this to good use, though.

    You can also just do procedural content that randomly (or specifically) integrates hand-made content. A game that does this right now is the roguelike, Dungeon Crawl Stone Soup. You’ll run around a procedurally generated dungeon and run into not just hand-made bonus levels (already done in roguelikes for a long time now), but also hand-crafted rooms and even large structures blended into the random levels.

    Not sure why this isn’t done more often. I only really notice it being done in roguelikes, and not even too many of those blend the two concepts in the same level. It’s something I’m particularly interested in.

    1. Yeah, roguelikes were doing things 10 years ago that mainstream games haven’t even begun to touch. I always find that vaguely shocking.

  2. The thing every designer needs to remember is that procedural generation is a tool, not a principle. Like every other tool, it can help, and it can screw things up, depending on how it’s used. It’s not “the way forward” as so many people keep claiming, but an additional ability game designers can now use.

    The best game I’ve played is still Gothic II, where every detail is hand-crafted. It feels the most natural.

    As for content production… I’m not sure that’s entirely true. I think it’s how it *ought* to be, but not necessarily how it *is*. I see a lot of games where the developers clearly put all their effort into improving the technical side of the game, and then ran out of time to actually create enough interesting content. The hollow shell that is Gothic III is a good, if depressing, example.

    1. …which of course still proves that content production takes a lot of time. It’s just that not everyone factors that into their planning.

  3. Procedural generation is definitely useful. If you like the idea, Love, ( is a game made by a single developer with a world generation mechanic and AI building system that’s to rival Minecraft.

    I have implemented designed for procedural content that SHOULD work to create massively complex, interesting, and elaborate (and hopefully fun) levels- given a couple thousand years.

    Obviously, we have a long way to come, but better algorithms should help us make better content.

    1. Love is really interesting, although I found it terribly disorienting when I tried it out briefly. It’s notable that both Minecraft and Love allow for player modification of the landscape, which can help to mitigate the feeling of repetitiveness that’s often present in procedurally generated worlds.

  4. The best kind of games for this kind of generated content would have to be story-light exploration games. The problem is that such games are relatively uncommon these days. In fact, the only good example of such a game I can think really dates me: EA’s Seven Cities of Gold. There is an implicit story–Spanish explorers sailing to find riches in the unexplored world–but it really only shaped the simulation aspects of the game. Despite the game’s many frustrating aspects (like going from a peaceful negotiation with a tribal chief to a whole-sale slaughter with an accidental nudge), there was certainly a thrill in mapping out the new world.

  5. What about user created content? I think that is what makes minecraft so compelling, the sheer volume of well thought and well executed art, puzzles, and mods. It isn’t about creation so much as it is that you know other people will see, as well as the reverse, seeing the work of others. And I suppose that looking through mass volumes of mediocrity, looking for those gems of craftmanship also relates to this topic as well. I have noticed that on such games (with both user and creator produced content) (and by this I mostly mean halo 3) that I spend much more time playing and adding to content produced by players than by the creators, and the file sharing method of content disbursement guaranteed that the very best was widely spread. Creators creations are higher quality, but smaller quantity, procedurally produced creations are lower quality, but infinite. User-creations (I believe) maintain both qualities.

    1. User-created content can be great, but it takes some work to shape it to your vision of a game (if that’s even what you want to do). It’s really a different sort of animal, and one that presents different issues than developer-crafted or procedural content.

  6. oh, by the way, where does the user picture come from? As far as I know, i’ve never seen mine before, and certainly didn’t pick it.

    1. The user picture is generated by picking a chunk of one of several images I selected for the site; I believe you have the right side of Alyx Vance’s face. If you want a different one, you can sign up for a Gravatar.

Comments are closed.