I don’t want to design levels any more
Something has been bugging me for a while now: Level Designers are not very efficient.
This creeping realization is a problem for me, professionally, because of course: I am a Level Designer.
I think that Level Designers should get out of the business of making gameplay spaces. I want to design systems that make gameplay spaces.
Auto Dungeons and Systemic Worlds = Inevitability
Systemic content isn't a new thing: As a kid I played Rogue, an ASCII dungeon crawler where every room and encounter was systemically generated. More recently, Torchlight did generated dungeons pretty well.
What is new is the processing power and tools capabilities to make systemic content that could soon overlap our quality expectations for “triple A games”. While few people would argue that No Man’s Sky will be a blockbuster-grade console product, it’s certainly extremely exciting. Maybe the quality of each individual explore-able planet won’t be ultra high but there are a lot of planets.
One of the phrases I hear tumbling around my workplace these days is “we’re in an arms race for content”, and it’s a smart statement. As a guy who makes games content for a living, the message is pretty clear: We need to create more hours of high-quality player experience for less money.
And, as you have guessed by now: I think systemic content is going to be the answer. I can see several different ways to form a relationship between content-generating systems and AAA game production.
**Author’s Note: I’m NOT saying all levels should be purely systemically generated. I AM saying that seeing level data as ‘a system of moving parts’, and using systems in tools and game design to manipulate those parts, is going to produce great player experiences faster. **
Make It Faster: Procedurally-Accelerated Level Design
People are already doing cool things with tools for Level Designers to handle some of the ‘busywork’ that level designers and environment artists do. Houdini is a standout example, and works well with Unity and UDK.
In many cases the systemic generation here is really being the Level Designer’s assistant; it handles things like ‘okay, let me connect that hallway for you’, and ‘I’ll make sure that those stairs actually reach that platform’. This allows a level designer to work very quickly, and for as long as the level metadata remains connected to the procedural tool, the level designer can iterate rapidly.
For example, if the level designer was working with a procedural tool instead of a modelling suite and wanted to move a platform in a large objective area 5 meters higher, it would be nice if the procedural tools took care of adjusting the staircases, and moving any assets associated with the high platform.
Make It Play Differently: Procedure-Friendly Data
The most basic way to envision a level that plays in a variety of ways is to imagine a day/night cycle in a game world. Players have a different experience of the space, and different things can happen. This idea can be driven quite far however, so long as level metadata is constructed in intelligent ways that support player experience systems.
Traditional Level Data: Roads are modeled onto the terrain, and have a fixed texture and shape like any traditional 3D asset.
Procedural Level Data: Roads are a spline in the game world, and will inherit an appearance (dirt road, tarmac, aging concrete) at load time. This allows for interaction with other systems and gameplay events, as well: (road could be wet if the weather system says there is rain, or the road could be paved with gold bricks once the player has slain the evil wizard).
Tools Evolve But Design Must Drive the Process
Perhaps the most exciting promise for designers is to let software handle much of the variety-generation and connecting-of-dots that we now do by hand. As a Level Designer in triple-A games, I’ve spent a huge amount of time placing level metadata. I’ve spent man-months placing and tweaking search and patrol nodes in levels, for example. That’s a lot of time that I could have spent designing player experience.
For example, let’s say that instead of hand-placing spawners to create ambush points throughout an open world game, I worked instead on a system that would instance ambush spawners according to a set of rules that I design and iterate on? If I spent two months working on my system and its rules, I’d bet I could start to get a level of quality that begins to resemble the kind of precision that experienced Level Designers now produce by hand. And, my content would be scalable, allowing me to manufacture more fun in a shorter time.
Ultimately, the only reason not to do systemic content is the initial pain and suffering for Level Design and Level Art. Experienced designers and artists are confident about the quality they can produce because they do it all by hand, and they will crunch to polish and perfect every corner of their work. This hand crafting produces very high quality work but is intensely expensive.
I’m excited about a future in which I spend all of that same time polishing, tuning, and tweaking procedural systems instead of just content. This will be an utterly different way to work, and probably difficult to start… but I’m convinced this is future of Level Design. Spaces will be created by systems, and systems will populate spaces.