Read on Twitter
So, I want to talk about procedural generation. Simple concepts, you probably already know them, but it's good to stay grounded.
Basically, I'm thinking of making a video about of this.
Basically, I'm thinking of making a video about of this.
In addition to creating a loooot of procedural content tools and prototypes, I also play a lot of indie games with procedural content.
Now, we need to make a distinction between a procedural art tool and procedural in-game content. The important difference is a human editor.
Now, we need to make a distinction between a procedural art tool and procedural in-game content. The important difference is a human editor.
Today we're talking about procedural in-game content. Something the player will see, but the devs will never be able to veto or massage or even know exists.
They can plan it, but they simply aren't around when it happens.
They can plan it, but they simply aren't around when it happens.
This means the human editor has to do all the editing before anything even really happens. The human has to set up the procedure in advance.
You know how much you tweak a procedural tool's output? Look at it, change it, undo, mutate, tweak, retry...
That, but without seeing it.
You know how much you tweak a procedural tool's output? Look at it, change it, undo, mutate, tweak, retry...
That, but without seeing it.
Most procedural content in indie games ends up feeling aimless. And it is. Because there's nobody around to aim it.
This is why a common recommendation is to use proc content for content "the player won't care about". If it's bland, it's fine, it's supposed to be.
This is why a common recommendation is to use proc content for content "the player won't care about". If it's bland, it's fine, it's supposed to be.
But that's not really ideal, is it.
Is it possible to make proc gen content that matters? That pops? That flows and is fun?
Well... probably not to the extent of authored content, but let's talk about it a little.
Is it possible to make proc gen content that matters? That pops? That flows and is fun?
Well... probably not to the extent of authored content, but let's talk about it a little.
The question isn't "how do I generate a billion unique planets" or "how do I generate every person in London" or "how do I make a dungeon".
The question is "what is the player doing? What's fun?"
The question is "what is the player doing? What's fun?"
What part of exploring the planets will be fun? What part of the population of London will be fun? What part of the dungeon crawl?
This simple question will tell you the kind of content you need to make.
This simple question will tell you the kind of content you need to make.
"You can hack and recruit anybody in London, they're all unique and-"
Why does it matter that they're unique?
What part of hacking or recruiting them varies? That should be the part you focus on.
Their visual distinction and names are just background fluff.
Why does it matter that they're unique?
What part of hacking or recruiting them varies? That should be the part you focus on.
Their visual distinction and names are just background fluff.
We've known about this forever. Hell, Rogue knows about it: levels are often constructed out of cornerstone challenges or structures on a regular beat.
That was in NINETEEN EIGHTY.
Rogue is FORTY YEARS OLD. And it knew this shit.
That was in NINETEEN EIGHTY.
Rogue is FORTY YEARS OLD. And it knew this shit.
Much more recently, Diablo also understood this, assembling dungeons out of large prefab tiles that offer specific authored challenges with specific timing.
What's fun about Rogue? About Diablo?
It's the roll. The flow. The strain and push.
Not "the fighting".
What's fun about Rogue? About Diablo?
It's the roll. The flow. The strain and push.
Not "the fighting".
When the Rogue and Diablo devs asked themselves "what's fun about these dungeons?" they were smart enough to answer "the flow", not "the fighting".
The fighting is not part of the dungeon design. It's a different part of the game.
These questions are not so easy, are they?
The fighting is not part of the dungeon design. It's a different part of the game.
These questions are not so easy, are they?
(Obviously the fighting is technically part of the dungeon design, but that's like saying your clothes are part of your body. It's kinda not the point.)
So when we ask "what's fun about exploring the random planets?" we can't answer "fighting". We can't answer "finding things".
Those are *on* the planet, but they're not *the planet*.
Do you see?
Those are *on* the planet, but they're not *the planet*.
Do you see?
We need to generate the planet's overall flow. What makes the planet's heart beat? What makes the player's eyes get wider and wider and makes them want to see the next part of our planet?
We have a lot of options, but any answer will redefine our game.
For example, we could go for a more literal flow - make every planet something you race across, down into the canyons cut by ancient space lasers or something.
For example, we could go for a more literal flow - make every planet something you race across, down into the canyons cut by ancient space lasers or something.
Or we could make it a desperate search for rare resources as our air and health run low.
But once we choose it, whatever it is, the challenge becomes how to procedurally create content that focuses that play. Creates it in a way that has medium- and long-term pull.
But once we choose it, whatever it is, the challenge becomes how to procedurally create content that focuses that play. Creates it in a way that has medium- and long-term pull.
Every step of the way, we risk losing that focus and making procedural content just because we can.
No Man's Sky is an example of that. They answered the question as we did - desperate search for resources - but their procedures don't generate with that in mind.
No Man's Sky is an example of that. They answered the question as we did - desperate search for resources - but their procedures don't generate with that in mind.
There's a ton of approaches we could try for any given kind of content. But as we approach each technique, we need to ask:
Will this create the play the player needs?
If not... put it back. Look for another approach.
Will this create the play the player needs?
If not... put it back. Look for another approach.
It's so easy to fall in love with a particular approach to procedural content generation.
Good way to end up with a big ol' pile of procedural oatmeal.
... And that's that.
Basic stuff, as I said.
Good way to end up with a big ol' pile of procedural oatmeal.
... And that's that.
Basic stuff, as I said.
... I also want to talk about using procedural content as an isolation cell for the player, or as an adjunct to another form of unfolding play, but I think this thread's already gone on plenty long.
