Generative grammars as a form of procedural content generation

At their core, generative grammars are all about the rules and replacements of a language. In our case, that language is game content. Let's look at how we can take the simple concept of word replacement and use it to create more thoughtful and interesting procedurally generated content.
Definition and example: 00:00
Other uses: 03:11
Quest generation: 04:00
The main limitation: 05:48
Where's the code?: 06:41
Text version of this video: shaggydev.com/2022/03/16/gene...

[@BackroomsFR]

Your channel is criminally underrated god damn

[@retro5533]

This looks like it might exactly what I needed for my narrative/character driven RPG

[@TheShaggyDev]

Yeah, I think it has a lot of potential for that kind of game!

[@jvsvogler]

I actually was messing with a Python integration for Godot and the 'markovify' module which is a Markov Chain generator. This kind of approach to create content procedurally can definitely be useful or at the very least super funny haha Your videos are awesome, keep it up!

[@TheShaggyDev]

Thank you! And I like that! I can see how it could be very funny to use a markov chain to generate some ridiculous combinations!

[@monstereugene]

loving these vids. so informative and fun

[@TheShaggyDev]

Thank you! Glad you like them!

[@delqyrus2619]

Pretty interesting approach. Seems a little bit like a one-dimensional approach of the WFC-Algorithm, where you do basically the same, but in a different manner. In the WFC-Algorithm you describe objects and their possible outcomes in some n-dimensional manner. That way you start with some object(like the entrance to a maze) and from there you (randomly) pick one of the possible outcomes. But in your approach you seem to be focused more on the overall-layout, which gets pretty hard in the WFC-Algorithm to do right. So i could think of some combination of those, to make it more easy to create interesting content.

[@TheShaggyDev]

For sure! This method doesn't handle the concept of space very well since it's one-dimensional, but it does do well with overall flow, so I'd imagine using this to drive or maybe filter WFC rules could end up with some really nice results. For example, designing mini-encounters for wave function to build from and then filtering, modifying, or just populating based on the grammars.

[@tumbleweb]

I was JUST looking into this. Same wavelength?

[@TheShaggyDev]

Whoa! Kind of crazy since this seems like a bit more niche than some other pcg methods. Haha.

[@nuhuhbruhbruh]

Ooh unexpected Runestone Keeper. I have a huge love-hate relationship with that game: relatively unique gameplay that's easy to pick up, but outcomes feel so dominated by randomness that once you figure the system out it's just infuriating most of the time.