The lowest level of procedural generation is random combination. You have one pile of one thing and one pile of another thing, and then you put them together, like those dress-up games - five different coats, five different necklaces, five different skirts, five different eye-colours. The method is very straightforward, and results can be counted, like in the above example, 600 different outcomes! Wow!
The problem is that nothing new is really created. Perhaps coat 1 matches particularly well with skirt 4, but you cannot call this a novel result. It is not a thing on its own, it does not amount to more than the sum of its parts.
Let's talk about memes.
Not genes, not right now, just the colloquial use of the word. A meme is a combination of an image and a text. These two are related, sometimes quite loosely. Humour is derived from juxtaposition between fitting yet disparate elements, combined arbitrarily. I say arbitrarily, because one text can strangely often be combined with another. There are rules, sure, but they are much more loose.
Let's try to procedurally generate memes. One day, I might do this a bit more seriously, but in honour of the source material, I will make this generator in the most low-effort fashion possible.
But it's not that simple. Not only does it randomize text, it also has a chance to replace one face with another. Here, it seems to change who from the picture is speaking in the bottom text. The bottom text is... not great? Most often, I would take full captions, but sometimes I decided to cut up sentences for better randomization. This is one of the consequences.
This man now gives some sage advice and is paired with a new face that randomly fits the new caption.
Sometimes, especially when there are several captions, it veers off into a territory of nonsense. Still, with a bit of effort, you can piece together some semblance of a story.
Here, I have left several captions unchanged, as they are necessary for the setup of the joke. This is one way to avoid the above problem.
This one is great. It feels like it is referencing something you just do not get, and that lingering feeling of not fitting in is, after all, the purpose of all memes.
Add some citations marks to make it a dialogue and you can almost feel the tension:
Today:
Him: "WWE is just anime for rednecks"
Me: "of course see you later"
We saw each other at McDonald's during lunch.
So if I were to work more on this, that might be something to consider, actually considering sentence structure.
Apparently, both people who are born and raised in their home state and some biblical dude named Galatians keep telling others that they are at the store.
This really is just a bot spitting out random words, but then the final caption enchants it, once again, with the idea that all of it is just a reference you do not get.
My reaction too.
I love this. Putting Thanos' face there, and the "Ya know what? Good post OP." that just works in every single placement, it's just great.
As I said.
I feel that way too.
While working on the project, I thought that it actually was working out. I had grand plans to continue onto randomized captions using a Markov-chain like approach, but then, as I added more and more content, I realised slowly that -
The reason this works at all is that most text in a meme is one-sentence punch-lines, like, "Yes! I actually lost a girlfriend over this very argument" or one-sentence set-ups like "Me after drinking approximately one bottle of water". Putting them together creates something that looks so much like a joke that you almost find it funny by habit.
And sometimes it is funny, but only because the original pieces were funny. Combining them randomly really does not add anything, or if it does, it is a one in a thousand chance.
Evaluation:
This has all been a colossal waste of time. Considering that I wanted to work with memes, I must then say, it went as well as it could.
The problem is that nothing new is really created. Perhaps coat 1 matches particularly well with skirt 4, but you cannot call this a novel result. It is not a thing on its own, it does not amount to more than the sum of its parts.
Let's talk about memes.
Not genes, not right now, just the colloquial use of the word. A meme is a combination of an image and a text. These two are related, sometimes quite loosely. Humour is derived from juxtaposition between fitting yet disparate elements, combined arbitrarily. I say arbitrarily, because one text can strangely often be combined with another. There are rules, sure, but they are much more loose.
Let's try to procedurally generate memes. One day, I might do this a bit more seriously, but in honour of the source material, I will make this generator in the most low-effort fashion possible.
Let's start simply. We have an image and a caption. Of course, the original post had a different caption, but it has been replaced with another caption from another post of approximately the same length and... that's it. I do not check for semantic information or anything like that. Somehow, it tends to magically work out.
Today:
Him: "WWE is just anime for rednecks"
Me: "of course see you later"
We saw each other at McDonald's during lunch.
So if I were to work more on this, that might be something to consider, actually considering sentence structure.
The reason this works at all is that most text in a meme is one-sentence punch-lines, like, "Yes! I actually lost a girlfriend over this very argument" or one-sentence set-ups like "Me after drinking approximately one bottle of water". Putting them together creates something that looks so much like a joke that you almost find it funny by habit.
And sometimes it is funny, but only because the original pieces were funny. Combining them randomly really does not add anything, or if it does, it is a one in a thousand chance.
Evaluation:
This has all been a colossal waste of time. Considering that I wanted to work with memes, I must then say, it went as well as it could.
Comments
Post a Comment