All made with stable diffusion on my old ass PC with an RTX 2060 6GB
He has the power to know how many coats equals one yard of linen
Marx on his way to make the Black Book 100 million higher.
you didnt need to do use an AI for images of captain communism, you could've just taken screenshots of Love Sausage
I found the answer: https://www.factcheck.org/2022/08/is-it-really-ALMOST-a-square-or-did-you-typo-in-the-resolution/
It’s a 10x10 grid and each image is 521x512. I think squares don’t always look squarish on screens because our screens are always rectangles. There’s lots of optical illusions like that
How do you make these things? I've always wanted to try messing around with AI.
For these I generated them on my home machine with a modified version of the CompViz/StableDiffusion code and the model they just released, once it's working it's pretty easy to use from the command line, you basically just tell it the prompt you want and settings like # of steps, size of image, etc. Currently you need an nvidia GPU to make it work, and "enough" VRAM. There are also some GUI based interfaces for it which I haven't tried, and a bunch of online services where you can do it even more easily, both paid and free, and surely there will be more to come. I'd recommend trying it out online then if you get hooked and have the hardware you can take the time to set it up locally, it's a bit of a pain, but worth it cause you can just let it run overnight and make hundreds of images.
Is there a guide somewhere?
I used to play a AI generated text game(AI Dungeon) which requires 6GB of VRAM so I just assumed generating images must require dozens of GBs.
Nevermind, found a youtube tutorial and got it working. Had to look through a few, most of them just tell you to use some third party app. Then there's this one with a guy in a green bodysuit actually tells you how to download everything and how to run it.
Almost certainly if you've got the GPU, you may need to download/update Nvidia CUDA drivers
https://huggingface.co/spaces/stabilityai/stable-diffusion
If you have a graphics card with at least like six gigs of video ram or want to try setting up a Google colab, you can find a guide on the "stable diffusion" subreddit.
My card should be powerful enough to run it but there's some issue w/ half precision on 16XX cards so I don't have the throughput to make anything but cloudy blobs
Nvidia gimped the compute capability of 16xx intentionally :torvalds-nvidia:
"this pic goes hard feel free to screenshot" - Karl Marx
Full prompt: Movie screenshot of Karl Marx as captain communism, Marvel_Studios, detailed, glorious, cinematic, 8k uhd
I tried making a bunch of pictures of Lenin in various zany situations, but the model sucks at drawing Lenin for some reason.
Other than that it's great fun. You can try it for free on huggingface.
Pretty sure you can download the whole thing too if you have a few gigs to spare
Yeah, I know, but my gpu is like ten years old. I could probably try running it on Google colab...
