Stable Diffusion
Posted: Tue Nov 22, 2022 7:05 am
If anyone's seen DALL-E or other online art generators powered by AI then they're probably familiar with this. Basically a local version of those websites that have an AI generate photos based on prompts for you. Saw it on /b/ last night, of course they use it to generate porn of the degenerate variety but I've been having a lot of fun creating ridiculous stuff with it. If anyone wants to create shit with it, here's a link to the github for the project. You will need git and Python 3 (a specific version I believe). Other than that it installs its own dependencies. You will also need a model to base your images off of, theres tons online, and an official one, too. I've been using the official one but as I understand it there's some with anime bias, etc. I'll post a few sample images at the bottom.
As a disclaimer, I'm not really good at making everything look realistic yet, but some people are really talented with making prompts that generate photo realistic images. In particular the AI is kinda sucky at making hands, eyes, and for some odd reason guns? Using the negative prompts has really helped a lot.
Note a decent graphics card is also required, I used a 2060 super with no problems. I believe 4GB VRAM and above is good for it, and there are command line options to set if you're not working with much VRAM. If you have any really good pics please post them, and if you need help with prompts go to lexica.art and look up Stable Diffusion tutorials.
https://github.com/AUTOMATIC1111/stable-diffusion-webui
As a disclaimer, I'm not really good at making everything look realistic yet, but some people are really talented with making prompts that generate photo realistic images. In particular the AI is kinda sucky at making hands, eyes, and for some odd reason guns? Using the negative prompts has really helped a lot.
Note a decent graphics card is also required, I used a 2060 super with no problems. I believe 4GB VRAM and above is good for it, and there are command line options to set if you're not working with much VRAM. If you have any really good pics please post them, and if you need help with prompts go to lexica.art and look up Stable Diffusion tutorials.
https://github.com/AUTOMATIC1111/stable-diffusion-webui