dnd-ai-art-bot VS stablediffusion-infinity

If you're no stranger to deploying bots, managing .env files, etc., you can just go straight to the repo. Grab your API token from Dreamstudio and get after it!

Yup: https://github.com/invictus2010/dnd-ai-art-bot

Stable Diffusion Infinity exists as well, though it's an entirely separate install (unless they've changed it to work with the A1111 api). Similar to the openOutpaint extension, but it's standalone.

The reason I do this is that I break a complex image on photoshop into squares in a grid and I want , after the batch is done with one particular stye, to glue it with stablediffusion-infinity.

Stable Diffusion Infinity, which I recall was one of the earlier ones.

is infinity canvas the same as stable diffusion infinity? if so, very yes; it's great and was a big influence on openOutpaint, but i didn't really want to have a second python venv full of very similar libraries so here we are :)

I generated this at 576x1024 resolution, using a tall format so I could post on TikTok and other social media. I had really like the worms that had been one part of an animation I made for Halloween, so I decided to use the same prompt again. I generated several scenes in Stable Diffusion. For most of the animation, I went through and did a Photoshop action that enlarged the image but kept the canvas size the same, then ran that result through img2img at 0.1 to make only a slightly different frame. This kept zooming in for each frame, but to transition into the next scene, I needed to start zoomed out, so I used outpainting with StableDiffusion-Infinity to expand the scenes, with more trees and worms and things around the central image. SD-Infinity worked pretty well, but it was a pain to install (I had to follow a tutorial video from Nerdy Rodent on youtube on how to install it) and took a few steps to run each time. So I also tried using painthua.com which gave similar results but was easier to set-up and run, for the most part. Once I had outpainted the 576x1024 images into bigger 2048x4096 images, I had room to work backwards and zoom out from the scenes I was transitioning to, with an img2img change at each frame, so that once a new scene was revealed we could zoom into it.

Similar editing, outpainting, erasing, inpainting, etc. is already available to use locally in SableDiffusion-Infinity. I have it running on my Windows machine, and between that and Automatic1111's web GUI I don't have any reason to use a paid service like DALL-E for anything. I followed this tutorial video for how to get it installed: https://www.youtube.com/watch?v=93Nx22HkHoI

Github

The best thing would be to use interrogator to extract prompt from image. If you run this SD version -> https://github.com/lkwq007/stablediffusion-infinity the interrogator tool is present there. I cant remember what was the exact prompt but one of the examples that interrogator did was: "a close up of a clay object, a raytraced image by Bruce Nauman, cg society contest winner, neoplasticism, biomorphic, 8k 3d, volumetric lighting"

Made with Stable Diffusion Infinity. Would share the prompt, but there wasn't any one prompt - there were a lot of different ones to build this thing up piecemeal & iteratively refine.