It clearly struggles with faces (every face has weird eye distortion, f-ed up nose, crooked lips, and asymmetrical jawline) but is genuinely impressive when it just has to create a 'form' without giving it much detail. Enjoy!ĪI-generation works best with non-descript prompts & abstract concepts. He is so sad he has two frowns!Įdit: A few mentioned liking Nautilus so I max upscaled that image here. Hope everyone enjoys these as much as I do! Which is your favorite? Personally I am a big fan of the Amumu. One small caveat regarding the lore blurbs: there are various banned words that the AI will not allow image generation from (sex, gore, etc.) If I came across a banned word in the lore I would substitute in a terms and conditions appropriate replacement. Since I was already going to be curating my images from a group of four and upscaling from there I wanted this project to attempt to create what I felt like were believable champions. Some of the initial four had to be re-rolled because the AI would spit out amorphous blobs of color or some truly horrifying versions of Olaf that might get the Disney lawyers after me. Upscaling generally made the images higher quality and added in more fine detail. From there you can generate a new square of four variants or begin upscaling a single image if it was great. Once the four base images were generated for each champ my friends and I would talk about it and pick our favorite. Welcome to the Midjourney AI Champion VU! That seemed to work way better than I expected so after trying a few more I decided to bite the bullet and go all in and see how all 161 champions would turn out.
After a making a handful of those I thought it would be cool to see what the AI would generate from the lore blurb for a champion. I would vaguely describe them and add keywords and get some decent images. Initially I started using the AI to create images based around mine and my friends favorite champions. You can enter a string of words, phrases, or even image links to have the AI bot spit back out a block of four images generated from your input. Hey everyone! Recently I have been goofing around and having fun with an AI image generating Discord bot called Midjourney.
0 kommentar(er)
