The key to predictability seems to be in looking at the dataset before constructing the prompt, to determine the keywords that describe what you want in a most precise way.
Was the prompt: "Booba overflowing with milky goodness as if : Greg Rutkowski had personally spent 1000 hours of artistic talent specifically on how much they overflow"
the fuck outta here son. my question is only for the big dogs
Nolan Roberts
The first 100 search results on google.
Ian Walker
>the fuck outta here son. my question is only for the big dogs I'm serious, Greg Rutkowski the prompt has now created more work and better work than Greg Rutkowski the man will in his entire life
It's not a stretch to argue the prompt is now more prolific and important than the artist.
weird, can someone generate this and tell me which of these images it matches? The one on the left is the one I got on the other machine, the one on the right is the one I get now on my local one.
"oil painting of a rugged wanderer in a trenchcoat, leaning on his cane, standing in front of a huge looming ornate tower, casting shadow on everything nearby, non-euclidean architecture, the tower incredibly distant from the man but still huge, fog clings to the tower, the tower is punctuated by sparse foliage nearby, highly detailed, masterwork by Caspar David Friedrich" Steps: 50, Sampler: k_lms, CFG scale: 7.5, Seed: 8705882, Width: 512, Height: 640
Is it worth teaching my friends code or just let them wait until it’s retarded down into something they can understand like the retard proof GUI
Ryder Reed
in the previous thread, user linked his krita plugin >So yesterday I wrote a simple Krita plugin. It's just a weekend project, but I want you guys to check it out. It's pretty fun to work with this model from an actual image editor. github.com/sddebz/stable-diffusion-krita-plugin >The repo README has demo video.
>Is it worth teaching my friends code of course, always, regardless of the %current_fad%
Ian Morgan
>Show me one image that is better than any of his work I'm not going to, instead I'm just gonna point out no actual adult considers work like this art. It's manchild shit.
Can I just use the AI to colour my drawings? I though I would just redraw what the text2img spits out and use the img2img to colour it and give it shadows and shit.
gaussian blur in any tool you want, advanced editor is experimental, it's excellent work just has limitations to what people are expecting to be able to do when they want a very specific thing