- cross-posted to:
- aigen
- cross-posted to:
- aigen
cross-posted from: https://aiparadise.moe/post/13227
Prompts are embedded in the PNGs. I used the Dynamic Prompts and Regional Prompter extensions, and a wildcard setup that didn’t embed because of all the variables. This managed to keep the artists mostly separate from their art.
Models were CyberRealistic 3.1 and/or qgoPromptingReal (which did much better at understanding the prompt).
A bunch of these are slightly different but my basis was:
${age=25yo} ${person_description=!with perky breasts and ({short pixie-cut|shoulder-length|tied up}) __haircolor__ hair} ${context=against a stark empty wall } ${location=at an avant garde art gallery show} ${perspective=orthographic} An ${perspective} photo of a ${age} artist beside a self-portrait ${location}. ADDBASE An ${perspective} photo of an embarassed ${age} artist ${person_description} wearing a contrasting dress and smiling at the viewer ${location}. ADDCOL An ${perspective} photo of a single life-size beautifully expressive pornographic self-portrait of an identical nude ${person_description} {shocked|overwhelmed} by an (extremely intense orgasm:1.2) while (riding:1.1) a (painting of a pussy vibrator:1.2), gasping, moaning, (disheveled:1.2), ${context} ${location}.
I wrote out the prompt with variables in it, so I could change the variables or get them automatically filled in by randomly picking from lists. See https://github.com/adieyal/sd-dynamic-prompts#combinatorial-generation and the part about variables in the extension’s help.
But the real thing that gives the images structure is the regional prompter extension, which interprets the
ADDROW
/ADDCOL
stuff and makes the one thing be on one side of the image and the other thing be on the other side, by doing different parts of the image with different prompts in the same generation.Awesome, thanks for the info! I’ll have to read up on all of this. I haven’t done anything other than using web clients but have been itching to dig deeper. Much appreciated!