Show newer

telling stable diffusion to render photos of liminal spaces with different sampling methods, first try for each.

stable diffusion seems to like black and white quite a lot. perhaps the model i'm using, since it's trained from danbooru, has a lot of line drawings or sketches that are black and white it was trained on.

it does seem to know generally what a liminal space is, which is kind of cool considering these ones don't exist.

seems like nighttime shots are a little harder for it. i noticed it was definitely harder on the upscaling. either loss of detail or visual artifacting so i went with the former.

"photograph, red hair, braids, girl, 19 years old, forest background, blue eyes, cyborg ears, night time, nude, smirking"

Show thread

stable diffusion is the first AI i've tried that's been able to get damaris looking even somewhat close. for some reason, it doesn't recognize that the knit sweater is supposed to be blue and makes it green instead. either way, it's cool seeing what she might look like if she was real.

"photograph, red hair, braids, teenage girl, forest background, blue eyes, bell collar, blue knit sweater, cyborg ears"

two of my players' d&d characters in stable diffusion. i like the results of the second character way more.

"photograph, man, 37 years old, short brown hair, silver armor, purple robes, muscular, blue eyes, holding sword made of vines, autumn"

"photograph, bald man, 30 years old, auburn beard, viking outfit, muscular, blue eyes, holding sword made of lightning, snowing"

gameliberty stable diffusion generations (dpm2 a).

"realistic photograph, crusader knight, holding video game case, usa flag background"

QUNA CONCERTS ARE BACK FOR A LIMITED TIME

THIS IS NOT A DRILL

Show older
Game Liberty Mastodon

Mainly gaming/nerd instance for people who value free speech. Everyone is welcome.