oh wait naw it’s the same image
The ! Makes a lot of diffference damn
really interpreting language correctly and s***
!dream cave painting on stone, finger painting, heavy texture, rendition of Artificial intelligence

that ! was gamechanging
makes sense tho. became an image of a dog covered in blue, paint(ing)
!dream the essence of humanity depicted as a tessellated fractal, concept art, trending on artstation, 4k

all punctuation has an effect I think
I just ran a test it does do something, but it read the punctuations differently
so strange wonder what the difference is
here is the example
my beautiful dark twisted fantasy rodeo Travis Scott astroworld whole lotta red playboi carti metal abstract octane render sharp focus 8k extremely detailed Shot on 70mm F/2.8 high Contrast --seed 1262389093
no punctuation

commas

semicolons

periods

wish stable diffusion's features were better documented
!dream inspirational boomer meme, script font over nature image --seed 666

!dream inspirational boomer meme, script font over nature image! --seed 666

cave painting on stone, finger painting, heavy texture, rendition of Artificial intelligence

cave painting on stone, finger painting, heavy texture, rendition of Artificial intelligence

damn that ai getting sturdy as hell on em
!dream human blood vessels seen under an electron microscope shaped like the Nazca Lines from Peru

!dream the liminal space deep within the human soul, abstract concept art, 4k

I just ran a test it does do something, but it read the punctuations differently
so strange wonder what the difference is
here is the example
my beautiful dark twisted fantasy rodeo Travis Scott astroworld whole lotta red playboi carti metal abstract octane render sharp focus 8k extremely detailed Shot on 70mm F/2.8 high Contrast --seed 1262389093
no punctuation

commas

semicolons

periods

wish stable diffusion's features were better documented
hm I think you’re misunderstanding maybe? They’re not Intentional prompt modifiers, the punctuation in a prompt affects the image the same way a word might. Just to a lesser degree based on what the model learned during training on the laion 5B dataset i think?
it would be a small difference tho like how i imagine using “a” instead of “the” or “an” etc in a long prompt
btw side note you said 6 seconds on a 3060? you mean per image or per inference step?
hm I think you’re misunderstanding maybe? They’re not Intentional prompt modifiers, the punctuation in a prompt affects the image the same way a word might. Just to a lesser degree based on what the model learned during training on the laion 5B dataset i think?
it would be a small difference tho like how i imagine using “a” instead of “the” or “an” etc in a long prompt
btw side note you said 6 seconds on a 3060? you mean per image or per inference step?
what other modifiers you got on this fork