Give me a prompt and i'll make an image

Yeah, it just knows, from looking at images of bartenders, that bartenders nowadays in the US (where the dataset was first focused) tend to have them.

This AI -- and really, all non-fictional AI -- are just trying to guess what comes next based on what the data set says is most likely to come back. When drawing hands, it knows that most fingers have another finger beside them, leading to "MOAR FINGERS!" and when drawing a bartender, especially young attractive ones, the data set suggests "more tats." If this were a few years ago, the male bartender would have a fancy waxed mustache, too.
so, here's the thing I'm using multiply checkpoints that have been trained by others

so depending on the check point i use and even the sampling method i can get widely different out puts.
using the Least mean squares (LMS) sampling and using the same check point as the above and the handsome prompt
00067-2594770005.png


Diffusion probabilistic model solve
00068-3717350096.png


and just for giggles an anime based checkpoint
00070-571527980.png
 

log in or register to remove this ad




Have messed around with some web app AI tools, with some success, but often get something that is vaguely cubist...

Would love to.see something better..

Moonlit treehouse village nestled within the canopy of colossal tropical trees, traversed via rope bridges, small gliders, and giant mounted insects. The ground invisible beneath the clouds below.
 

Have messed around with some web app AI tools, with some success, but often get something that is vaguely cubist...

Would love to.see something better..

Moonlit treehouse village nestled within the canopy of colossal tropical trees, traversed via rope bridges, small gliders, and giant mounted insects. The ground invisible beneath the clouds below.
This is what i got using a model that was developed for landscapes @ 🤖 🖼 AI Art Generator: Create Stunning AI Art

TUmTlDz2JTj7A9qnXtaz.jpg
 


This is what i got using a model that was developed for landscapes @ 🤖 🖼 AI Art Generator: Create Stunning AI Art

TUmTlDz2JTj7A9qnXtaz.jpg
Was never able to get darkness to look ok. This looks better even with the multimoons.

My best version of this idea after a number of iterations was probably this..
Screenshot_20230406_144817_Chrome.jpg


I will say that one of the fun/frustrating aspects of trying to get something that looks ok, is the way that it can bubble up things you might not expect.
 

Have messed around with some web app AI tools, with some success, but often get something that is vaguely cubist...

Would love to.see something better..

Moonlit treehouse village nestled within the canopy of colossal tropical trees, traversed via rope bridges, small gliders, and giant mounted insects. The ground invisible beneath the clouds below.
Got inspired by this thread and started to fool around with whatever I could find on the www AI-wise, too.

Neural.Love (category Anime) did this on its first attempt with your exact prompt:

mlt1.jpg


mlt2.jpg


mlt3.jpg
 

Unseen monster hiding in the shadows of a dungeon doorway along a dimly lit corridor. Black and white pen and ink drawing in the style of Advanced Dungeons & Dragons.
Neural.Love (category Prompt Editors) did this in 3 attempts with your prompt:

umadd1.jpg


umadd2.jpg


And this in 9 attempts the prompt modified as follows: Line art in black and white, an unseen monster hiding in the shadows of a dungeon doorway along a dimly lit corridor, black and white pen and ink drawing in the style of Dave Trampier.

umdt1.jpg


umdt2.jpg


umdt3.jpg


umdt4.jpg


umdt5.jpg


Perhaps still too "Oriental" and perhaps more Elmore with some Easley here and there than Trampier, but still.
 

Remove ads

Top