entrep.lynxQ: I tested this and compared the outputs to those of Kittl.
Even though you're both using stable diffusion, Kittl's renders are more spot on to the text description. And they're also better looking. I tried it several times for 2 prompts, and always Kittl's are better:
* prompt 1: (logo with) 1 palm-up hand underneath 2 human stick figures; muted colors with blue chroma; highly detailed; simple, smooth and clean vector illustration; no jagged lines
* prompt 2: (stock image of) a leader having discussion with 2 subordinates; office setting; desks and computers in the background
In both cases, your version often misses the count (e.g. number of people the leader is discussing w/ OR number of human stick figures). Furthermore, in your outputs the renders often have mutations (twisted hand, crooked keys in keyboard, etc).
Could you shed some light on why that is? Is it just a training issue. If so, when can we expect your renders to be at par w/ those of Kittl?
Bugrahan_StockimgAI
May 15, 2024A: I tried your prompts on the platform you wrote and the results were really funny. We are surprised that you are not satisfied with our service, we are improving our models every day. Thank you for your feedback.
Thank you for this post. It's extremely helpful. I have 2-3 other AI image-generation sites/plans, etc. and none are perfect. This one, at least, claims that you can use the generated images commercially, which about 98% of the others do not say--or they say that you're on your own.
I appreciate your critique. Probably saved me from a mistake--Hitch