Given that it's seemingly art someone trying to make ppl pay money for on patreon, you'd hope it involves some "private" work and that they aren't just ripping off other creators as well as the ppl stupid enough to pay for "art" that's not unique, easily duplicated and "sold" in unlimited amounts...
You're better off finding your own "style" though, "art", as much as it is, shouldn't be copied/replicated, you should rather find your own take on it.
You could look at finding some lighting loras to "dim" things, or some kind of bluring, seems like you're getting too much/bright colors so some of the weighting might be causing issues and with the lora you're getting oversaturating.
== Updated ==
looking at the lora on civitai and one of the creators images the model perfectWorld_perfectWorldBakedVAE seems like a close match for the background, it's on huggingface.
I agree completely. It's impossible to know for sure how any image has been generated without the png info. Finding your own style is part of the fun, however it's a good learning exercise to try replicate other's images first. I would try the most popular typically used checkpoints wich ever theme or style one tries to replicate.
Besides Perfect world, there is the different versions of
AbyssOrangeMix, ReV Animated, NeverEnding Dream (NED), Kotosmix and many more. Try the samus loras on civit,
You must be registered to see the links
,
You must be registered to see the links
etc.
Try the very popular upscaler that many who make this style uses
You must be registered to see the links
.
When using hiresfix always use hiressteps=2x samplesteps, meaning if 20samplesteps use 40hiressteps.
I always recomend 20-30 samplesteps and 40-60 hiressteps in order to conserve the image composition.
Thanks to
devilkkw for the tip.
Try to use Restore Faces GFPGAN and/or postprocessing GFPGAN. About the prompt. Write in positive the things you wish to see in your image, write in negative the things you don't wish to see..
Also you can use negative to affect the positive, if you want really large breast put "large breast in positive and small breast in negative.
Adding weight is even more powerful, example:
Positive (large breast:1.2)
Negative (small breast:1.2)
Use a value over 1 for an increase and less than 1 for a decrease.
Don't forget to add terms that describes the genre style and terms that describes the image/photo style such as "wide angle lens", "depth of field", "sharp focus" etc, also terms for the color and light such as (vivid color:1.2) or (vivid color:0.8) for less, (diffused light:1.2) or (diffused light:0.8) for less.
