How to prompt this model
Write in prose, not keyword lists. One flowing paragraph in this order:
Subject — who, once, precisely ("an adult woman with long wavy blonde hair")
Pose/action — what the body does, one line, sides only if asymmetric
Clothing/materials — described physically, how they catch light
Setting — where, with 1-2 concrete objects
Light — source, direction, interaction with surfaces
Camera — angle and framing anchors ("low-angle from foot level", "head at the top and feet at the bottom with everything inside frame")
Describe materials physically instead of using quality tags. "Glossy black leather with sharp reflections" does more than "masterpiece, best quality, ultra detailed" — generic quality tags are unnecessary on Krea2 and can push a plastic over-sharpened look.
No negative prompt needed. Krea2 reads everything as positive, so negatives like "bad anatomy, worst quality" can backfire. Omit what you don't want by silence, and describe each element once, precisely — repetition gets exaggerated instead of obeyed.
For difficult anatomy (hands, feet in foreground), keep it to one simple line and anchor positions in the frame. If a foot still comes out wrong, inpaint just that region instead of re-rolling dozens of seeds.
Recommended settings (Krea)
Sampling steps: 8-16
CFG scale: 1
Sampler: Euler
Did you like the content? I would be grateful to receive a coffee
Usage responsibility: this is an open model with no content filter. You are solely responsible for the images you generate and publish with it. Do not create illegal content, non-consensual intimate imagery, sexual content involving minors (real or fictional child-like characters), or any other material prohibited by Civitai's Terms of Service and the laws of your jurisdiction. The model author assumes no liability for misuse.
Description
FAQ
Comments (9)
Having some issues with this model. I think it looks amazing and I want to start using it but all my pictures become scrambleld. I tried it on two computers, once it started working and then it stopped. Not really sure what is going on. The image just comes out at some grayish pattern. Even if I only type "girl" with no negative prompt.
I believe it could be a cache problem, try clearing the cache on your Stable Diffusion and download the model again, then test.
I usually have worse problems, but with all the models in stable diffusion, and I still haven't found out what it is, and even after reinstalling the problem continues, every now and then my Stable Diffusion loses its Precision and all the images are bugged. and honestly I still haven't discovered the cause of the problem, and sometimes I even get discouraged from creating images because of it.
@BerserkFG Yeah there's something weird going on. I got it working for a while on my other computer but then I did something and I don't remember... Now it doesnt work on neither computer nor on a reinsntallation of auto1111. This is what it produces:
https://i.imgur.com/ar1Bra1.png
@nsfw_master I'm about to launch a 2.0 version of the model today, where I did some tests and apparently it's generating prettier images, so in this case, delete this 1.0 version and test 2.0 to see if the same thing happens. I just haven't launched it yet, because I'm generating the images for the catalog. I hope this problem doesn't continue to happen to you.
@BerserkFG It's just some weird bug tbh. Will test your 2.0 model def
@BerserkFG Got it working with your 1.1 model. Awesome
@nsfw_master I'm glad you solved it, apparently there was a problem with my old model that was having trouble with its stable diffusion
@nsfw_master this time I included the VAE in the model, that may have been it










