Experimental models that can run unstable. You have been warned.
For those who like to experiment with artist tags.
->Read the description for each model.<-
Continuation of HL's styles series
Description
I'm figuring out the model. In process rn.
So far I can tell you this:
Was trained and tested on the WAI-NSFW-illustrious-SDXL v12. I haven't tried other models yet.
You can use weights in the range of 0.7-1, I stopped at 0.8-0.9
VERY RECOMMENDED to use artist tags. Improves quality.
LoRA shows itself quite well with the following tags:
limited palette - the basis of this LoRA, I recommend using it.
-any color- theme - just add a theme.
multicolored background - if you're tired of one color simple backgrounds
realistic - improves style, but it's a matter of taste.
film grain - it depends. Sometimes it improves image quality, sometimes it downgrade it. Also fine to put this tag to the negative prompt. Try it.
And ye, experiment.
FAQ
Comments (5)
Maybe it's a little unstable, but so is most of Illustrious right now. This is still a reliably beautiful aesthetic influence, I hope you'll keep working with it.
Would you be interested in training shiun aka by ziyun tag on AidXL/Artiwaifu ?
Preferrably on Noob Vpred 1.0 for use on ΣΙΗ. :)
Honestly, it's not a very strong LoRA. It just subtly directs the model in a direction that I can't even describe. The results presented by the author on the LoRA page were mostly achieved by prompting artists, with the LoRA serving as just a final touch.
It's not a secret. I wrote that in the description for the LoRA.
worked fine for me. maybe try typing a few more tags after 1girl
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.