Okay so we have one of my favorite styles here: The art of the artist Queen Complex!
It is cute, it is goofy aaaand nearly every style-lora fucks this style up. So I did it myself like with my Sabudenego-Lora. It is not so hard to crop out watermarks and upscale your shit before you train on it... seriously...
NO Watermark! (If you get some, it is not the fault of the lora)
Krea 2 (even works with other concept-loras that are trained on realistic images)
One of my favorite style!
trigger: qu33n, illustration
strength: 0.7-1.0
Illustrious
Strength: 0.8-1.0
Trigger: qu33n_d4rk,
Recommended: solo (if too many girls)
I used Wai-Checkpoint and Euler A Sampler
I only fucked up one thing, but I will have to retrain it: The Pics of Queencomplex are often showing groups of people. If you want to keep the girl "single" also in the act with several boys, use the prompt "Solo"
Queencomplex has in my opinion two Styles (3 if you count the 3D art). One is fairly simple and bright and the other is more dark and moody (this one here).
The dark and moody one has way more complex backgrounds. But because some Lora Creators don't care, they push everything into one lora. So Backgrounds with Queencomplex-lora are always a bit weird...
Not anymore! Here I will give you the "bright" version with friendly colors and simple Background.
I will also upload the bright style. more friendly, more simple and more bright.
Description
FAQ
Comments (7)
Awesome style!
Which WAI checkpoint are you using? I've tried WAINSFW v14 and v15 and not getting even close to queencomplex's style with this
Also, I realized that may have come off a little aggressive in text, and I don't mean it that way. I'm genuinely trying to figure out how to make the best use of this lora :)
I am using Wai v14.
My current setup for the pics are:
Wai
5 Step Lora (I believe it is this. Was a long time ago, that i downlaoded it:https://civitai.com/models/1857032/vvlxviii-speed-boost-asset-cfg-1-steps-5-dmd2)
Queen Complex lora at strength 1.0
euler a
cfg1
8 steps
It could be, that the 8 step lora enhances the style. let me try it out.
So most important probably: I don't use negative prompts. sometimes there are stuff like low details and so on in negativs.
AAAND I only used: masterpiece, best quality; in positive prompts.
the thing is, many people always are having their: masterpiece, ultra detailed, best quality, intricate details, cinematic lighting - in their prompts. which are great for awesome pics. but this will destroy often the style.
Queens style is simple, cartoony. When you force the ai to generate it with additional details and more complex lighting than it has, obviously it will look different.
the message wasn't aggressive :)
Okay so i think I found the solution:
so it seems my 5 step lora is making artstlye stronger than they should be. but not a lot.
I got good results without it with:
strength of 1.1-1.2 and set "3D" as negative prompt too.
@quadratrund Awesome, thanks a bunch for the reply! I'll try it both with more strength and with the 5 steps lora and see what works, but I really appreciate it!
@treiceleitner feel free to write me if it has worked!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.