I hate early access, so no early access :D Also general tricks here in description:
ArtBrush Cartoon
Integrated Brush style. Gives the checkpoint a very artistic flair. It was just something t play around with. so I added it here to the other checkpoints.
V3 realistic Proportions
Tried to keep the style but make the body proportions more realistic and less exaggerates like in cartoons. Goes a little bit into semi-realism.
V3 Ultra refined
After long sessions, I managed to clean the style up entirely! You can compare the images from V3.0 to the other versions. Everything is waaay sharper and facial details are super detailed, without explicit prompt. AND I didn´t used a adetailer for the sample images! Very happy with this version!
V2.0 way more defined, way less blurry!
I aimed for a Style, that is cartoony but with an artstyle that creates depth. So no flat colors and complete 2D style. Overall goal was an elegant cartoon style with highlights.
V1.0
I put like 30+h in there and combined 60+ styles until I found a mix that I liked. AND I FUCKING LOVE IT!!!
I backed in a strong CFG control so the faces are way more detailed than normal Illustrious models.
I didn’t used adetailer, but feel free.
This Checkpoint eliminated every Anime existing in Illustrious. At least I didn't find an animestyle it responds to.
On the other side, If you want a Style, that is totally flat.... well it will adapt but wont make it flat 2D.
The sample images are complete generated without loras.
The samples are done with CFG 4, 6 and 7. you can already get decent images with CFG4 and 24 Steps.
My Settings:
Euler
220-30 Steps are okay, fell free to use more.
Clip Skip 2
CFG 4-7
Also because I saw many people missing it, some systems allow:
Instead of writing:
((((nude))))
you can also write:
(nude:1.4)
Or
((((Nude)))), ((((ass)))),
you can write:
(Nude, ass:1.4)
this will make your prompting a bit more less chaotic.
This is mostly a NSFW Model. Please handle with care and don't harm the platforms guidelines. I take no responsibility for anything. I just leave it here.
Description
FAQ
Comments (4)
Pretty good model/merge for 2.5D western art! Most Loras that I tried work fine but some need really high strength to take effect (up to 4.5 😱) !!!
Creates great scenery but is biased for certain poses ("girl straddling guy in POV view"). The "pov crotch" sometimes even appears when using only scenery prompts.
I uploaded a bunch of (cherrypicked) gens that dropped out of my ModelTester workflow.
oh hmm lets see if i can fix that in future version.... I actually never use pov😅
regarding the Loras with 4.5 strength: is it style related? cause i got quite aggressive against anime style.
@quadratrund it is kinda strange - you can see the "pov crotch" in this image: https://civitai.com/images/62721774 but it's possible that my random prompt generator accidentally prompted this... let me check that
@quadratrund mh, no, the prompt is
"diffuse lighting,diffused light,moon,(fisheye view:1.25),straight-on view, (illustration:1.2), source_cartoon, masterpiece, very aesthetic, absurdres, western art, (inside, indoors:1.3), hotel room,Photographs on the wall"
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.