Kageillustrious
Woohoo, this is my first merge and personal mix I would like to share with you! This model's main FOCUS is the MALE physique and HOMOsexual themes. This mix would not be possible without the amazing HomoSimile XL model as it's base. Other ingredients include, Coco Illustrious and Illustrious Gehenna. And don't worry, also included is the up to date version of NoobAI's datasets. So artist tags and prompt adherence are on point. And finally, a large mix of personal Lora's I've crafted to help influence the overall male theme.
Model Looks Best at These Settings:
Sampling method: Euler a
Sampling steps: 20
Width: 896
Height: 1152
CFG Scale: 6
Positive Prompt:
masterpiece, best quality, amazing quality, very aesthetic, absurdres, very awa (better for styles/poses)
-or-
masterpiece, best quality, amazing quality, very aesthetic, high resolution, ultra-detailed, absurdres, newest, colorful, rim light, backlit, highest detailed, very awa (better for characters)
Negative Prompt:
worst quality, bad quality, bad hands, very displeasing, extra digit, fewer digits, jpeg artifacts, signature, username, reference, mutated, lineup, manga, comic, disembodied, cropped, amputee, bad foreshortening, what, guro, logo, bad anatomy, bad perspective, bad proportions, artistic error, anatomical nonsense, amateur, out of frame, multiple views, text ( copyright name:1),
Helpful tags for styles and characters can be found here - https://www.downloadmost.com/NoobAI-XL/danbooru-artist/
And as usual I always recommend Adetailer for the faces and other details.
Check out my other work! - https://beacons.ai/kagemaboy
Description
Brighter colors/ Better Anime Style / VAE Included
-Still Work In Progress-
FAQ
Comments (16)
Can't use the v2..the color seems really good in k-samper,but when it comes to saving image it'll be like less color.
I've checked my custom nodes in ComfyUI, they are not differ from what I used before. I haven't change anything besides the checkpoint.
Here are the links:
V2:Could not use...Why? | Civitai
Same nodes and settings V1:Image post by Gay_FlipFuck | Civitai
I'm looking into what happened to fix it. Thank you for letting me know!
Looks like it's a VAE issue. I'm going to try and bake a VAE into the model so it should fix the problem :)
Please redownload and test it again. I've added a VAE into the model. Thanks!
@Kagemaboy OK! I've delete my posts~ Thanks for your works!
Hmmm, I tried it on my own diffusion. But the coloring seems tends to be darker than the v1. But the problem seems to be resolved by using VaE (I dont need to use Vae on v1). I will keep try to see how is it compared to old one, but do you also have any vae recomendation for the V2? Thanks beforehand~!
sdxl_vae works well with it
Please redownload it as well. I have applied a VAE to the model. Thanks!
Hnnn, after some test, I think i must say that I think v1 is more consistent in generating image. While v2 is greater in making poses, I feel that it tends to alter the style and also body types of the lora used, unlike V1, also, I feel that the datasets became smaller in some aspect? Character that can be generated without loras in V1 cant be generated anymore in V2. And it keeps making distorted images that never happened in v1.
@RaidenFan86 Thank you for the feedback! I hope to fix these issues moving forward.
I don't know if it's just me, but V2 seems to have lost knowledge of certain stuff (like foreskin)
I just noticed that too! I will keep working on it to bring back what was lost. Please let me know if you notice other concepts missing. Thank you!!
Please redownload and test again. I improved this version. Thanks!
@Kagemaboy Works really nice now, thanks!
@cwhk Yay, I'm so happy!! Thank you!


















