Hello everyone,

today I upload some character-Model of a requested anime "Ghost Hunt". Now iam starting with the side characters, next one is Chiaki Kasai. She has her appearance in the third case.

Prompt: look at "Trigger Words" on the right side 👉

Changelog: look at "About this Version" on the right side 👉

Some word about the outfits: The outfits are somewhat improvise and can be a bit unstable, because THEY CHANGE THE FU**ING OUTFIT [Color] EACH EPISODE OR LESS😠😢😭. Furthermore, the anime is pretty old, and I didn't find an FHD Version of it. Many full body images turned out pretty bad, because of the poor quality, so I had to remove them from my training-Data Finally some characters with not so many outfits! 🎉🤣😅

quality tags I use:

oekaki, anime coloring, anime screencap, masterpiece, best quality, amazing quality

My Settings (forge)*:

*I made it more detailed, as some user ask for it

*note, the example images only use high-res fix.

Checkpoint: waiNSFWIllustrious_v150

Sampler: Euler a

Scheduler: Automatic

Samplings Steps: 30 - at least use more than 15! I often see people generate with 8 steps. This it not enough xD (I mean, you should generate you pics the way you want, but the results should be better with more steps)

Hires.fix: R-ERSGAN 4x+Anime6B, 20 Hires steps, Denoising Strength: 0.3-0.5 (I use 0.5)

Extensions (nice to have for little more Details):

Adetailer (highly recommended)

model: 1st 99coin_anime_girl_face_m_seg.pt

Detection: Confidence - 0.7

Mask Preprocessing: Mask erosion: 24

Impainting: Mask Blur 12, impaint only masked 96, impaint denoise 0.4, use seperate with/height 1024

model: 3nd Nipple-yoro11x_bbox.pt (optional, mostly for NSFW):

same as 1st + Adetailer prompt: <lora:IL20_NP31i:0.6>nipples,

*You need a lora for above prompt, just search on civitai

4nd model: pussy_v4_best.pt (optional NSFW)

same as 1st

CFGRescale Integrated (forge): Default (0.7)

SDXL Styles (anime)

CD Tuner: detail(d1): 2, everything else default (mostly 0)

Detail Daemon: Default

DynamicThresholding: Default

FreeU: B1: 1.1, B2: 1.2, S1: 0.6, S2: 0.4 (leave the Rest deafult)

SelfAttentionGuidance: 0.5 (leave the rest default)

img2img (for upscaling):

Resize by 1.5

extension above, but either use them in text2img OR img2img NOT both!

Script: Ultimate SD upscale (extension)

target size type: From img2img, Upcaler: 4x-Ultrasharp (need to download and copy in your ERSGAN folder - should be looked like this YourDrive:\webui_forge_cu121_torch231\webui\models\ESRGAN), Type: Chess Tile width: 1024, Tile Height: 0, Mask Blur, Padding: 52