V3.1 - Rebaked all styles then merged on ChenkinNoob-V0.2, made results a lot cleaner.










If the images are too dark for your taste put "dark theme" into the Negative prompt.
Note all previous illustrious-0.1 and NoobAi loras work perfectly fine from what I have tested on ChenkinNoob.
If you are using Reforge WebUi turning on EPS scaling can help improve gens further, I personally also use FreeU for the samples above. Gen settings I use are the same as previous versions.
Fixed some of the quirks from V2.2 but please do share if you find anything I may I have missed.
Sampler: Euler A AYS
Sampling Steps: 24
Upscaler settings:
1.5x - 2.0x Upscale
10 Hires steps
Denoise 0.1-0.6 (Use lower denoise on higher Upscale)
Suggested Dimensions: 832x1216, 1216x832, or 1024x1024
CFG Scale: 4-6
ADetailer along with Anzhc's Face seg 640 v3 y11n
Inpaint Denoise: 0.2-0.5
Use Separate width/height:
1024x1024
Inpaint mask blur: 24
Inpaint only masked padding, pixels: 48
Positive prompt base:
masterpiece, best quality,
Negative prompt base:
worst quality, low quality, bad , watermark
The finger issues from V2.1 have been much reduced and somehow after lots of mixing it came out even cleaner so I hope you all enjoy. Worst case use fewer digits in the negative for more finger stability.
Was noticing various finger and hand issues with blur and artifacts that bugged me so I went and tried to fix it. Should be a lot better and cleaner now but if anything is consistently odd please let me know.
If you have some issues with fingers add (fewer digits, 3 fingers:1.3) to the negative and it should help until I update the model later.
Goal of V2 was to use Anzhc's new SDXL Anime VAE Dec-only B3 VAE which is baked into the model, I suggest to use this VAE or none as its baked in. Otherwise I wanted to make the mix cleaner and sharper with much better dark gens.
Use: Dark, dark theme, darkness for general dark gens or night, dim lighting for light specific darkness.
There are some mild quirks, if animal ears fail make sure to add animal ears tag to stabilize it. Sometime extra people leak into gens if you forget solo.
Will fix most of the issues with a V3 but I need to remake and find more styles before I can do that.
Sampler: Euler A AYS
Sampling Steps: 24
Upscaler settings:
1.5x - 2.0x Upscale
10 Hires steps
Denoise 0.1-0.6 (Use lower denoise on higher Upscale)
Suggested Dimensions: 832x1216, 1216x832, or 1024x1024
CFG Scale: 4-6
ADetailer along with Anzhc's Face seg 640 v3 y11n
Inpaint Denoise: 0.2-0.5
Use Seperate width/height:
1024x1024
Inpaint mask blur: 24
Inpaint only masked padding, pixels: 48
Positive prompt base:
masterpiece, best quality,
Negative prompt base:
worst quality, low quality, bad , watermark
Important note: Baked in characters for illustrious/NoobAi are often burnt so you may need to reduce weight so it doesn't make the ckpt look strange.
New merge to tickle a different craving I was having, spent some time improving my ability to train styles and used that to make the styles I merged into this model. I hope you all enjoy.
Check out the samples above for inspiration and/or ideas.
V3.1 - Rebaked all styles then merged on ChenkinNoob-V0.2, made results a lot cleaner.
1. Hak untuk model yang diposting ulang adalah milik pembuat aslinya.
2. Pencipta asli yang ingin mengklaim model harap hubungi staf SeaArt AI melalui saluran resmi. Klik untuk mengklaim
