generait girl having foxy eyes
can generait male Character and nsfw image
Not good hands
▼Please check settings and prompts▼
Merge model(V2)
Nabimix_V1
Dalsefopainting_V4
Treebark
Murata Yuusuke (OPM Mangaka) LoRA
Recommended setting
Sampling method : DPM++ SDE Karras, (DPM++ 2M Karras, DDIM)
Clip skip : 2
Hires steps : 13
Hires.fix upscaler : R-ESRGAN Genaeral WDN 4xV3, (R-ESRGAN 4x+Anime6B)
CFG Scale : 5~10
VAE : berrysmix = kl-f8-anime2
Recommended prompt
Prompt : (masterpiece, best quality)
Negative : (worst quality, low quality:1.4), negative_hand-neg, verybadimagenegative + (blush)
Textual Inversion : negative_hand-neg, verybadimagenegative
Description
Merge model
dalcefoPainting v4
SamDoesArts (Sam Yang) Style LoRA
Process
0.6(nijiv5style_v1) + 0.4(abyssorangemix3AOM3_aom3a3) = A
A + 0.67(AU (delete_au) Style LoRA) + 0.22(Disco Elysium - Style LoRA) + 0.25(Yoneyama Mai Style LoRA) = B
0.85(B) + 0.15(dalcefoPainting v4) = C
C + 0.13(SamDoesArts (Sam Yang) Style LoRA) + 0.2(Kidmo//style LoRA) = D
0.8(D) + 0.2(cetusMix_Version3.5) = E
0.75(E) + 0.25((nijiv5style_v1) = F
F + 0.3(Murata Yuusuke style LoRA) = NabiMix
Recommended setting
Sampling method : DPM++ SDE Karras, (DPM++ 2M Karras)
Clip skip : 2
Hires steps : 13
Hires.fix upscaler : R-ESRGAN Genaeral WDN 4xV3, (R-ESRGAN 4x+Anime6B)
CFG Scale : 7~10
VAE : kl-f8-anime2
Recommended prompt
Prompt : (masterpiece:1.2, best quality)
Negative : (worst quality, low quality:1.4), negative_hand-neg, verybadimagenegative + (blush)
Textual Inversion : negative_hand-neg, verybadimagenegative
FAQ
Comments (6)
Getting very different results from the samples shown. Using entire generation data + correct VAE before you ask though.
I see what happened, the samples are from img2img, they need to be generated in 512x768 and upscaled not generated straight away as 1024x1536.
Still on some images not even remotely close, still a good model but keep that in mind. I'll upload a review once possible.
I talked to him before i asked why didn't i get the same image as him, he said it might be xformers but i personally think he removed some of the prompts when he upscaled to Img2Img so it might be like this.
Original prompt in txt2img:
(masterpiece:1.2, best quality), 1lady, solo, upper body, huge breasts, (latex skin, selfie:1.15), (face focus:1.2), top view, white maxi dress, Museum, (black short hair), round earrings,
Original Negative:
(worst quality, low quality:1.4), multiple views, multiple panels, blurry, watermark, letterbox, text, EasyNegative, badhandv4, (elf ears, nsfw:1.2)
Img2Img prompt:
(masterpiece:1.2, best quality), 1lady, solo, upper body, huge breasts, selfie:1.15), (face focus:1.2), top view, white maxi dress, (black short hair), round earrings,
So my theory is I think we're missing some of the prompts because he might've removed it and that is why we don't get a similar result. You can take a look of my post in this model to see the difference in our results.
As a result of checking, I found three reasons why images are generated differently.
1. As you said, the image size recorded in Exif during i2i process has changed.
(Original Image Size 512x768)
2. Seed of some images has changed in the i2i process. The seed changed because I removed seed and used SDupscale.
3. denoising strength was changed when generated i2i image.
(most of original image's denoising strength is 0.5)
Thank you for pointing out the problem.
just discovered your models ! love it, do u have a discord channel or smth i can join and stay up to date ?
You can visit Arca live
https://arca.live/b/aiart/72189422
the merging recipe alone ald drops my jaw.. simply amazing..