Introducing AIイラストおじさん
with latest version ---> V3-Exp
Compatible with WAI-NSFW-illustrious-SDXL and NoobAI-XL (NAI-XL).
Tips:
Parameters:
• CFG: 3 - 5.5
• Steps: 24 - 35
• Sampler: Euler and DPM++ 2M Karras for more details.
Strength LoRA:
• 0.3 - 1.2 (1.0 for soft tones, 1.2 for stronger tones).
Hi-res fix ---> Denoise:
• 0.05 - 0.65
Recommended Positive Prompts:
determined by the pattern: ([Quality] + [Era/Style] + [Classification] + [Character/Subject] + [Clothing/Features] + [Action/Pose] + [Setting/Background]
(masterpiece, best quality, newest, highest, absurdres)
shinny skin, wet
Negative Prompt (add only what is necessary to achieve balanced quality):
((censored)), 3d, (worst quality, low quality, normal quality, lowres, blurry, jpeg artifacts:1.2), (bad anatomy:1.3), (bad hands:1.4), (missing fingers:1.4), (extra fingers:1.4), (fewer fingers:1.4), (fused fingers:1.4), (too many fingers:1.4), (extra limbs:1.3), (malformed limbs:1.3), (mutated hands:1.4), (poorly drawn hands:1.4), (bad hand anatomy:1.3), (missing hand:1.3), (text, watermark, signature, logo), monochrome, grayscale, extra head, extra body, bad perspective, deformed face, poorly drawn eyes, poorly drawn mouth, duplicate, error, cropped
Description
FAQ
Comments (15)
I love this Lora. Can you make an animagine 3.1 version, if possible?
Excellent, def must try.
One minor thing is that Lora tend to produce low angle pictures by default. Sometimes its hard to force it to average mid angle.
Well, I think this lora can probably do that. and this is what I will try to improve in version 2.0. Hope this helps you :>
possible have this for 1.5?
I only tried it once and the result is excellent. Absolutely a perfect Lora.
I got into this space solely because of this LoRA. On my first generation, the image turned out perfect, with almost no flaws. You definitely deserve recognition from the community. unless you’ve already received it! Regardless, I’m sure it must have been incredibly challenging to achieve such an amazing result with this LoRA.
No, I havent been recognized yet, but I dont mind. It was a bit challenging, yes, but the trial and error was definitely worth it :)
This LORA instantly upgrades the quality of most pony models into the next level.
it is possible to look at the datasets? I want to know how you made it.
I made several sets, it was trial and error until I found one that really worked, so it will be a little difficult to find the right set that works. just like this version
@katohh did you generate the dataset using AI? or did you find the images?
Great lora. It's kinda annoying that it really wants to generate "from below" images
滅茶苦茶よかった。
noobAI用も作ってほしい
nice
Awesome style!
Details
Files
Available On (6 platforms)
Same model published on other platforms. May have additional downloads or version variants.