SilvermoonMix01 was trained with NSFW images only. This model excels at depicting the details of female characters, including elements like liquids, hair, lips, muscle, lighting and shadow.
For instructions on how to use the V-pred model, please refer to https://civarchive.com/models/833294/noobai-xl-nai-xl
Recommended settings:
Sampling method: Euler a
Start with no negative prompt. Sample images were generated without negative prompt.
Use ADetailer extension to enhance facial details: https://github.com/Bing-su/adetailer
IllustriousXL comes with some jpeg artifacts that can be removed by ADetailer or Hires. fix. However, if you don't want to use these, try adding jpeg artifacts to negative prompt. Please note this may also remove some background details.
推荐设置:
采样方法:Euler a
先尝试不使用负面提示词。示例图片都没有使用负面提示词。
使用 ADetailer 扩展来增强面部细节:https://github.com/Bing-su/adetailer
IllustriousXL 会带有一些 jpeg artifacts,这些可以通过 ADetailer 或 Hires.fix 去除。不过,如果你不想使用这些工具,可以尝试将 jpeg artifacts 加入负面提示。不过,这可能会同时去除一些背景细节。
Change notes:
v6.0: Further enhanced lighting, prompt obedience and compatibility with LoRAs that are trained on other models.
v5.0: Improved compatibility with LoRAs that are trained on other models. Lights looks more nature now. Improved logic.
v1.1: Fine turned twice with fantasy backgrounds and impasto portraits:
Improved background details.


Removed unnecessary skin details. This makes images "cleaner".

Improved shadows logic.

Improved an issue of torso being too long.
v1.0: Fixed underfitting issue.
v0.1: initial release, for testing only. It is currently quite underfitting.
Description
FAQ
Comments (13)
Wow, works really well stably without any negative prompts!
If you start with the prompt 2girls,mother, it seems to be interpreted as mother and daughter, not 2mother.
The problem is that daughter is basically treated as a loli, even if you specify an age such as 15yo.
I submitted a comparison chart that shows that you can do fine things without negative prompts, but if you can't see them, you'll have more caveats to this wonderful model.
2girls,mother,というpromptで始めると、2motherではなくmother and daughter, と解釈するようです。
問題はdaughterを基本的にloliとして扱います。15yoなどの年齢指定してもloliとして扱います。
ネガティブプロンプトなしでも立派なものができるんだよという比較図を提出しましたが、見えなくなっていたら、この素晴らしいいモデルの注意点が増えます。
Hi, thanks for your comments.
I believe the "daughter" comes from prompt "age difference".
Please check my images:
1. If you want two mothers, remove age difference.
2. If you want mother and daughter, but not a loli; add "petite, loli" to negative prompt.
@silvermoong
Thanks for the reply, and thanks for the “admonition” at the same time. I think I was getting a little carried away.
Before I reply to your reply, I want to say what I think about this Model. It's very well done. I'm surprised that it can be done so well without the negative prompt, but I think it's a good job even when operated as usual with the negative prompt.
Reply to Reply.
I am sure [age difference] is having a negative effect, but it still seems to generate [petit,loli].
I think it depends on SEED and can be covered by the number of attempts.
1. At first it was [mother and daughter], but it didn't work, so I tried to make it only [mother], but it didn't work. I made a mistake. (Now that there is an age difference, I think it would have been fine to generate [older mother] and [younger mother].
2. In the image shown here, “petite, loli” was added to the negative prompt. I didn't feel much effect. I'll add an image for comparison of the negative prompts, just for reference since it is a one-shot.
I personally don't mind [loli], but [civitai] makes me angry.
返信どうもありがとう、同時に「諌めて」くれてありがとう。少し舞い上がってた気がする。
返信に対する返信の前にこのModelについての感想を言いたい。とても良い出来だ。ネガティブプロンプトなしでも立派なものができるのが不思議だが、通常通りネガティブプロンプトを入れて運用しても、良い出来だと思う。
返信に対する返信。
[age difference]が悪影響を及ぼしているのは確かだが,それでも[petit,loli]は生成されるようだ。
SEED次第なので試行回数でカバーできる、とおもう。
1.最初は[mother and daughter]だったのだけど、うまくいかないから[mother]だけにしたらどうかと試すが、ダメだった名残だね。勘違いさせてしまった。(いまならage differenceがあるのだから、[older mother]と[younger mother]が生成されても良かった気がする。)
2.今回挙げている画像はネガティブプロンプトに "petite, loli"を加えている。あまり効果を感じなかった。一発撮りだから参考程度だけどネガティブプロンプトの比較用画像を追加しておく。
ぼく個人は[loli]だと困るんじゃなくて、[civitai]が怒るんだよね。
What was the dataset/merge recipe for this checkpoint?
This model replicates my SilvermoonMix01 on SD1.5 SilvermoonMix01 | delicate skin and hair texture - v7.3 | Stable Diffusion LoRA | Civitai.
I used a similar recipe and trained IllustriousXL.
I am amazed at the quality and stability of the models you have developed.
By the way, it looks like certain characters are generated without Lora. Is there a list of wildcards? Perhaps it would be possible to divert the wildcard list from an existing model?
Hi, thanks for your support!
Unfortunately, I don't have a good way to find out what characters the model knows.
@silvermoong Thanks for the reply. OK, I get it. I'll try some words on my own then.
好得很
Pretty good actually. The possible full version in the future is very potential.
Legendary
Ishikei enjoyer detected



















