|
--- |
|
license: creativeml-openrail-m |
|
library_name: diffusers |
|
pipeline_tag: text-to-image |
|
tags: |
|
- ' stable-diffusion' |
|
- stable-diffusion-diffusers |
|
--- |
|
|
|
|
|
# what is "SSSSLLDDLL" |
|
|
|
このプロンプトが全てだ。 |
|
check this prompt, its that all. |
|
|
|
> shiny sissy luxury latex dress for doll, subtle ambient glow, global illumination, <lora:flat2:-1>, |
|
> Negative: EasyNegativeV2,badhandv4,worst quality,(noise:1.1),unnatural pose,improper proportions,facial asymmetry,disconnected elements,irrelevant background, stain, animal ears, |
|
|
|
|
|
|
|
 |
|
|
|
|
|
|
|
# V3 is comming! |
|
|
|
このマージモデルはラテックス・ラバー・エナメルの美しさを、アニメ風のイラストに反映させる目的で作られました。V3はその目的を維持しつつ、より体型の崩れを抑えて、顔を可愛くしました。また、従来よりも若干イラストに近い2.5Dモデルを目指し、使いやすくしました。 |
|
|
|
This merge model was created with the goal of reflecting the beauty of latex rubber enamel in a anime like illustration; V3 maintains that goal, but with less body shape and a prettier face. i also aimed to create a 2.5D model that is slightly closer to the illustration than before, making it easier to use. |
|
|
|
|
|
|
|
そのため、SSSSLLDDLLv3は普通のアニメモデルとして使用することができます。この画像の通り、ごく普通の日本の成人女性も、かわいく出力できます。瞳の表現に自信があります。 |
|
|
|
Therefore, SSSSLLDDLLv3 can be used as a normal anime model. As shown in this image, even a very ordinary Japanese adult woman can be output in a cute way, confident in the expression of eyes. |
|
|
|
> standing, wide angle, (1girls:1.3), harem anime official art, seifuku clothes, school uniform, wide angle, glossy detailed pantyhose, large eyes big, cute beautiful eyes detailed pupils cute, circle white pupil cute, town park background, 35mm f1.4, iso100, raw, high quality,subtle ambient glow, raytracing, global illumination, ultimate quallity, |
|
|
|
 |
|
|
|
|
|
|
|
しかしこの画像を見た多くの人は、こう考えるはずです。「この女性にピチピチのラテックススーツを着せたい。いや、彼女のカバンの中にはラバースーツがあるに違いない」 そう、このモデルは貴方の夢を叶えるために誕生しました。 |
|
|
|
アニメ顔を維持しながら、実写のようなファブリックのディテールを実現しています。"イラストや3DCGでは困難だが、AIモデルだからこそ表現できるアート" だと考えています！ |
|
|
|
But you guys who see this image will think: "AYAYA I want to put this woman in a tight-fitting latex suit. No, I bet she has a rubber suit in her bag!" , Absolutely, this model was created to make your dream come true. |
|
|
|
i have achieved fabric details like live-action while maintaining the cartoon face. i believe this is "art that is difficult to achieve with illustrations or 3DCG, but can only be expressed with AI models" ! |
|
|
|
> sitting desk, wide angle, cowgirl position, ([precure:lovelive:0.1] face:1.1), (black serafuku:1.2), (fusion black tight school uniform and leotard with sissy frills:1.3), black cameltoe, pantyhose, black gloves, wet clothes latex, lip gloss, makeup, sweat, heavy breathing, (hime cut:0.8) ,black hair, (school classroom:1.2), evil smile, (vulgarity:1.2), school girl, blue eyes, skin glossy, shiny clothes, glistening fabric insanity detailed, subtle ambient glow, global illumination, <lora:flat2:-0.4:NF>, |
|
|
|
 |
|
|
|
|
|
|
|
# BREAKING: v32 "Fixed Face" available now! |
|
|
|
## 2023-05-28 updated! |
|
|
|
 |
|
|
|
良いニュースがあります！ v32(3.2)がstableになりました！ 顔をよりアニメ・イラスト寄りに近づけ、可愛くなっています。"V32FF"のFFには、その意味を込めています。 |
|
|
|
また、イラスト・3DCG・実写のバランスを調整し、より私好みで、より他のモデルにない雰囲気で、よりユニークな、イラストを生成できるようになりました。 |
|
|
|
つまり、SSSSLLDDLLらしい画像を、もっと楽しめるようになりました！ |
|
|
|
BREAKING! v32 (3.2) is now STABLE! |
|
|
|
The faces have been made prettier, closer to the anime/illustration. The FF in "V32FF" means just that. |
|
|
|
and the balance between illustration, 3DCG, and live-action has been adjusted to generate illustrations that are more to my liking, more unique, and with an taste not found in other models. |
|
|
|
In other words, you can now enjoy more SSSSLLDDLL-like images! |
|
|
|
 |
|
|
|
### HIGHLY RECOMMEND !! |
|
|
|
For better latex gloss and detail, use iroiro-lora! |
|
|
|
 |
|
|
|
- flat1 or flat2 : minus improves detail. Gloss is also improved. |
|
- deluster1 or deluster2 : Minus for more luster |
|
- hairdetailer : Better detail in hair, better detail in clothes. In SSSSLLDDLL, it makes the hair especially silky. |
|
|
|
and i recommend extension "Self Attention Guidance". Gloss is improved too. |
|
|
|
for example: |
|
 |
|
|
|
## Why no V2? |
|
|
|
実はV1とは別のモデルで作成したV2も一部で公開していたのですが、手足の安定性が低く悩んでいました。しかし神のようなLoRAをマージしたところ、人体の補正もキャラの顔も向上しました。これをV3と呼称することにしました。 |
|
|
|
Actually, V2, which was created with a different model than V1, was also partially published, but it suffered from low stability of the limbs. However, when we merged the GOD-LIKE LoRA, both the correction of the human body and the face of the character were improved. im decided to call this V3. |
|
|
|
> wide angle, missionary, ((kinky harajuku kei)), (sissy:1.4), (silky satin sissy long dress with many ruffles:1.3), (textile shading lolita fasion satin dress:1.3), (Glossy clothes luster:1.3), garter straps, (thighhighs:1.1), gothic latex corset, boots, (torpedo breasts:1.0), heavy makeup, eyeshadow, eyebrows hidden by hair, cheek blush, thick rouge lips, narrow waist, wide hips, large eyes big, circle white pupil anime cute, imperial palace room, subtle ambient glow, global illumination, ultimate quallity |
|
|
|
 |
|
|
|
そのLoRAの名は [Big Hard Latex](https://civitai.com/models/51221/big-hard-latex) 。副作用として、ラテックスアウトフィットも実際に販売されているものに近いデザインになったかと思います。 やはりLATEX……！ LATEXはすべてを解決する……！ |
|
|
|
それとは別に、huge breastsで出力されるバストがより大きくなるようにしました。 |
|
|
|
The name of that LoRA is [Big Hard Latex](https://civitai.com/models/51221/big-hard-latex). As a side effect, I think the latex outfits are now designed closer to what is actually sold. It is still LATEX ......! LATEX solves everything ......! |
|
|
|
Apart from that, we made the busts output with HUGE breasts larger. |
|
|
|
> ((goddess)), (textile shading Navy Blue pvc suit with many ruffles), (angel wing, angel halo), gothic corset pvc, ((torpedo tits, wide hips, gigantic boobs, narrow waist, oppai loli)), Gray hair, seductive smile, in heat, , shiny clothes, glistening fabric insanity detailed, subtle ambient glow, global illumination, |
|
|
|
 |
|
|
|
|
|
|
|
# tips (simple or detailed) |
|
|
|
このモデルでシンプルな表現を出そうと思った場合、以下の設定を推奨します。またHiresは必ず使いましょう。 |
|
|
|
If you are trying to produce a simple expression with this model, the following settings are recommended. Also, be sure to use Hires. |
|
|
|
|
|
|
|
Sampler: DPM++ 2M (or DPM++ SDE) |
|
CFG scale: 4-6 |
|
Hires upscaler: 4x_RealisticRescaler_100000_G (or ESRGAN4x AnimeVideo) |
|
Steps:over 16 |
|
Hires Steps: 0 or half of Steps （少ないほどいいかも） |
|
Hires Denoise: 0.6 |
|
|
|
|
|
|
|
この画像を基準にして、色々と変更してみます。この画像以下からは未修正です（指など）。 |
|
|
|
Using this image as a basis, make various changes. Unmodified from below this image (fingers, etc.). |
|
|
|
|
|
|
|
> 1girl, dynamic pose, dynamic angle, cinematic shadow effect, looking at viewer, cygames mecha_musume, Transformation by mechanization, demon lord girl, (kawaii:1.2), elf ears, demon horns, (metalic bodysuit:1.2), ([Mechanical exoskeleton:boned armor:0.5]:1.0), (glowing mechanical wings:1.2),((floating_armor)), female focus, beautiful detailed eyes, post apocalypse, subtle ambient glow, global illumination, ultimate quallity, |
|
|
|
 |
|
|
|
|
|
|
|
ここからupscalerをlatentに変更、CFGは3に変更。明瞭度がCFG効果で下がりましたが、ディテールは向上しました。はっきりとした絵、または光沢を強くする場合は、CFG7-8を推奨します。9以上は唇辺りの描写が悪くなる場合があります。 |
|
|
|
From here, upscaler was changed to latent and CFG was changed to 3. The clarity was reduced by the CFG effect, but the detail was improved. CFG 7-8 is recommended for a clear picture or for a high gloss effect. 9 or higher may result in poor rendering around the lips. |
|
|
|
 |
|
|
|
|
|
|
|
またSamplerをUniPC、3-5のCFGにして、Hires StepsをStepsの値以下にすることで、3DCGやフィギュアみたいな雰囲気になります。逆にHires Stepsの値を増やせば増やすほど書き込みが細かくなります。 |
|
|
|
Also, by setting Sampler to UniPC, 3-5 CFG, and setting the Hires Steps to a value less than the Steps value, you can create a 3DCG or figure-like atmosphere. Conversely, the more you increase the value of Hires Steps, the more detailed the writing will be. |
|
|
|
 |
|
|
|
|
|
|
|
他にも光沢を美しくするような設定を探しながらモデルをマージしていたら、ここ２週間ほどで、画像表現における２つのゲームチェンジャーが登場しました。（AI業界ではよくあることです） |
|
|
|
I was merging models while looking for other settings that would beautify the gloss, and in the last two weeks or so, two game changers in image representation have emerged. (so often the case in the AI industry LMAO) |
|
|
|
|
|
|
|
１つはextensionの [Self Attention Guidance](https://github.com/SusungHong/Self-Attention-Guidance) です。これは強度を0.75から1にすることで、明瞭度を上げたような画像にすることができます。これは写真撮影で、F値で被写界深度を調整する方法に似ています。作例ではシンプルな背景のままでしたが、背景の書き込みも向上する場合があります。 |
|
|
|
One is the extension [Self Attention Guidance](https://github.com/SusungHong/Self-Attention-Guidance). This can be done by setting the intensity from 0.75 to 1, which makes the image look like it has increased clarity. This is similar to how depth of field is adjusted with f-stop in photography. In the example, the simple background was left as is, but the background writing may be improved as well. |
|
|
|
 |
|
|
|
|
|
|
|
もう１つはiroiro-loraの[flat2](https://huggingface.co/2vXpSwA7/iroiro-lora/tree/main/test2) です。このLoRAをマイナスで適用することにより、ディテール・書き込みを増大させることができます。この例では上記SAGに加え、flat2:-1を付与しています。 |
|
|
|
The other is iroiro-lora's [flat2](https://huggingface.co/2vXpSwA7/iroiro-lora/tree/main/test2). This LoRA can be applied in the negative to increase detail/writing. In this example, in addition to the above SAG, we have given it a flat2:-1. |
|
|
|
 |
|
|
|
|
|
この２つのトピックは互いにシナジーがあります。他のプロンプトでも、衣装のシワや髪の毛のディテールが向上します。tiled diffusionなども効果が見込めるはずです。どれも私のお気に入り！ ぜひ使ってみて、そのディテールを体感してみてください！ |
|
|
|
These two topics have synergy with each other. Other prompts such as costume wrinkles and hair detailing could be improved, tiled diffusion should also work, etc. All of these are my favorites! Please try them and experience the detail !! |
|
|
|
> highest quality, (trending on artstation), poster illustration, (2girls:1.2), (a High Quality Victorian Era cute girl), (luxurious intricate-harem dancer outfit with pelvic curtain:1.2), (cameltoe:1.2), anklet, (neck ornament and cleavage and covered nipples), bob cut, (intricate halter backless dress:1.2), sideswept hair, pixie cut, (see-through_dress:1.2), (beautiful face), (sexy eyes), detailed face, detailed eyes, slight smile, glitter many particles, <lora:flat2:-1>, |
|
|
|
 |
|
|
|
|
|
これらを書いた後に気づいたことなのですが、UniPCに限り、CFG1.5が実用レベルになります。あなたの聞き間違えではありません。CFG1.5です。SAGを1、flat2を-1にして使ってください。 |
|
|
|
I realized after writing these that CFG1.5 is the practical level only for UniPC. You are not hearing me wrong, it is CFG1.5, use it with SAG set to 1 and flat2 set to -1. |
|
|
|
 |
|
|
|
|
|
## misc. |
|
|
|
最後に、このモデルと、このモデルで生成されたメディアを使用したことに起因する責任は、すべて利用者のものになります。 |
|
|
|
NSFWのメディアに関して生成できるかは試しておりません ╮( -᷄ ᴗ -᷅ )╭。裸体などは各掲載媒体においてAIで検知される場合が多いため、レギュレーションや各国の法律に注意しながら利用してください。 |
|
|
|
Finally, any liability arising from the use of this model and the media generated by this model is solely that of the user. |
|
|
|
We have not tested whether it can be generated for NSFW media. Nudity, etc., ¯//_(ツ)_/¯ is often detected by AI in each publication, so please be aware of regulations and the laws of your country when using this service. |
|
|
|
|
|
|
|
しかし、肌色が少なくても、あなたとあなたの性癖を駆使することで、セクシーでマスターピースなアートワークを生み出すことが可能です。 |
|
|
|
そう、LATEXならね。 |
|
|
|
But even with less skin tone, you and your proclivities can create sexy, and masterpiece artwork. |
|
|
|
Only on the LATEX. |
|
|
|
|
|
|
|
> standing, wide angle, (1girl:1.0), corruption, (dark persona magical girl:1.2), colorful hair, colorful clothes, dynamic poses, wide angle, latex gloves, highleg playboy bunny suit, showgirl skirt, glossy detailed pantyhose, large eyes big, cute beautiful eyes, effect in dark background, high quality,subtle ambient glow, raytracing, global illumination, ultimate quallity, <lora:flat2:-0.6>, <lora:wowifier_v2:0.5>, <lora:Villagesuccubus_v30:0.8:NCNF> |
|
|
|
 |
|
|
|
# specs |
|
|
|
## used models (OpenRAIL-M) |
|
|
|
- latexGenerator15 |
|
- aoaokoPVCStyleModel_pvcAOAOKO |
|
- ambientgrapemixForAnime2D |
|
- amanomix |
|
- BalorMix |
|
- dorayakimix |
|
- stylized3DModelLora |
|
- milfWeightEdit |
|
- bigHardLatex_v10 |
|
- DefbeaCF3AR_mix_v2 |
|
|
|
|
|
## licence |
|
|
|
**CreativeML OpenRAIL-M** |
|
|
|
❌ 本モデルを商用の画像生成サービスで利用する行為 <br> |
|
Use of this model for commercial image generation services |
|
|
|
❌ 本モデルや本モデルをマージしたモデルを販売する行為<br> |
|
The act of selling this model or a model merged with this model |
|
|
|
❌ 本モデルを使用し意図的に違法な出力をする行為 <br> |
|
Intentionally using this model to produce illegal output |
|
|
|
❌ 本モデルをマージしたモデルに異なる権限を与える行為 <br> |
|
Have different permissions when sharing |
|
|
|
❌ 本モデルをマージしたモデルを配布または本モデルを再配布した際に同じ使用制限を含め、CreativeML OpenRAIL-M のコピーをすべてのユーザーと共有しない行為 <br> |
|
The act of not sharing a copy of CreativeML OpenRAIL-M with all users, including the same usage restrictions when distributing or redistributing a merged model of this model. |
|
|
|
✅ 本モデルで生成した画像を商用利用する行為 <br> |
|
Commercial use of images generated by this model |
|
|
|
✅ 本モデルを使用したマージモデルを使用または再配布する行為 <br> |
|
Use or redistribution of merged models using this model |
|
|
|
✅ 本モデルのクレジット表記をせずに使用する行為 <br> |
|
Use of this model without crediting the model |
|
|
