It can make a live-action person's face resemble an Asian person in an animated style.

that it be used in conjunction with other facial LoRA.

I recommend that it be used in conjunction with other facial LoRA.

weights: 0.6~0.8.

no trigger word. (1girl)

A LoHa made using a 30 image dataset and my own methods. 15 batch size.

This was a test after being requested, it uses somebody else's dataset that was left untagged save for the triggerword. Uses my own method. 824 images 15 batch size.

A LoHa trained with 12 images. 1 batch size. Uses my own method.

A LoHa trained on a measly 10 images using my method. 1 batch size.

If you'd like to support me and do more:
https://ko-fi.com/edgonkofi - commissions open!
If you're looking for a >>LoRA Making Tutorial<<

Urban Lingerie

We all like hoodies and lingerie.

Why not combine urban wear and lingerie then?

Get creative and mix prompts from my Urban Gal and KPG models!

This is a LoHa model trained from 123 various sources images. The tests found that portraits performed better,which could create a very dynamic expression,and that's where the name comes from.

You can use the LoHa to design eyes & face.

Weight 0.9-1.0 will perform better

Of course the tirgger word is still 光头强lora

Support my video here:https://www.bilibili.com/video/av823626310

My QQ group:https://jq.qq.com/?_wv=1027&k=o89HEhaS

It's a loha model trained wtih Pixiv painter opinew6.

The cute and vivid style always hits my heart straightly.

So I'd like to share it with you guys all.

It works well when mixed with other Loras, so you can just create different stylized images of your own Lora.

Weight from 0.6 to 1.0 works ok.

目前使用权重在0.6到1.0均可。

You should refer it in prompt words instead of Additional Network.

注意loha模型不能够在Additional Network 中使用，你应当在prompt中引用。

这是一个画风LoCon，

This is an artstyle LoCon,

目前看来不是很稳定，当我学习了新的训练方法，我会更新这个LoCon。

It doesn't seem very stable at the moment, I will update this LoCon once I learn new training methods.

注意：触发词为：chibi，

Note: The trigger phrase is: chibi,

推荐配合角色lora使用。

I recommend using it with a character Lora.

推荐把这个LoCon的强度设定为1，角色lora的强度设定为0.8，把CFG设置为7，使用 anything-v3进行图片生成。

Setting the weight of this LoCon to 1 and the weight of the character Lora to 0.8，set CFG to 7,use https://civitai.com/models/66/anything-v3 。

根据实际情况进行调整以获得更好的画面效果。

Make adjustments according to the actual situation to achieve better visual effects.

如果有画风训练相关的心得，欢迎分享给我。

If you have any insights on artstyle training, please feel free to share them with me,

以便我能修改这个LoCon.

so that I can modify this LoCon.

• Weight: 1.0 - 1.1

• CFG Scale: 5.5 - 7

• Recommended Prompts: manaka nemu, brown hair, purple eyes, blunt bangs, school uniform, neckerchief, hair bun, sidelocks, long hair

Buy us a cup of coffee and we’ll spill the beans.

https://www.buymeacoffee.com/CoSubcreation

Or Fanbox

https://co-subcreation.fanbox.cc/

Or爱发电

https://afdian.net/a/Co_Sub_creation

Your support is the fuel that powers our world-building. Your ideas are the bricks that shape our world-making!

あなたの応援で、世界は生まれ変わる！あなたのアイデアが、世界を作り上げる原動力になるよ！

您的支持将成为创造世界的动力，您的想法将成为构建世界的基石！

If you have any ideas, suggestions, or questions during use, you can ask them on Discord ↓

ご利用中に何かアイデアや提案、質問がございましたら、Discordにてお問い合わせいただけますようお願いいたします↓

使用中有任何想法&建议&问题，可以去Discord提问↓

https://discord.gg/tVpJMhjUwx

If you’re new to LoHa, please read this:

This model was trained with LoHa, so you need to install the WebUI extension. Trust me, it’s worth it.😉

https://github.com/KohakuBlueleaf/LyCORIS

LoHa model path: models/Lora

Usage: Click on the 🎴 button under “Generate” in the WebUI, choose the Lora tab, select the model, and adjust the weights. It’s as easy as pie.🍰

第一次使用LoHa请看：

本模型训练使用了LoHa，需要安装WebUI的扩展。

https://github.com/KohakuBlueleaf/LyCORIS

LoHa模型路径: models/Lora

使用：点击WebUI“生成”下面的🎴，选择Lora 选项卡，选择模型，更改权重。

LoHaを初めて使う場合は、こちらをご覧ください：

このモデルトレーニングにはLoHaが使用されており、WebUIの拡張機能をインストールする必要があります。

https://github.com/KohakuBlueleaf/LyCORIS

LoHaモデルのパス：models/Lora

使い方：WebUIの「生成」の下の🎴をクリックし、Loraタブを選択して、モデルを選択して、重みを変更してください。

Recommended parameters:

推奨パラメータ：

推荐参数：

Model: blueboys2D_v10

Vae: kl-f8-anime2

Clip skip: 2

weight:0.7~0.9

Sampler: DPM++ 2M Karras

Steps: 25

Size: 512x512|384x640|640x384

CFG Scale: 9 or 14~17(see Instructions: need extension)

Extension setting↓

Enable Dynamic Thresholding (CFG Scale Fix)

Mimic CFG Scale:7

Advanced Options▼

Top percentile of latents to clamp:100

Mimic Scale Scheduler:Constant

Minimum value of the Mimic Scale Scheduler:0

CFG Scale Scheduler:Half Cosine Up

Minimum value of the CFG Scale Scheduler:3

Preferred Prompts:

推奨TAG：

推荐提示词：

Prompt:

1boy,solo,upper body,fata, <lora:Style_GlassyandGlaze (FATA)_v1.0:1:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1>(see Instructions)

Negative Prompt:

EasyNegative, badhandv4

EasyNegative | Stable Diffusion TextualInversion | Civitai

badhandv4 - AnimeIllustDiffusion | Stable Diffusion TextualInversion | Civitai

Instructions:

Pulling the CFG too high can lead to flower patterns on your artwork, but fear not! You can install the Dynamic Thresholding (CFG Scale Fix) plugin to fix the issue:

https://github.com/mcmonkeyprojects/sd-dynamic-thresholding

We recommend increasing the weight, but be warned, it may lead to overfitting. We suggest using the Lora block weight extension to adjust the layering (you can use the recommended values directly from the prompt):

https://github.com/hako-mikan/sd-webui-lora-block-weight

It works well with some Sairu and Hira-nuri base models, but since the model was trained on Blueboys2D, we recommend sticking with the original base model for the best results. It does not support pastel colors, thick paint styles, 2.5D, or 3D base models.

It's great for drawing upper bodies and faces, but drawing full-body images may cause distorted faces, so you'll need to import the image into the image-to-image local redraw tool.

The default hair color is black and the hairstyle is an M-shape. To change the hair color, hairstyle, or eye color, you'll need to add specific prompts (as shown in the example image).

I don't use high-definition restoration because running multiple HD images at once during the lottery stage can't guarantee high-quality results. Instead, I prefer to batch process several good-quality images using the txt2img tool (usually with good composition), then use img2img to increase the resolution of individual images and determine the level of detail to add by adjusting the denoising strength. Finally, for those who really need HD images, I'll use Extras to generate high-definition restoration images at the end.

説明：

CFGの値を高くしすぎると、花模様が発生することがあります。この問題を修正するために、Dynamic Thresholding (CFG Scale Fix)プラグインをインストールすることをお勧めします。

https://github.com/mcmonkeyprojects/sd-dynamic-thresholding

重みは少し高めに設定することをお勧めしますが、オーバーフィッティングが発生する可能性があります。分層を調整するには、lora block weight拡張機能を使用することをお勧めします（推奨ヒントの数値を直接使用できます）。

https://github.com/hako-mikan/sd-webui-lora-block-weight

セルルロース、平塗りタイプの背景をサポートしていますが、学習モデルの背景はblueboys2Dであり、より良い効果を得るためには、元の背景を使用することをお勧めします。また、粉彩系、厚塗り系、2.5D、3Dタイプの背景はサポートされていません。

顔や上半身の描写には慣れていますが、全身の描写は難しいため、一部分を描いた後、txt2imgで画像を再構築する必要があります。

デフォルトの髪色は黒色で、髪型はM字型です。髪色、髪型、瞳の色を変更したい場合は、適切な指示語を追加する必要があります（例えば、図のように）。

抽選フェーズで多数の画像を一度に生成すると、高品質の画像を保証することができません。通常、私は画像の品質が良好で構図が適切な画像を一括生成し、img2imgを使用して単一の画像の解像度を高く調整し、Denoising strengthを調整して詳細度を調整します。最後に、Extrasで高品質の修正画像を生成することができます。

说明：

CFG拉的过高会导致花图，可以安装插件Dynamic Thresholding (CFG Scale Fix)来修复：

https://github.com/mcmonkeyprojects/sd-dynamic-thresholding

权重建议高一点，但可能会导致过拟合，建议使用扩展lora block weight来调整分层（可以直接使用推荐提示词里的数值）。

https://github.com/hako-mikan/sd-webui-lora-block-weight

这里是大佬的一些关于这个插件的教程和预设@大江户战士

https://www.bilibili.com/video/BV1Wv4y157NH/

支持一些赛璐璐、平涂类底模，但由于训练的底模就是blueboys2D，所以如果您想得到较好的效果，还是建议使用原底模。不支持粉彩系、厚涂系、2.5d、三次元底模。

擅长画上半身和脸，画全身容易崩脸，需要导进图to图局部重绘。

默认发色是黑色，发型是m形，想要更改发色、发型和瞳色需要加提示词（如例图）。

我没有使用高清修复是因为，在抽卡阶段一次性跑多张无法保证画面质量的高清图比较浪费时间，通常我比较喜欢在txt2img批量跑出还不错的图后（通常是构图比较好），转img2img直接拉高分辨率跑单张，并通过调整Denoising strength来决定添加细节的程度，最终跑出符合预期的图。如果确实需要高清图，在最后才会拉到Extras生成高清修复图。

In fact, you can also use img2img to generate beautiful anime girls, for example: ↓

実は、img2imgを使って美しいアニメの女の子を生成することもできます。例えば、↓

其实它也可以用img2img来生成好看的妹子，比如↓

This is a costume LoHa. Not character LoHa.

weight: 0.6-0.7

Walkure Romanze style

Can generate knight armor with pleated skirt and school uniform of Walkure Romanze.

No trigger words are set for the characters.

Samples are generated by t2i -> i2i -> Tiled Diffusion+Tiled VAE. Some prompts were deleted by Tiled Diffusion+Tiled VAE process, so the prompt of the samples is incomplete.

Samples full prompt:

1. (masterpiece, best quality,extreme detailed beautiful eyes), best illumination, perfect lighting,look at viewer, (1 girl, solo,street, upper body,dynamic angle),(victory pose:1.2), (smile, shiny_eyes,large breasts),sidelocks,long hair, armor, (white hair:1.5), (ponytail,golden eyes:1.4),breastplate, red skirt, thighhighs, greaves, gauntlets, lace_trim,<lora:WalkureRomanze_lycoris2_bf16-000025:0.6>

2. (masterpiece, best quality,extreme detailed beautiful eyes), best illumination, perfect lighting,look at viewer, (1 girl, solo,street, upper body,dynamic angle),(spread legs:1.2), (smile, shiny_eyes,large breasts),sidelocks,long hair, , (white hair:1.4),(ponytail:1.5), (golden eyes),white competition swimsuit, thighhighs,black greaves,black gauntlets, lace_trim,<lora:WalkureRomanze_lycoris2_bf16-000025:0.6>

3. (masterpiece, best quality,extreme detailed beautiful eyes), best illumination, perfect lighting,look at viewer, (1 girl, solo,bedroom, upper body,dynamic angle),(spread legs:1.2), (smile, shiny_eyes,large breasts),sidelocks,long hair, (white hair:1.4),(ponytail:1.5), (golden eyes:1.3),military_uniform, thighhighs, aiguillette, epaulettes,pleated_skirt, zettai_ryouiki,panties, <lora:WalkureRomanze_lycoris2_bf16-000025:0.4>

This is retrained from thesauceer dataset, just a version of his Tutu LoRA, please check out his original work! He was kind enough to share his dataset.

This is a LoHa LoRA! You need the extension!
Why? LoHa have virtually no effect in output art style and they do a lot better with multi-concepts.

Trigger words for V2.20:

positive: <lora:Tutuv220:0.5>(tutu)

You can also try "ribbon-trimmed skirt", "layered skirt" and "miniskirt".
If failing to make full shirt, try adding either "leotard" or "tank top" or "covered navel" to prompt.

Keep in mind you might have to adjust LoRA and/or tag weight for it to appear, defaults to "ribbon-trimmed skirt" most of the time! Dataset wasn't an exactly split between the different concepts.

Follow the steps in the Model Version notes to reproduce my images.

Sample images generated with Hassaku.

All negative embeds should be easy to find. and Prompts should be included in the sample images! Click the little icon in the bottom right to see it.

It was trained on 62 512x512 images provided by thesauceer, 20 epochs using Animefull-final-pruned for the Training Model.

Danbooru tags generated by Waifu Diffusion 1.4 Tagger.

LyCORIS of Jarosław Kaczyński

LyCORIS of Janusz Korwin-Mikke

Weights: 0.8-1.4

This is an experimental and somewhat unstable attempt at squeezing 9 different versions of Shinobu Oshino/Kiss-shot from the Monogatari series into one model. Getting desired results may be difficult, please check the preview images for the most reliable prompts for each version of the character!

Versions and tokens:

• Adult Kiss-shot | kiss-shot acerola-orion heart-under-blade

• Teen Kiss-shot (Kizumonogatari) | kizu_teen

• Teen Kiss-shot (TV) | tv_teen

• White dress (Kizumonogatari) | kizu_long

• Pink dress/short hair (Kizumonogatari) | kizu_short

• White dress (TV) | tv_white

• Pink dress (TV) | tv_pink

• School uniform (TV) | tv_student

• Gothic dress (TV) | tv_gothic

Again, check the preview images for example prompts!

LyCORIS of our boy Asmongold.

Suggested weights: 0.8 - 1.4

LyCORIS of Antony Starr as Homelander.

This was trained on purely Homelander pictures by design. Still, it is sometimes possible to get other characters.

Suggested weights: 0.8 - 1.4

LoCon trained on the artstyle of kojima ayami, who did the art for a lot of Castlevania games.

It is not quite her style but still looks good enough on it's own.

Trigger word is SotnStyle.

I recommend weight 0.7~0.8, or 0.9~1 for higher resolutions.

Trained on NAI.

LyCORIS of Thomas Jane

Suggested weights: 0.8 - 1.4

A Character LoCon for Niki Shiina (椎名ニキ) 【Ensemble Stars!】

Weight:0.4～0.7

Hires. fix

trigger words: shiina niki\(ensemble, stars!\)

※以下日本語で失礼します。

運要素が絡みますがぽい衣装を出現させることは確認しています。

共通呪文:shiina niki\(ensemble, stars!\), long hair, bangs, blue eyes, shirt, grey hair, hair between eyes, ponytail

Crazy:B衣装

出現率アップ呪文: jacket, choker, belt, necklace, black pants, open jacket, white shirt, white jacket, Crazy:B jacket, Crazy:B Uniform

ホットリミット衣装

出現率アップ呪文: navel, jacket, purple pants, necklace, ring, tank top, yellow jacket, open jacket, open clothes, hot limit_hot summer

ホットリミット衣装 (SCR)

出現率アップ呪文:navel, jacket, purple pants, necklace, ring, tank top, purple jacket, open jacket, open clothes, hot limit_hot summer

Please see https://huggingface.co/alea31415/bofuri-full for details

It seems that many people are not aware of my bofuri model so I extract a locon as usual for easier use.

The model is trained on top of ACertainty with clip skip 1.

Trigger words

• BoDrag

• BoDread

• BoFrederica

• BoIz

• BoKanade

• BoKasumi

• BoKasumi sarashi

• BoKuromu

• BoMaple

• BoMaple angel form

• BoMaple beginner outfit

• BoMaple black armor

• BoMaple machinegod form

• BoMaple pajama

• BoMaple sheep form

• BoMaple uniform

• BoMarcus

• BoMay

• BoMii

• BoMisery

• BoPain

• BoSally

• BoSally cape

• BoSally pajama

• BoSally turtleneck scarf

• BoSally uniform

• BoShin

• BoYui

• Dorazou

• Fox Oboro

• Monster Atrocity

• Turtle Syrup

• Hydra

Anything + dataset(FineTune)

I've been thinking about things lately, you know, like the things we think about when we're not really thinking about anything.

This is still an early version and will be improved in the future.

Enjoy!

Contact → My Twitter

Note: This is a Locon. Install the extension to get it to work. I've found them to do better when working on ideas that I want easily added to one of my LORA characters. Also, they're WAY less damaging to images, even at high weights.

I had made a couple NSFW style Locons (and I have a few more in development with sneak peaks here), but I don't always want my waifu's to become horny and evil. So. Here I am combating the evil part with an Angel (warning, still can be horny though). It was trained on animefull_final with 29 images with the usual self-pruning and tagging and a wide range of different images. I like keeping my prompts simple for the first few pictures so that you can see it without a whole bunch of mumbo jumbo. Make your pictures better than mine though!

What does it do you ask? A Locon that will transform your waifu into an angelic form. Usually with wings. I leave these Locon's very general, so if you want to push a specific direction with your images (like a halo/wings), I'd tag it. Otherwise it'll be random on what form it takes.

How to use it: I've found that with these Locon's I've been submitting, weight's can be wide and varied. 0.5 is fine is you want only a small effect. But a lot of time you can go as high as 1.5 if you want to really push the LORA you're changing.

Also the trigger word is:

AngelT - Only trigger word currently added.

Optional - Of course you can add other words that describe angels if you want specific components and to make sure they get into the picture

Extra, Extra!: Should be well trained and usable in most models. Images shown for example.

Picture Descriptions:

1-3: AnythingV3 using just this Lora. Showing different types of outfits can be called when just using this.

4: AnythingV3 combo'd with my Seraphina (Disgaea 5) Lora.

5: AnythingV3 combo'd with my in progress Ghislaine (Mushoku Tensei) Lora.

6: AnythingV3 combo'd with my in progress V2Corruption Locon.

7: AnythingV3 combo'd with my in progress Brainwashing Locon.

8: AnythingV3 combo'd with my in progress "Symbioted Taking Over" Locon (basically a symbiote merging with the user).

9: HassakuHentaiModel just with the Lora.

10: AbyssOrange with just the Lora.

11: PerfectWorld using just this Lora.

What's next for this? Unsure. Feedback will be the decider if I do a V2.

What's next that isn't this?

More character Loras. (Ghislaine from Mushoku Tensei, Conqueror Lilias from E7, Rozalin from Disgaea 2, Tharja from Fire Emblem, Catherine from Catherine Full Body, Miku + Piscalat from Idaten Deities, Kuribayashi Shino from GATE, Rin Sohma from Fruits Basket, "Kasumi" from P5, Ann from P5, Ikumi Mito from Food Wars, Alice Nakiri from Food Wars, Chika Fujiwara from Kaguya-sama, Lulu from FF10, Momo from My Hero, Vivy from Vivy, Dominique de Sade from Vanitas, and Dos from One Punch Man.) And maybe Wilhelm van Astrea, my favorite Re:Zero husbando or a Prinny from Disgaea.

V2's of some of the previous Loras. Specifically make them more diversely acceptable with other models and LORA's.

I also have some more quite NSFW Locon's. Specifically, being pet play, being stepped on, brainwashing, corruptedv2, and one that might be too much for me to even upload here >.<.

And, uh, I want to make a few "normal" Locon's so I have an "knight-ify" your character one coming up. And a forest fairy one as well. And a soldier one. And a barbarian one.

Or just give me some random suggestion or priority opinions. That works too.

Two versions, soft and hard.

enjoy yourself!

• 欢迎加入AI绘画模型博物馆QQ群

• ①群：791222249

• ②群：599962550

Locon trained on 81 cum filled shoes.

Performed well when drawing shoes only.

Have a great tendency of making any shoes into high heels.

A Locon of ba-zerarudo from Frame Arm Girls.

Trained Trigger words:

ba-zerarudo | leotard | pantyhose | rabbit tail | high heels

Pictures below are generated on Any4.5 with Controlnet.

BOOOOBA Eldridge :D

Trained trigger words:

Eldridge

default outfit | Kitty idol | Holy Night's Embrace | The Inquisitive Recluse | New Year's Outing

The "New Year's Outing" is overfitted and may have poor performance.

I trained this LyCORIS by collecting about 200 images generated by NijiJourney V5 on the internet.

Set weights to 0.7~0.8. I generated the example images with AOM3.

loha training just for fun

Hires step 20

Denoising 0.6

This model was made trained on 160 images made by Chelodoy, most of theme were NSFW, but the model works flawlessly with any type of image, it is black magic.

Trigger word is chelodoystyle. But it is unnecessary to use it. Model should work just fine at 1 weight.

A Character LoCon for rinne amagi (天城燐音) 【Ensemble Stars!】

Weight:0.4～0.7

Hires. fix

trigger words: amagi rinne\(ensemble stars!\)

※以下日本語で失礼します。

運要素が絡みますがぽい衣装を出現させることは確認しています。

共通呪文:amagi rinne\(ensemble stars!\), red hair, short hair, blue eyes, hair between eyes, jewelry, earrings, ring

※hair between eyesは私服を再現しようと思う時に邪魔になります

Crazy:B衣装

出現率アップ呪文: jewelry, jacket, black pants, open jacket, chain, ring, white shirt, white jacket, Crazy:B jacket, Crazy:B Uniform

私服（ヘッドバンド）

出現率アップ呪文:headband, black shirt, pants

ラッキーチェイン（フィーチャースカウト燐音編衣装）

出現率アップ呪文:jacket, necklace, off shoulder, tank top, torn clothes, fur trim, black shirt, black pants, ring, fur-trimmed jacket, red belt, chain necklace

Flambé！衣装

出現率アップ呪文:gloves, hat, jewelry, jacket, earrings, black gloves, pants, fingerless gloves, collar, black jacket, torn pants, black pants, ring, peaked cap, single glove, torn pants, military cap

A dream style LoRA for Stable Diffusion webui.
一个梦幻风格的LoRA。

Girl in the Flask (瓶中少女）

An Utility/Artstyle LoCon

You can use this Locon model to bottle up the characters you want and take them with you.

你可以使用这个locon模型可以让你把你想要的角色装进瓶子里带走。

• Recommeneded prompts: simple background, full body, air bubble, water,submerged (Please check my sample work for more details)

• Trigger word: in bottle,in container

• You can use this LoCon with other characters' LoRA.

• Recommeneded Weight: 0.6-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5, Dan mix

• This LoCon may have some small issues, I will fix it soon.

• How to use LoCon/LyCORIS? Please read instruction of a1111-sd-webui-locon.

Human kind can not gain anything without first giving something in return. To obtain something of equal value must be lost. - Fullmetal Alchemist

If you like my work,please give me a like <3.

如果你喜欢我的模型请给我一个赞 <3.

もし、私のモデルが気に入ったら、「いいね！」をクリックしてください<3。

::: 该页面为中国传统纹样loha的使用介绍，关于详细的训练过程，分层调试方法，可以到我们的wiki上去查看。https://aigc.ioclab.com/sd-showcase/chinese-ornament.html :::

中国传统纹样lora by IoC Lab

训练及数据：@shichen @ciaochaos

版本 介绍 2023.4.1 上传lora，介绍内容（本篇） 2023.4.4 更新训练方法（wiki待上传）

在图片中你可以看到有大约25个以上的版本，我挑选了其中最有特点的几个版本，你可以在version中挑选下载，或者前往huggingface下载全部的版本

https://huggingface.co/ioclab/ioc-lab-models/tree/main

使用方法：

在训练集的标注中，分为以下五个语段：

traditional chinese pattern

with circular symmetrical composition
with square symmetrical composition
with special-shaped symmetrical composition

red and blue color scheme

featuring horses and clouds
featuring flowers and branches

a flower in the center

所以你可以围绕着这五种固定句式进行扩充，如需要作为背景，可以调低权重，增加 1 girl

分层拆解：

在分层图片中可见，在高权重下，通过仅开启 midd，outd 等 lora 层的权重，可以做到相比于直接降低权重，更能保留纹样风格的人物背景图。

推荐模型版本介绍：

这里是一些效果比较特色的模型版本，但出图效果与参数，种子等都有关系，你可以尽情尝试，期待你的评论

特色模型版本

以下是一些特色模型版本：

• 4：特点是纹样更卡通化，色彩更缓和

• 7：特点是对关键词的还原更好，更符合数据集的倾向

• 10：特点是更对称，同时构图倾向会偏向于非中心对称构图

• 13：特点是相比10的颜色会更缓和，高权重下会出现第三种散布构图

• 15：特点是高饱和度和高横向对称构图

• 16：特点是高饱和度+散布构图

人物配合版本

以下是容易和人物配合的版本：

• loha01_old3: 更容易出卡通光影，构图和花纹装饰性强

• loha04_old3: 构图容易出全身，光影保留效果好

• loha06_old1: 对构图限制小，人物动作自然

• loha02_old2: 背景层次更多，对称性强

数据集标注及loha训练方法，见wiki链接

https://aigc.ioclab.com/sd-showcase/chinese-ornament.html

Chinese Ornament Wiki in English

::: This page is an introduction to the use of traditional Chinese pattern Lora, for detailed training processes and layer debugging methods, please refer to our wiki. https://aigc.ioclab.com/sd-showcase/chinese-ornament.html :::

Chinese traditional pattern Lora by IoC Lab

Training and Data: @shichen @ciaochaos

Version Description 2023.4.1 Uploaded Lora, Introduction Content (This Article) 2023.4.4 Updated Training Method (Wiki To Be Uploaded)

In the pictures, you can see more than 25 versions. I selected the most distinctive versions for you to download, or you can go to Huggingface to download all versions.

ioclab/IocLab_Models at main (huggingface.co)

How to use:

In the annotation of the training set, the following five language segments are divided:

traditional chinese pattern

with circular symmetrical composition
with square symmetrical composition
with special-shaped symmetrical composition

red and blue color scheme

featuring horses and clouds
featuring flowers and branches

a flower in the center

So you can expand around these five fixed sentence patterns. If you need to use it as a background, you can reduce the weight and add one girl.

Layered Decomposition:

As shown in the layered images, under high weight, by only turning on the weights of midd, outd, and other Lora layers, you can achieve a person-background image with a style more retained than directly reducing the weight.

Recommended Model Version Introduction:

Here are some model versions with distinctive effects, but the output effect is related to parameters, seeds, etc. You can try them out and we look forward to your comments.

Distinctive Model Versions

Here are some distinctive model versions:

• 4: The characteristic is that the pattern is more cartoonish and the color is more gentle.

• 7: The characteristic is that it better restores the keywords and is more in line with the tendency of the dataset.

• 10: The characteristic is that it is more symmetrical, and the composition tends to be non-centrally symmetrical.

• 13: The characteristic is that compared to 10, the color will be more gentle, and the third scattered composition will appear under high weight.

• 15: The characteristic is high saturation and high horizontal symmetry composition.

• 16: The characteristic is high saturation + scattered composition.

Character Matching Versions

The following are versions that are easy to match with characters:

• loha01_old3: More likely to produce cartoon light and shadow, with strong composition and decorative patterns.

• loha04_old3: The composition is easy to produce the whole body, and the effect of retaining light and shadow is good.

• loha06_old1: The restriction on composition is small, and the character's movements are natural.

• loha02_old2: More layers in the background, strong symmetry.

For dataset annotation and Lora training methods, please refer to the wiki link.

https://aigc.ioclab.com/sd-showcase/chinese-ornament.html

Weight : 0.4～0.6?

The model for the illustration for the sample is based on 7th_anime_v3_A and has been merged several times, including AOM3.

7th_anime_v3_A will be a close illustration.

Please use it in an environment where LoCon is available.

AnyLoRA is used as the model for the study.

A Character LoCon for Ijichi Seika(伊地知 星歌) | Bocchi the Rock!

She may not be as popular as other characters in Bocchi the Rock, but I still like her.

• Recommeneded prompts: red eyes,long hair,ahoge,blonde hair,black choker,off shoulder,bangs (Check my sample pics for detail)

• Trigger word: red eyes, blonde hair

• Recommeneded Weight: 0.6-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5, DanMix

• How to use LoCon/LyCORIS? Please read instruction of a1111-sd-webui-locon.

• This LoCon still have some small issues, I may fix it later.

If you like my work,please give me a like <3.

如果你喜欢我的模型请给我一个赞 <3.

もし、私のモデルが気に入ったら、「いいね！」をクリックしてください<3。

「貴方が今も、月の光を忘れぬように……」

• This is a lycoris model, it requires the locon extensions to be installed.

• The trigger words for alternative costumes are still under active testing and may not be stable

• Recommened prompts: 1girl, solo, chain necklace\(arcueid\), black pantyhose, white long sleeves turtleneck sweater, blue pleated skirt, looking at viewer, brown footwear, knees boots

• Weight 0.7-0.9 works best for me.

• CFG Scale: 5.5-7

Model based on Erkerut art style. Works best with MeinaHentai (v2 used for example images) at 0.9.

个人感觉八宝备老师画风前后期差异挺大，我选择的是偏后期的画风。

我使用anythingv4.5,权重0.5-0.9。最好使用Hires. Fix。

Tigger words使用Happoubi Jin，可以触发画风。

我顺便做了一下几个主要角色，你可以使用Happoubi Jin+角色名召唤她们。

可以召唤的角色有：蓮華Tsukioka Karen，燈露椎Dorothy，亜璃子Alice，篝ノ霧枝Kagarino Kirie，稲森はるInamori Haru，沢渡雫Sawatari Shizuku。其中Dorothy需要结合short hair使用，其他的使用名字即可。召唤角色时候推荐权重0.7-0.9。

I feel that there is a significant difference in the before and after style of Happoubi Jin's artwork, and I chose the processing style.

I used anythingv4.5, with a weight of 0.5-0.9. It is recommended to use Hires. Fix.

Trigger words: Happoubi Jin, which can trigger the art style.

I also made several main characters, and you can summon them using Happoubi Jin + character name. The characters that can be summoned are: Tsukioka Karen, Dorothy, Alice, Kagarino Kirie, Inamori Haru, and Sawatari Shizuku. Dorothy needs to be used with "short hair" to summon, while the others can be summoned using their names. It is recommended to use a weight of 0.7-0.9 when summoning characters.

Hitoimim Style

An Artstyle LoHa

V2b is out!

Important: I only recommend using this LoHa to generate upper body images for the best results.(我只推荐用这个LoHa生成上半身的图像以达到最好的效果.)

• Recommeneded prompts: simple background,upper body,cropped shoulders,hair ornament,ribbon,bow,jewelry (Please check my sample work for more details)

• Trigger word: you don't need any trigger word for this LoHa

• Recommeneded Weight: 0.6-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V4.5, Dan mix

• This LoHa may have some small issue, I will fix it later!

• How to use LoHa/LyCORIS? Please read instruction of a1111-sd-webui-locon.

If you like my work,please give me a like <3.

如果你喜欢我的模型请给我一个赞 <3.

もし、私のモデルが気に入ったら、「いいね！」をクリックしてください<3。

When you don't just want bigger; you want more too. Example images generated from 1 set of 8 using AnythingV4.5

Neo Universe from Uma Musume

300+ Ayaka HQ images dataset

A Character LoCon for mao isara (衣更真緒) 【Ensemble Stars!】

Weight:0.4～0.7

Hires. fix

trigger words: isara mao\(ensemble stars!\)

※以下日本語で失礼します。

運要素が絡みますがぽい衣装を出現させることは確認しています。

Trickstar衣装（!時代）

出現率アップ呪文: plaid_red jacket, white shirt, red_plaid pants, black gloves, purple wristband, jewelry,

Trickstar衣装（!!時代）

出現率アップ呪文: plaid_red jacket, white shirt, black pants, black gloves, fingerless gloves, jewelry,

制服（冬）

出現率アップ呪文:hood, necktie, white shirt, open blazer, blue blazer, school uniform, plaid, pants,

髪下ろし（通常　hair ornament, hairclip, bangs pinned back,　と呪文指定しなくてもヘアピンと前髪をあげてしまいますのでおろしたい場合）

出現率アップ呪文:down hair, hair between eyes

ネガティブにhair ornament, hairclip, bangs pinned back,で多少は髪を下ろしている感じに出来ます

お世辞にも使い勝手が良いと言えないですが使っていただけたら幸いです。

weight: 0.6-0.8

recommend prompt: pauldrons

Benito Mussolini LoCon

IMPORTANT NOTES:

>This model is NOT meant to share or glorify any political view.

>Trigger word: "BenitoMussolini". Recommended weight: 0.7

>Trained on mostly b&w pictures, in order to get good results appropriate prompting and model is required.

>The best results achieved with realistic models (tested with Realistic Vision).

>Probably the easiest way is to just use inpaint.

>This time I’m testing LoCon, feel free to post your feedback.

Prompting

>Highly recommended to add this to your prompt:

Positive: in color

Negative: oversaturated, comics, fake, black and white, b&w, blurry, image noise, simple background, monochrome, (scar)

>Trained mostly on upper body pictures, might be hard to get full body.

>Negative prompts are powerful, use them to your advantage.

>I suggest using 768x768 resolution.

How to make a similar model

I know this model is far from being great, but if someone wants to know the workflow, here you go:

1. Find good images and crop them. 2. Upscale x2~x4 (in my case mostly R-ESRGAN). 3. Remove noise using any graphic software. 4. Scale down. 5. Train LoCon (using Google Colab, e.g. Kohya Trainer). 6. This time I’ve experimented with block weight adjusting (check out sd-webui-lora-block-weight and sd-webui-supermerger awesome tools for LoRa fine tuning)

I mostly use b&w photos, so following steps 2,3,4 leads to significant increase of quality and reduction of image noise.

Also:

>Why do you share such controversial models?

Mainly for fun, also because I can as it’s allowed and I’m not doing anything illegal.

>If you want to use my model anywhere feel free to do so, but it would be nice to put me in credits.

>Check out my models: Adolf Diffusion, Stalin Diffusion

>For any question you can contact me rafikxd123#7133

Note:

This is a LYCORIS/LoCon model, and you need to install a1111-sd-webui-locon extension to use this model.

Suggest weight:around 1

No trigger words,But suggest add chibi in prompt,When used to generate characters.

Based on novel-anime-full model.

(Not chi b i style)

Each trigger word is an independent concept and can be added as needed.

Weight 0.3-0.7

This is a LoHa LoRA! You need the extension!
Why? LoHa have virtually no effect in output art style and they do a lot better with multi-concepts.

Trigger words for V1.20:

positive: <lora:Mermaidv220:1>(mermaid, full body)

You can also try "monster girl", "pussy", "green scales" (or any other color WITH LIMITED results), "head fins"
LoHa is biased towards generating a "pussy" since the entire dataset had one! Use negative "pussy, nsfw" and "sfw" in positives for sfw gens!

Follow the steps in the Model Version notes to reproduce my images.

Sample images generated with Grapefruit.

All negative embeds should be easy to find. and Prompts should be included in the sample images! Click the little icon in the bottom right to see it.

It was trained on 75 512x512 images and 20 epochs using Animefull-final-pruned for the Training Model.

Danbooru tags generated by Waifu Diffusion 1.4 Tagger and manually fine tuned.

This is a LoHa model trained from 100 ノーコピーライトガール images and 100 flat color images. The tone is very close to Walden Blue,and that's where the name comes from.

You can use the LoHa to design hiphop girl,flat color design.

Weight 0.9-1.0 will perform better,also you can try new way to use,lol

2023/4/6 update

Upload Lycoris version. v1.0 version is conventional LoRA model.

Introduction

A conventional LoRA model of KDA Series, including 4 characters, Ahri, Akali, Evelynn and Kai'sa. 0.6~0.8 weights are good. The preview images are generated using 1.6 version (original) YesMix.

Usage

The universal trigger tag is 'k/da \(league of legends\)'.

• For Ahri, the trigger tag is 'ahri \(league of legends\)'. The key features are 'fox girl, fox ears, fox tail, blonde hair, choker, facial mark, whisker markings'.

• For Akali, the trigger tag is 'akali'. The key features are 'baseball cap, choker, cropped jacket, ponytail, mask (optional), kunai (optional)'.

• For Evelynn, the trigger tag is 'evelynn \(league of legends\)'. The key features are 'choker, crop top, feather trim, midriff, hair between eyes, purple hair, claws, sunglasses (optional)'.

• For Kai'sa, the trigger tag is 'kai'sa'. The key features are 'choker, double bun, hair bun'.

See the preview images for details. Btw, this model is trained with some images that contain the baddest outfits and prestige outfits. Therefore, images that contain the above outfits can be also generated using this model, theoretically.

Cautions (Must Read)

• Some preview images are generated using hires fix. Thus, if you wanna generate the preview images, pls download the original preview images and analyze them using 'PNG INFO' in webui to get all generating information.

• The maximum of generating resolution should not exceed 1024 theoretically, otherwise extra/deformed head/limbs will be waiting for you :). Try to use 'hires fix' to prevent this problem and help to generate higher-resolution images.

• For 784mb VAEs (NovelAI, Anythingv3, Orangemix, Counterfeit) users, make sure add '--no-half-vae' in command line of webui to prevent generating black images. See voldy's log for the details of edit method. For high-performance GPU, make sure 'medvram' and 'lowvram' mode are disabled.

Introduction

A Lycoris model trained with ~180 images by LAM. The trigger tag is ‘lam-style’. The 784mb VAEs (NovelAI, Anythingv3, Orangemix, Counterfeit) are recommended. 0.6 ~ 0.8 weights are good. The preview images are generated using 1.6 version (original) YesMix.

Cautions (Must Read)

• Some preview images are generated using hires fix. Thus, if you wanna generate the preview images, pls download the original preview images and analyze them using 'PNG INFO' in webui to get all generating information.

• The maximum of generating resolution should not exceed 1024 theoretically, otherwise extra/deformed head/limbs will be waiting for you :). Try to use 'hires fix' to prevent this problem and help to generate higher-resolution images.

• For 784mb VAEs (NovelAI, Anythingv3, Orangemix, Counterfeit) users, make sure add '--no-half-vae' in command line of webui to prevent generating black images. See voldy's log for the details of edit method. For high-performance GPU, make sure 'medvram' and 'lowvram' mode are disabled.

Intro

This is a LoHa / LyCORIS model of a cute Japanese girl face. Model was trained on various irl photos, however it does not depict any real person. Consider it a virtual character.

Trained on v1-5-pruned. Offset noise has been applied.

(Translation are all from chatGPT.)

これは可愛い日本人の女性の顔のLoHa / LyCORISモデルです。このモデルは、さまざまな実写の写真で訓練されましたが、実在の人物を描写しているわけではありません。仮想キャラクターと考えてください。

这是一个可爱的日本女孩脸部的LoHa / LyCORIS模型。该模型是基于各种实际照片进行训练的，但它并不描绘任何真实的人物。可以将其视为虚拟角色。

Settings

1 - Not necessary but highly recommended

To make this and any LoRA/LoHa more compatible, use this extension. I made some custom weight preset to seperate face from the rest of the contents. Copy and paste below lines into your LoRA block weight extension textbox:

このLoRA/LoHaと他のものをより互換性が高くするには、この拡張機能を使用してください。顔を他のコンテンツから分離するカスタムウェイトプリセットを作成しました。以下の行をコピーしてLoRAブロックウェイト拡張テキストボックスに貼り付けます。

为了使此模型及任何LoRA/LoHa更兼容，请使用此扩展。我制作了一些自定义权重预设，以将脸部与其他内容分开。将以下行复制并粘贴到您的LoRA块权重扩展文本框中：

----LyCORIS

-Keep only body-

LyCOBODY:1,1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,1,1,1,1,1,1,1,1

LyCOBODY0.5:1,1,1,1,1,1,1,1,1,0.2,0.2,0.5,1,1,0,0,0,0.2,0,0,0.8,1,1,1,1

-Keep only face-

LyCOFACE:1,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0

LyCOFACE0.5:1,0,0,0,0,0,0,0,0,0,0,0,0,0,0.2,0.5,0.8,1,1,1,0.2,0,0,0,0,0

-Keep only clothing (not tested)-

LyCOCLOTH:1,1,1,1,1,1,1,1,0,0.2,0.2,0.2,0,0,0,0,0.5,0.8,1,1,0.2,0,0,0,0,0

-Keep only pose (not tested)-

LyCOPOSE:1,0,0,0,0,0,0,0,0,0.2,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0

Like so. Remember to click Save Presets and Reload Tags. Then you use it like in the preview images, <lora:AG-yan:1:LyCOFACE>.

Reminder: As per the ReadMe of the extension, you cannot use the built-in presets (17 blocks) on a LyCORIS (26 blocks). Just keep in mind that you need to know clearly which it is that you apply onto.

If you don't want to use this extension, that's fine I guess, but things might got messed up pretty quick if you load more LoRA together.

このようにしてください。プリセットを保存してタグをリロードすることを忘れずに。そして、プレビュー画像のように使用します。 <lora:AG-yan:1:LyCOFACE>。

リマインダー：拡張機能のReadMeによると、ビルトインプリセット（17ブロック）をLyCORIS（26ブロック）で使用することはできません。適用するものがどちらかはっきりと把握しておく必要があることを念頭に置いてください。

この拡張機能を使用したくない場合は、それでも構いませんが、LoRAを一緒にロードするとすぐに混乱する可能性があります。

就像这样。记得点击保存预设并重新加载标签。然后您可以像在预览图像中一样使用它。

提醒：根据扩展的ReadMe，您不能在LyCORIS（26块）上使用内置预设（17块）。请牢记，在应用预设时需要清楚了解模型的类别。

如果您不想使用此扩展，我想这也没关系，但是如果同时加载更多LoRA，事情可能会很快变得混乱。

2 - Parameters

base res is 512*768

Sampler: DPM++ 2M Karras v2 (Basically the improved version of old 2M Karras)

Hires or img2img upscale is essential if you want full body portraits

upscale by 2 or higher.

Hires step: 36 or higher

Upscaler: R-ESRGAN 4x+ / Latent

Denoising strength: 0.5

Intro

This is a LoHa / LyCORIS model of a cute Asian girl face. Model was trained on various irl photos, however it does not depict any real person. Consider it a virtual character.

Trained on v1-5-pruned. Offset noise has been applied.

(Translation are all from chatGPT.)

これは、美しいアジアの女性の顔をモデルにしたLoHa / LyCORISモデルです。モデルはさまざまな実際の人物の写真で学習しましたが、実在の人物を描いているわけではありません。仮想キャラクターと考えてください。

这是一个可爱亚洲女孩面部的LoHa / LyCORIS模型。该模型通过训练各种真实人物照片而得，但并未描绘任何真实人物。因此请将其视为虚拟角色。

Settings

1 - Not necessary but highly recommended

To make this and any LoRA/LoHa more compatible, use this extension. I made some custom weight preset to seperate face from the rest of the contents. Copy and paste below lines into your LoRA block weight extension textbox:

このLoRA/LoHaと他のものをより互換性が高くするには、この拡張機能を使用してください。顔を他のコンテンツから分離するカスタムウェイトプリセットを作成しました。以下の行をコピーしてLoRAブロックウェイト拡張テキストボックスに貼り付けます。

为了使此模型及任何LoRA/LoHa更兼容，请使用此扩展。我制作了一些自定义权重预设，以将脸部与其他内容分开。将以下行复制并粘贴到您的LoRA块权重扩展文本框中：

----LyCORIS

-Keep only body-

LyCOBODY:1,1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,1,1,1,1,1,1,1,1

LyCOBODY0.5:1,1,1,1,1,1,1,1,1,0.2,0.2,0.5,1,1,0,0,0,0.2,0,0,0.8,1,1,1,1

-Keep only face-

LyCOFACE:1,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0

LyCOFACE0.5:1,0,0,0,0,0,0,0,0,0,0,0,0,0,0.2,0.5,0.8,1,1,1,0.2,0,0,0,0,0

-Keep only clothing (not tested)-

LyCOCLOTH:1,1,1,1,1,1,1,1,0,0.2,0.2,0.2,0,0,0,0,0.5,0.8,1,1,0.2,0,0,0,0,0

-Keep only pose (not tested)-

LyCOPOSE:1,0,0,0,0,0,0,0,0,0.2,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0

Like so. Remember to click Save Presets and Reload Tags. Then you use it like in the preview images, <lora:AG-yan:1:LyCOFACE>.

Reminder: As per the ReadMe of the extension, you cannot use the built-in presets (17 blocks) on a LyCORIS (26 blocks). Just keep in mind that you need to know clearly which it is that you apply onto.

If you don't want to use this extension, that's fine I guess, but things might got messed up pretty quick if you load more LoRA together.

このようにしてください。プリセットを保存してタグをリロードすることを忘れずに。そして、プレビュー画像のように使用します。 <lora:AG-yan:1:LyCOFACE>。

リマインダー：拡張機能のReadMeによると、ビルトインプリセット（17ブロック）をLyCORIS（26ブロック）で使用することはできません。適用するものがどちらかはっきりと把握しておく必要があることを念頭に置いてください。

この拡張機能を使用したくない場合は、それでも構いませんが、LoRAを一緒にロードするとすぐに混乱する可能性があります。

就像这样。记得点击保存预设并重新加载标签。然后您可以像在预览图像中一样使用它。

提醒：根据扩展的ReadMe，您不能在LyCORIS（26块）上使用内置预设（17块）。请牢记，在应用预设时需要清楚了解模型的类别。

如果您不想使用此扩展，我想这也没关系，但是如果同时加载更多LoRA，事情可能会很快变得混乱。

2 - Parameters

base res is 512*768

Sampler: DPM++ 2M Karras v2 (Basically the improved version of old 2M Karras)

Hires or img2img upscale is essential if you want full body portraits

upscale by 2 or higher.

Hires step: 36 or higher

Upscaler: R-ESRGAN 4x+ / Latent

Denoising strength: 0.5

Note: This is a Locon. Install the extension to get it to work. I've found them to do better when working on ideas that I want easily added to one of my LORA characters.

Another character changing Locon, this one specifically around taking your character and "bimbofying" them. It was trained on animefull_final with 63 images with the usual self-pruning and tagging. I like keeping my prompts simple for the first few pictures so that you can see it without a whole bunch of mumbo jumbo. Make your pictures better than mine though!

What does it do you ask? If you use this LORA without any other character LORA or descriptions, it'll create a someone with a thicc body, big breasts, and a relatively slutty outfit. I personally like combo'ing it with another LORA and at different variable to see different stages of transformation.

How to use it: The Locon strength best used is anywhere up to 1. Depends on how much effect you want it to have. Also the trigger word is:

BimboT - Only trigger word currently added.

Currently I have trained it to lean more towards bra/panty style outfits (though a usage of clothing trigger words should let it be swayed.) I've debated on making it lean slightly more towards skirts on a second version, so let me know what you think.

Extra, Extra!: Should be well trained and usable in most models. Images shown for example.

Picture Descriptions:

1-3: AnythingV3 using just this Locon.

4: AnythingV3 with my Shadow Sae Lora and this Locon.

5: AnythingV3 with my Sae Lora and this Locon.

6: AnythingV3 with my Hibana (Fire Force) Lora and this Locon.

7: AnythingV3 with my in progress Seraphina (Disgaea 5) Lora and this Locon.

8-9: AbyssOrange using just this Locon.

10: PerfectWorld using just this Locon.

What's next for this? Potentially a version that more balanced between the bra/panty combo and skirts when unprompted? Dunno if it's worth the effort though.

What's next that isn't this?

More character Loras. (Ghislaine from Mushoku Tensei, Conqueror Lilias from E7, Catherine from Catherine Full Body, Miku + Piscalat from Idaten Deities, Kuribayashi Shino from GATE, Rin Sohma from Fruits Basket, "Kasumi" from P5, Ann from P5, Ikumi Mito from Food Wars, Alice Nakiri from Food Wars, Chika Fujiwara from Kaguya-sama, Lulu from FF10, Momo from My Hero, Vivy from Vivy, Dominique de Sade from Vanitas, Dos from One Punch Man, and Seraphina from Disgaea 5.) And maybe Wilhelm van Astrea, my favorite Re:Zero husbando or a Prinny from Disgaea.

V2's of some of the previous Loras. Specifically make them more diversely acceptable with other models and LORA's.

I also have some more quite NSFW Locon's. Specifically, being pet play, being stepped on, brainwashing, corruptedv2, and one that might be too much for me to even upload here >.<.

And, uh, I want to make a few happy Locon's so I have an "angel-ify" your character one coming up. And a forest fairy one as well. And a soldier one. And a barbarian one.

Or just give me some random suggestion or priority opinions. That works too.

Tall dark matter choco cosmo caramel with dark choco sauce extra whip extra gravity

Tall dark matter choco plasma schwarzschild choco sauce with neutrino extra quasar

• CFG: 5.5 - 7

• Weight: 0.8 - 1.1

Features SN Chapayev from Azur Lane in her original costume & White Cavalier's Respite casual costume, trained on animefull-latest. See trigger words section for costume prompts.

Tested and works fine on high weight range up to 1.0, however finer details hadn't improved much from the earlier LoRA version.

Due to limited training data quality, it is highly recommended to apply with the following LoRA block weight:

1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0

to prevent style/image composition leakage.

A Character LoCon for Ijichi Nijika (伊地知 虹夏) | Bocchi the Rock! / 下北沢の大天使

[かき鳴らせ 交わるカルテット 革命を 成し遂げてみたいな]

• Recommeneded prompts: red eyes, side ponytail, sidelocks,school uniform, skirt,bangs

• Trigger word: red eyes, side ponytail

• Recommeneded Weight: 0.6-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5, DanMix

• How to use LoCon/LyCORIS? Please read instruction of a1111-sd-webui-locon.

• Some sample figures were generated with my Hitoimim Style - LoHa and Vivid Impactful Style - LoRA

If you like my work,please give me a like <3.

如果你喜欢我的模型请给我一个赞 <3.

もし、私のモデルが気に入ったら、「いいね！」をクリックしてください<3。

Another Loha trained to create women looking like Laura-Matsuda, it is not easy, but you can get some really near aproximations of this character.

This is a LoHa LoRA! You need the extension!
Why? LoHa have virtually no effect in output art style and they do a lot better with multi-concepts.
I honestly don't know if v1.20 or v2.20 is better, v2 have more tags for control tho.

Trigger words for V1.20:

<lora:CheekBulgev120:1>(cheek bulge, fellatio, penis, 1girl, 1boy)

You can try "deepthroat", "irrumatio", ":>=", "deepthroat", "cum in mouth", "saliva", "pov", "from above", "from side" (or no camera angle at all for some kind of front/side angle).
V2.20 also have on top of V1.20 tags: "head tilt", "finger to cheek:1.4", "penis grab", "lipstick on penis" and tagged for penis sizes but kinda just doesn't work.

I'm so happy with this that I couldn't stop generating with it!

Follow the steps in the Model Version notes to reproduce my images.

Sample images generated from Grapefruit.

All negative embeds should be easy to find. and Prompts should be included in the sample images! Click the little icon in the bottom right to see it.

You might want to add some negatives, for example:

It was trained on 326 512x512 images scraped from the internet and 20 epochs using animefull_final_pruned.

Danbooru tags generated by Waifu Diffusion 1.4 Tagger and fine tuned for specific tags.

Trained on anything v3.

Pictures are done with EasyNegative, kl-f8-anime2 VAE and various models.

Hikaru from (R)ape Gouhouka!!! (The name seems to avoid this LoCon to be published, so decided to tweak it a bit.)

• Trained with Anime (final-pruned) model.

• Comes in 4 versions; Anime-6 (Trained with 32 images), Extended-5 (Trained with 82 Images/5 epochs), Extended-6 (6 epochs), Manga style (6 epochs).

• Tip: You can make any combination of this LoCon's version to get interesting and high quality images; Put any "base" LoCon in the prompt with 0.5+ weights, and add another to Additional Networks with UNet weight of 0.2 and Text weight of 1.0 OR 0.2 weights in prompt with UNet weight of 0.5 and and Text weight of 0.7+. Also it works great with ALL, OUTD, and OUTALL LoRA weight blocks.

• Tip: Extended-5 and Extended-6 tend to get the character's arms up; You can avoid this by putting (arms up:1.21551) in the negative prompt.

A little Lora without prentention able to do some Juri from street fighter. Trained On Never Ending Dreams, (NED)

This LORA is Hakamada Hinata from RO-KYU-BU!

inital tags:

"nagatsuka saki, long_hair", "glasses, twin braids" or "goggles, forehead, hairband"

You can also mix and match.

main costume:

"gym uniform, bike shorts"

"basketball uniform, shorts"

You may want to use 7thLayerAbyss or AOM3A.

show you guts cool say what saikou daze.

If you'd like to support me and do more:
https://ko-fi.com/edgonkofi - commissions open!
If you're looking for a >>LoRA Making Tutorial<<

SDoom2.1

This LoRA was specifically designed with @fking 's model in mind.

Therefore I would advise you to please use fking_scifi_v2 | Stable Diffusion Checkpoint | Civitai

That should give you the best results!

This LORA is Hakamada Hinata from RO-KYU-BU!

inital tags:

"hakamada hinata, very long_hair, pink hair"

main costume:

"gym uniform, buruma"

"kimono"

You may want to use 7thLayerAbyss or AOM3A.

show you guts cool say what saikou daze.

This LORA is Tomoka Minato from RO-KYU-BU!

inital tags:

"minato tomoka, short hair, one side up, mole"

main costume:

"gym uniform, bike shorts"

"basketball uniform, shorts"

"school uniform"

You may want to use 7thLayerAbyss or AOM3A.

Thank you.

2023/4/6 update

Just update preview images.

2023/4/5 update

Upload Lycoris version (smaller file size). v1.0 version is conventional LoRA model.

Introduction

A conventional LoRA model trained with ~70 images by Hyulla. The trigger tag is ‘hyulla’. 0.6 ~ 0.8 weights are good. The preview images are generated using 1.6 version (original) YesMix.

Cautions (Must Read)

• Some preview images are generated using hires fix. Thus, if you wanna generate the preview images, pls download the original preview images and analyze them using 'PNG INFO' in webui to get all generating information.

• The maximum of generating resolution should not exceed 1024 theoretically, otherwise extra/deformed head/limbs will be waiting for you :). Try to use 'hires fix' to prevent this problem and help to generate higher-resolution images.

• For 784mb VAEs (NovelAI, Anythingv3, Orangemix, Counterfeit) users, make sure add '--no-half-vae' in command line of webui to prevent generating black images. See voldy's log for the details of edit method. For high-performance GPU, make sure 'medvram' and 'lowvram' mode are disabled.

A Character LoCon for Hiiro Amagi (天城 一彩) 【Ensemble Stars!】

Weight:0.4～0.7

Hires. fix

trigger words: amagi hiiro\(ensemble stars!\)

※以下日本語で失礼します。

運要素が絡みますがぽい衣装を出現させることは確認しています。

ALKALOIDライブ衣装（出現率高め）

推奨呪文: uniform\(alkaloid_hiiro\), white pants, black jacket, fingerless gloves, hat

私服:秋〜冬（出現率高め）

推奨呪文: white jacket, black pants, open jacket, blue vest,

トランプ衣装（運要素強。ライブ衣装混同しがちですがそれっぽいのは数回確認）

推奨呪文: playingcards uniform, hat\(playingcards\), black jacket, black pants

ネガティブ:uniform\(alkaloid_hiiro\), white pants,

お世辞にも使い勝手が良いと言えないですが使っていただけたら幸いです。

Attention!

download Extension into sd-webui, and then use built-in lora system to run your model. LoHa Model supported

Not For Kohya-ss' Additional Network

• checkpoint: Anything v5

• trigger words: cat tail, cat ears

• recommond CFG Scale = 7.0

• recommond weight: 0.7~0.8

• recommond use hires.fix

training parameters

$network_module = "lycoris.kohya"
$resolution = "712,1024"
$batch_size = 3
$max_train_epoches = 8
$save_every_n_epochs = 1
$network_dim = 128 
$network_alpha = 32
$clip_skip = 2 
$lr = "1e-4"
$unet_lr = "1e-4"
$text_encoder_lr = "1e-5"
$lr_scheduler = "cosine_with_restarts" 
$lr_warmup_steps = 0 
$lr_restart_cycles = 1
$noise_offset = 0.05 
$keep_tokens = 3
$min_bucket_reso = 256
$max_bucket_reso = 1088
$use_lion = 1
$algo = "loha"
$conv_dim = 8
$conv_alpha = 1

This is the Vietnamese traditional dress for woman called Ao Dai.

Recommend Sampler is UniPC, CFG from 5 - 9.

Use hires fix for more than 800x800 image.

Please refer my prompt

This is the Vietnamese traditional dress for man called Ao Dai.

Recommend Sampler is UniPC, CFG from 5 - 7.

Use hires fix for more than 800x800 image.

Please refer my prompt

Trained by LoCON

Recommend Sampler is UniPC, CFG from 5 - 7.

Use hires fix for more than 800x800 image.

Please refer my prompt

If you'd like to support me and do more:
https://ko-fi.com/edgonkofi - commissions open!
If you're looking for a >>LoRA Making Tutorial<<

LnF Aeros Lingerie

This is a spin-off of the LnF Aeros concept, but aimed towards Asian lingerie.

Have fun! Don't forget to like and review!

A simple Locon trained on "Never Ending Dream" (NED) to call Cammy-white, with a description of her.

Can have some trouble with the braids lenght or number. tested with strenght of 0.8

This is an EXTRACTION of 1.2 which isn't far off the OG one.

IT was done by Kappa_Neuro

Why don't you go TAKE A LOOK AT HIS AMAZING CONTENT?

THE MODEL THIS GOES WITH IS HERE: https://civitai.com/models/30363/cheesed-out-anime-backgrounds

KAPPA RECCOMENDS WITH LYCORIS EXTRACTIONS NOT TO USE A VAE.
THIS EXTRACTION LIKELY WORKS ON ANY MODEL AT WILL, AND WE FOR NOW HAVE YET TO TEST IT SO HAVE ONLY ACTUALLY USED AN IMAGE FROM 1.2 <3

HOW TO SUPPORT US:

We stream a lot of our testing on twitch: https://www.twitch.tv/duskfallcrew

Any chance you can spare a coffee or three? https://ko-fi.com/DUSKFALLcrew

Request image gens via our pixiv: https://www.pixiv.net/en/users/70748346

Hang with us on discord: https://discord.gg/Da7s8d3KJ7

If you'd like to support me and do more:
https://ko-fi.com/edgonkofi - commissions open!
If you're looking for a >>LoRA Making Tutorial<<

Oriental Lingerie

Very happy with my caftan and nuisette models, I use LoCon fusion design to bring you another KICK ASS lingerie model!

Download, like, and share!

This way to another great lingerie model!

A Character LoCon for Kita Ikuyo (喜多 郁代) | Bocchi the Rock!

Kita Kita~★

[足りない 足りない 誰にも気づかれない 殴り書きみたいな音 出せない状態で叫んだよ]

• Recommeneded prompts: Please check my sample pics!

• Trigger word: red hair, yellow eyes /green eyes

• Recommeneded Weight: 0.6-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5, DanMix

• How to use LoCon/LyCORIS? Please read instruction of a1111-sd-webui-locon.

• Some sample figures were generated with my Hitoimim Style - LoHa and Vivid Impactful Style - LoRA

• This Locon model should be the best model of Kita Ikuyo at this time.

If you like my work,please give me a like <3.

如果你喜欢我的模型请给我一个赞 <3.

もし、私のモデルが気に入ったら、「いいね！」をクリックしてください<3。

Ikuyo Ikuyo ~★

少女歌剧塔罗牌风格LoRA模型，4版本

this Revue Starlight CG Tarot Style Lora，has 4 version.

普通LoRA版以及卷积Locon版、loha以及IA3

Common LoRA and Locon LoRA、loha and IA3

这是他们的不同之处↓

comparing with them↓

what is LOCON、loha、and IA3？

readme

How to use？

Download Extension into sd-webui, and then use locon model as how you use lora model.

两种尺寸图可选，长图512X768以及宽图768X512

two size to choose,512x768(Portait) and 768x512(landscape)

背景最终色彩根据明暗程度自适应调整

The final result of the background will be adjusted automatically according to the brightness

推荐权重：0.4-08

Recommend Weight:0.4-08

你可以使用以下关键字控制效果

you can use these words to control effect...

1. arcana_style\(starlight\)

   →正位

   Upright

2. arcana_style\(starlight\)_upside-down

   →逆位，效果不好建议改成negative prompt

   Reverse,but it works not good,maybe use in negative prompt...

3. roman_numeral

   →显示罗马数字
   show roman numeral

4. English_text

   →显示卡牌名框，可改成负面
   show tarot name frame,maybe use in negative prompt...

尤菲米娅的人物loha,底模为anything3.0。

推荐权重<=1。只保留了人物的特征，具体衣服请参照我的例图使用tag还原。

24 Cans of Monster (METEION FFXIV)

Note: IF YOU HAVENT PLAYED ENDWALKER AND YOU PLAY FFXIV DO NOT USE THIS. SPOILERS MUCH SPOILERS. Also uh, if you just like pretty things do download cause she's fantastic!

WILL TAKE SOME PRETTY HEFTY PROMPTING TO GET World Ending Depresso Meteion either that or we're just lazy.

A meme to explain the meme:

We stream a lot of our testing on twitch: https://www.twitch.tv/duskfallcrew

any chance you can spare a coffee or three? https://ko-fi.com/DUSKFALLcrew

If you want custom LoRA OR MODEL trained an option will become available on the Patreon: https://www.patreon.com/earthndusk

This LoRA will end the world if you don't each her the proper ettiqutte.

AND NO METEION IS NOT A CHILD ffs lmao.

She's a birb.

「戦いには向かぬ衣装だと思うか？ 花嫁衣装こそ、女の戦闘着だ！ いつも以上にマスターを守ってみせよう」

• CFG Scale: 5.5 - 7

• Weight: 0.8 - 1.0

• Tigger word: nero claudius \(bride\) \(fate\), nero claudius \(fate\)

• Recommended prompts: ahoge, blonde hair, head wreath, bridal veil, lock, green eyes, chain, elbow gloves, zipper, belt, hair intakes, bangs, hair between eyes, loose belt, center opening, cleavage, white bodysuit, puffy detached sleeves, leotard, zipper pull tab, single hair bun, white footwear, hair ornament, sidelocks, petals, white leotard, collarbone, strapless

What is this

这是一个风格LoHA，可以画出清新明快的水彩插图。风格来源于Shato。训练使用Anything v4.5，测试使用meinaUnreal。部分大模型会使色彩过重，请多尝试不同模型。

This is a style LoHA that can create fresh and bright watercolor illustrations. The style is inspired by Shato. Anything v4.5 was used for training, and meinaUnreal was used for testing. Some large models may result in overly saturated colors, so please try different models.

これはLoHAスタイルで、清新で明るい水彩のイラストを描くことができます。スタイルの元になったのはShatoです。トレーニングにはAnything v4.5を使用し、テストにはmeinaUnrealを使用しました。一部の大型モデルは色彩が過剰になる可能性があるため、さまざまなモデルを試してみてください。

本模型素材由他人提供，包含共计80张图片，分辨率为1024*1024。我会在Train Data中附上完整的训练集以及训练参数（脚本使用：lora-script）。

The training images was provided by others and contains a total of 80 images with a resolution of 1024*1024. I will attach the complete dataset and training parameters (script used: lora-script) in the Train Data.

このモデルの素材は他の人から提供され、1024×1024の解像度を持つ合計80枚の画像が含まれています。私はTrain Dataに完全なトレーニングセットとトレーニングパラメーター（使用スクリプト：lora-script）を添付します。

❓How to use / 使用方法

提示词 / Prompts ：

🔸watercolor \(medium\), byShato, <lora:LoHA-styleShato:1>,

注意事项：当角色离镜头较远时可能产生面部崩坏，这是因为该画风的五官用的是简笔画的形式表现。

Note that when character is far away, face might get distorted, there's nothing I can do due to how face was drawn in this style.

注意事項：キャラクターがカメラから離れると、顔が崩れることがあります。これは、このスタイルで特徴的な顔の表現がアニメ調の簡略化された形式で描かれているためです。

📝其他信息与参数:

Model: meinaUnreal,

Image Resolution / 分辨率: any, tends to work better on high resolution

Sampler / 采样器: DPM++ 2M Karras v2 (改进版2M Karras 采样，CLICK TO SEE HOW)

Sampling Steps / 采样步数: >30

💜 I hope you enjoy it as much as I do. 希望你喜欢这个模型.💜

There are some signs of overtraining, please use only 0.6 lora weight or lower when using

Similarly, a model trained for only 1 epoch, which can easily change its posture or art style.

类似的方法，1epoch炼出来的模型，没有过拟合很容易换装换画风

Recommended weight:

• No character Lora: 0.7-0.9

• When using with Character Lora : 0.4-0.6

This style Locon specialize in thick lines and bold colors. It works really well at generating colorful anime with cel shading. It is quite strong, so it's recommended to lower the weight when using characters Lora if you want to keep all the character's feature.

Not trained on NSFW images, not tested for NSFW

Exemples made with the following models :

• Anything V4.5

• Anylora

• Store Bought Gyoza

Recommended Embeddings:

• Easynegative

• bad hands v5

This was made using HollowStrawberry's Lora collab. I highly recommend their guide, has it makes making lora a breeze!

I'm learning how to make a LoHA/artstyle model. This is my first step result, my PC isn't very good for testing so I release it for more feedbacks from your side, so this is basically an Alpha test, the model do work but not very user-friendly.

If you want to get a good result, you should know how to use lora block weight.

If you don't know how to use it but still have interest on learning some new tricks, then read the following guide.

My goal is make a artstyle LoHA that is capable to most of chara LoRAs.

So I trained it on Animefull-pruned, works perfectly on that one and Anything V3, will work on AnythingV4.5 without character LoRA, but almost won't work on Counterfeit or AOM3. It will more likely a color filter than an art style on those models.

Maybe I should train another version based on AnythingV4.5. If you have any suggestion about art style training, let me know. I will be very appreciate of that.

It can work without tags, but sometimes will make weird color.

Here are some tag groups for more stable results:

• Blue Archive-ish theme (Millennium?): tree, cloud, road, city, day, blue sky, outdoors

• Japanese theme (Hyakkiyako?): pink theme, cloud, day, outdoors, cherry blossoms,

• Nightsky: blue theme, night, night sky, starry sky, cityscape,

• You can try other tags generated by taggers of Blue Archive CGs.

You 'd better use it with LoRA block weight if you want to input a character lora. You can get it as webui plugin. Otherwise you should lower this model's weight to below 0.6, which will work bit poorly but still work, or it will make the image color very awful.

↓If you get something like this, lower the artstyle model weight, or use the method below.

When you use it with LoRA block weight, there are 2 methods:

• A: By default, you should remove your charactor lora's artstyle by this:

  <YourCharaModelName:1:1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0>

  don't forget to add :ALLto this model or it may won't be working.

  e.g. <BArstylev1c:1:ALL>

• B: If you noticed that your character's appearance has been obviously affected by this model, keep only the color of the artstyle model instead like this:

  <BArstylev1c:1:1,0,0,0,0,0,0,0,0,0,0,0.8,1,1,1,1,1>

  In this situation, you can keep Charactor model in full.

  e.g. <YourCharaModelName:1:ALL>

• Basically, method A will works more like the original artstyle than method B.

• It seems that LoHA model have more block weights, I don't have much time to test since my PC will easily crash on it, if you have some information about it, please tell me. But so far using LoRA block weight is fine.

• ↓Still, if you get everything right but still got something like this, try to use the second way instead.

Method A when the charactor is affected.

Method B

A Character LoCon for Korwa

• Recommeneded prompts: Please check my sample pics.

• Trigger word: asymmetrical clothes,asymmetrical legwear (for the original outfit)

• Recommeneded Weight: 0.6-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5, DanMix

• Please check out my profile for more information.

• How to use LoCon/LyCORIS? Please read instruction of a1111-sd-webui-locon.

If you like my work,please give me a like <3.

如果你喜欢我的模型请给我一个赞 <3.

もし、私のモデルが気に入ったら、「いいね！」をクリックしてください<3。

Added bugs to fix in a later version

如题。

As title shows.

Tried a new method which reduce the probability of overfitting，100+ HQ kokomi images dataset

尝试了一个新方法，降低了过拟合得概率，100张+心海图像数据集

This is a LoHa LoRA! You need the extension!
Why? LoHa have virtually no effect in output art style and they do a lot better with multi-concepts.

Trigger words for V1.17 (LoRA weight above 0.75 is a big too strong but still fine):

<lora:DoubleBunsv117:0.75>(double buns) or <lora:DoubleBunsv117:0.75>(buns cover)

You can try "double buns" for uncovered buns, "buns cover" for covered buns (these 2 are mutually exclusive, chose one!), "hair ribbon", "hair ornament".

Notes: It doesn't do very well with the buns color! Its almost always white since that's mostly what the dataset was! You might get different colors, you can try very basic ones, that's one of the things I might try to improve for next versions!

Follow the steps in the Model Version notes to reproduce my images.

Sample images generated from Grapefruit.

All negative embeds should be easy to find. and Prompts should be included in the sample images! Click the little icon in the bottom right to see it.

You might want to add some negatives, for example:

It was trained on 211 512x512 images scraped from the internet and 13 epochs using animefull_final.

Danbooru tags generated by Waifu Diffusion 1.4 Tagger and fine tuned for specific tags.

This is my next LoCon that I trained thanks to this guide:

[Guide] Make your own Loras, easy and free | Stable Diffusion Other | Civitai

Argie Film - Portrait (SD 1.5)

Inspired and based from Argentina Film Photography Aesthetic and Style.

There are LoRas and Checkpoints that reference the Analog / Film style like RZ - Analog, which I admire and took inspiration from it.

So I recompiled a lot of film pictures from my own to trained. It was intended for portraits more than for full body, for landscapes and others, I am not 100% sure that it works well.

It was trained by approximately 1200 portrait-type photos extracted from my personal film pictures that I took in the past with a film camera.

In the training I used this checkpoint with SD 1.5 as a base, so I also recommend that you use this checkpoint in conjunction with the LoCon to obtain the most faithful result to the one I had.

Also I experiment in its use with this checkpoint. (Use its own prompts + prompt I recommend below)

(Surely there will be better combinations, better options, but that I have not tested or documented.)

How to install LoCon:

• How to use models · civitai/civitai Wiki (github.com)

How to use it (There is no a trigger word):

Prompts: analog style, film, dust, grain, (film grain:1), photography of a + yourprompt, detailed face,

I also recommend using epiNoiseoffset_v1.

Sample: analog style, film, dust, grain, (film grain:1), photography of a 20 years old Golden blond hair Belarusian woman at the beach, detailed face,

Negative Prompts: rz-neg-foranalogportrait (download here)

You can experiment with negative prompts, but for me ‘rz-neg-foranalogportrait’ give very good results.

I recommend these parameters:

• LoCon Strength: 1.0

• Size: 512 x 512 - Truly I have not tried with other resolutions, I think the most stable is this measure.

• CFG: 9.5 is perfect (in my opinion)

• Sampling Steps: between 20 and 40 is the most stable. (40 more details)

• Sampling Method: DPM++ SDE Karras works very well for portraits.

I used this LoCon in img2img and its not gives very well results. (But you can try)

I invite you to experiment with checkpoints, mix them with other Loras and share them!

Don’t forget to put the prompts you used, thank you!

This is a LoHa model for SRMY（Cookie⭐）, you may need to INSTALL this extenstion before you use it.

To enhance the character style, you can add ( hakurei reimu:0.9) , red dress, armpits, (armpits hair:1.2)` to your prompt