If you create any good images using this LyCORIS, please consider sharing them by leaving a review. Thank you!

• This LyCORIS can help generate anime elves with revealing clothing.

• Elf to trigger, and if you add Dark_skin, it can generate anime dark elves.

• Elves have Breast_curtains, Bikini, Pelvic_curtain as their clothing, along with Elbow_gloves, Thighhighs, Jewelry.

• The training data is mainly softcore NSFW images, and the basic clothing of elves is very revealing.

• Although Breast curtains are not the primary training target, adding Underboob can increase the success rate.

My first LoRA Model.

I used Kohya Dreambooth Colab to train this model using 40 images.

Since there were very few images, this model was trained using both anime screencap, fanart and AI generated images.

Works well around 0.75 ~ 0.9

This is a loha model. Click here to get more information.

In practice, however, you will find that some characteristics do not match. One of the reasons is the number of training sets are not enough.

You may need to enter some additional characteristics to separate. And I have recommended some in below.

It has 4 concepts:

• second ascension

  Type second ascension to trigger

  And to avoid concepts fusion, you may additionally need to enter tags such as pelvic curtain, horns, pantyhose, etc.

• ice warrior

  Type ice warrior to trigger

  And to avoid concepts fusion, you may additionally need to enter tags such as white scarf, ponytail, black bikini, etc.

• like a lady

  Type like a lady to trigger

  And to avoid concepts fusion, you may additionally need to enter tags such as red sweater, off shoulder, etc.

  The right eye of the character in this costume is different from the other costumes, so you can type alter eye to distinguish it but it doesn't work well.

• dream portrait

  Type dream portrait to trigger

  And to avoid concepts fusion, you may additionally need to enter tags such as white dress, belt, etc.

First ascension and third ascension are not included because I didn't find enough images to train

LoHa model for SZ, the trigger word is char_char_sz_loha

Hi, this is an experimental LoHA that I trained for fun. It generate what behancer called "Daily Render" images.

Or perhaps we should call this Secondly Render? LOL, just kidding.

See images for prompts and checkpoint I used.

Side note: I've been trying to nail down how to train a proper LoRA that can generate product design(for concept sketches and/or beauty shots), but so far no luck. Soo I might need some pro advices on which checkpoint should be used in such purpose, and if someone have done similar things, a how-to guide would be much appreciated.

It is the first time to make Lora, and I set the parameters by the way. I hope someone can provide a better configuration file. 500 datasets.

LoHa model for askICG

Sample images were generated with https://github.com/hnmr293/sd-webui-cutoff extension and target words is `green, pink`

Recommeded tags:

` green dress, strawberry hair ornament, pink capelet, pink neckerchief, pink sash`

YMN sik LoHa

Recommended tags:

` char_ymn_sik_loha, purple skirt, red ascot, white shirt, `

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

The locon is extracted from the full model

The associate hugging face repository contains original models https://huggingface.co/alea31415/watayuri (may update one version later)

I am at some airport at the moment that I upload this model. Therefore I will only be able to add more example images after the flight in like 24 hours.

Trigger words:

• ChibanaSumika

• ChibanaSumika, glasses

• KoshibaMai

• MamiyaKanoko, glasses

• MamiyaKanoko, hair bun

• NishideraNene

• ShirakiHime

• ShirakiHime child

• YanoMitsuki

• YanoMitsuki child

I have not tested them thoroughly and I doubt we can get good result for Nene now as there are too few images of her. Same for Sumika without glasses.

The model is only trained with screenshots for the moment, so all the images are tagged with aniscreen. It is up to you to decide whether you want to add it or not. Characters are separated by ;.

You can get the clothes correctly with general prompts like green school uniform or brown coat.

--

I don't need support or credit, but I would be glad to know that you are using the models I trained and find it useful.

Moreover, I would like to advocate for more franchise models.

You can take a look at my workflow https://github.com/cyber-meow/anime_screenshot_pipeline if you are interested.

I just want to spread the fact that there is no reason to encode a single concept in each lora.

A Character LoCon for Yu Mei-ren (虞美人) [Lancer]

• Recommeneded prompts: Please check my sample works!

• Trigger word: 1girl, bikini, brown eyes, brown hair,, hair ornament, nail polish, solo, sunglasses, sunglasses on head, swimsuit, robe,twintails, white bikini, very long hair, multiple earrings, medium breasts, bare shoulders, open clothes,cleavage, bangs, jewelry

• Recommeneded Weight: 0.8-1 (You can adjust it by your perference.)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5

• 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。

This is a loha, please place it in the models\LyCORIS folder, Download LyCORIS extension.

这是一个loha，请将它放在models\LyCORIS文件夹，并且下载LyCORIS扩展。

This loha is trained using Anything V3.5.

这个loha 使用Anything V3.5进行训练。

The presentation diagram was generated using anything v4.5

展示图使用anything v4.5进行生成

You don't want to generate wings, delete the tag 'fair aloe'.

你不想要生成出翅膀，删除fairy aloe 这个tag。

We will comply with the ANYCOLOR secondary creation guidelines.
https://www.anycolor.co.jp/guidelines/

It seems that no particular intention is indicated. I pledge to create and publish works that respect the spirit of ANYCOLOR content.

特に個別の意向が示されていないと思われます。ANYCOLORコンテンツの精神を尊重した作品を制作、公表していく事を誓います。

弦宝的loha底模，使用NAI 7G底模制作，经过了16000步终于把loss降到了<0.1....

推荐权重<=0.9

百炼嘉维尔的loha模型，使用NAI 7G版本底模，推荐权重<=0.9

凯尔希的loha，用NAI 7g版制作，追求更好的泛用性。

推荐权重<=1，当权重为1时一般不会画面崩坏，但是对风格会有很大影响。如果想尽量保持底模的风格，那么应该把权重调低，调低到什么数值最合适需要结合具体模型测试。0.7-0.8可能是不错的范围。

为了方便修改各种特征，本loha并没有对任何特征tag进行删去tag的操作，所以使用的时候请根据自己的需求复制版本说明中的特征词汇，这也是我做出的一种新尝试。

Loha by kal‘tsit was made using NAI 7g version, aiming for better versatility.

Recommended weight <=1. When the weight is 1, the image is generally not prone to breakage, but it has a significant impact on the style. If you want to preserve the style of the base model as much as possible, it is recommended to lower the weight. The most suitable value depends on specific model testing. 0.7-0.8 may be a good range.

For the convenience of modifying various features, this Loha has not removed any feature tags. Therefore, when using it, please copy the feature tags from the version description according to your needs. This is also a new attempt that I have made.

A LoHA of Mandi Collins

Model trained with help of the great tutorial https://civitai.com/models/45539/dreambooth-lycoris-lora-guide created by @malcolmrey .

Cropping of the training images with https://civitai.com/models/46364/web-bulk-image-cropper

This LoRA is not official or authorized.

I check Twitter every day, so I will respond when I receive a DM.
ツイッターは毎日チェックしているのでDMが来たら対応致します。

kiana kaslana (herrscher of finality)

There are two versions, base layer weight 0 and normal.

In my own testing, it seems that when using the normal version, the output is often closed eyes.

Trigger words:

kiana_herrscher_of_finality -> base trigger word

finality_dress -> dress

finality_dress_breasts_out -> just breasts out

finality_wo_dress -> naked but wearing shoes and socks

finality_nude ->nude

+_+ -> eye detail

例图使用t2i -> i2i -> Tiled Diffusion+Tiled VAE生成，在最后一步中删除了除画质之外的prompt

サンプル図はt2i -> i2i -> Tiled Diffusion+Tiled VAEの順番で生成、Tiled Diffusion+Tiled VAEの時一部のpromptsを消した

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.

Its a model to make cool swords, its pretty unwieldy though, you can try to get the type of sword you want, it kinda works, specially with high prompt weighting or low LoCon weight, but it is random.

Images were generated using Aurora. LoCon was trained on NAI.

download my sample and import to webui for prompt!!!

download my sample and import to webui for prompt!!!

download my sample and import to webui for prompt!!!

下载图片导入webui获得提示词！！！

下载图片导入webui获得提示词！！！

下载图片导入webui获得提示词！！！

V2.0版本发布了，这是我的迭代版本，也是一个基本自动的版本，如果你只是希望尝试简单的效果，请选择V2.0版本，他稳定，效果直接，但是会限制一些些你的创意！

如果你觉得自己可以控制这个模型，我依然坚持使用1.0版本，它有更高的自由度，但是你需要对人脸，画面进行更多的修正来获得更好的效果。

I will publish the V2.0 for easy use, it will be a auto version to help most people get the effect. however it could limit your creation!

for ultra advanced user, I will still prefer V1.0 although it need much more care about the face and other crashed things!

please ensure you read the instructions before you use it！all the pics are with generation data, just download it and see it in webui!

important

Basic setting Lora weight 0.65-0.8 and Denoising 0.2-0.3！！！skip clip 1 to get front, clip 2 to get 45 degree!!!

基础设置 lora权重0.65-0.8，重绘率0.2-0.3，最佳区间！skip clip 1正面，clip 2侧面!!!

前言:

如果你要支持我，请通过tangtuanzhu的微信联系我，给我支持...，这样我才有电费和显卡继续跑图。我的其他模型都在这里：ttplanet Creator Profile | Civitai

请下载我的范例图片并从webui中获得提示词，不知道是不是bug不能直接读取！！！

if you see missing generated data, please download the pic and get it back from webui! this is the new bug of Civitai！

Hires fix is highly recommended!

推荐一定要高清修复！

介绍 Introduction:

这个模型非常的好玩，可以把你的人物塞入一个包装盒里面，如果你组合使用其他模型就会出现玩具的效果。塞入包装盒的除了人物，还有和他相关风格的装饰，衣服等，你可以自己输入关键词来添加，虽然不能精准到每一个物品，但是也基本上八九不离十了。

目前没有加入动作支持，请自行使用controlnet控制动作，将在V2版本中迭代大量的动作，让使用起来更加简单随机，但是依然推荐使用controlnet进行控制以获得最好的效果。

如何使用 How to use:

注意事项，由于webui没有内置LyCORIS使用前一定要记得安装插件,

https://github.com/KohakuBlueleaf/a1111-sd-webui-locon

然后使用传统的方式调用lora, 比如： <lora:miniatureWorldStyle_v10:0.6>

请注意使用我的tag,你记得修改lora的名字，我的tag是本地的名字，和你下载的不一样！！

Reminder: if you want to use my tag, please check the lora name, see if it is consistent to what you downloaded!!! my local file name does not match the online version!!!

根据不同的模型，建议使用权重为0.6-0.8，但是不要使用1,基本上会爆炸。

使用结构为：

inbox\(ttp\)作为触发词，<lora:loraname:0.7>作为调用词，中间放入你放入的人物行了。使用很简单，人物描述尽量使用自然语言。

use 0.6-0.8 weight based on different SD models, but don't use 1

inbox\(ttp\) to trigger，<lora:loraname:0.7> to use the lycoris, and put down the person you want to put into the box, please use nature language to describe the person.

范例 example:

inbox\(ttp\), (8k, RAW photo, best quality, masterpiece:1.2), (realistic, photo-realistic:1.37),1girl, huge breasts, (holding a guitar:1.5), (solo), blonde hair, instrument, guitar, boots, pants, male focus, reference inset, standing, electric guitar, red pants, shirt, full body, microphone, long sleeves, amplifier, looking at viewer, white footwear, holding, realistic, shoes, blue eyes, black shirt, jewelry, fashion, short hair, english text, print shirt, glasses, holding instrument, full body, simple background,

<lora:inboxgirl_V1_loha-pickup1:0.65>

权重0.6-0.8是变成盒子的最好区间，低于0.6不会进入盒子但是会提高模型的身材，高于0.8会由于lyCORIS特点出现独特的画面劣化。请注意调整。日常使用请用0.65,这是我调节过最平衡的参数

please use 0.6-0.8 weight to get the best result, any weight lower than 0.6 won't trigger the box, and above 0.8 will destory the pic. for regular use 0.65 will be the best!!!

模型本身倾向使用平坦的胸部，请使用巨大来进行平衡。

the model is with high weight to be flat breast, so use huge to balance it

如何进阶使用 How to use(advanced):

1. 使用controlnet来控制你的人物动作，记得调整骨架的大概大小和位置。please use controlnet

2. controlnet也可配合大头骨架生成Q版造型，请查看我的范例图获得灵感。use controlnet can build the chibi outfit

3. 可以和很多lora进行融合，使用的时候请注意控制权重。本模型权重较大，请适当增加你配合的lora权重, can be used with other lora together

4. face editor plugin 强烈推荐使用，脸部会因为lora变成你不喜欢的样子，可以搭配不同的lora使用face editor提高效果。就算不用lora也会更好看，使用face editor的时候请移除本lora再使用，保留其他prompt