Weight : 0.6～0.7?

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.

ACertainThing.ckpt is used as the model for the study.

Outfit in the Gaze_On_Me! Dance.

Trigger words: shirt, blue/green/black/pink shorts.

A LoHa Chainsawman Style.
All image upscale 2x from 512x768 using Tiled Diffusion. Most images from wildcards.

Model: hassakuHentaiModel_v11, Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, 'Overlap': 48, Clip skip: 2, "{'Method': 'MultiDiffusion', 'Upscaler': '4x-UltraSharp', Model hash: 973cb455d0, 'Scale factor': 2, 'Keep input size': True}", 'Tile batch size': 8, Denoising strength: 0.35, 'Latent tile width': 96, 'Latent tile height': 96, Tiled Diffusion upscaler: 4x-UltraSharp, Discard penultimate sigma: True, Tiled Diffusion scale factor: 2

LoHa of King Halo (umamusume). Please install extension on WebUI.

After installing extension, put LoHa file in your sd/models/lora folder and write LoRA notation in prompt to applying LoHa. <lora:your_loha_name:weight>

All uploaded images are generated by AOM3A1

Recommended options

• LoRA weight 0.7 ~ 0.8

• Trigger words

  • Base

    • king halo \(umamusume\)

    • (optional) ojou-sama pose

  • Racing competition costume :

    • ear covers, green bow, hair bow, purple gloves, off shoulder, bare shoulders, short sleeves, sleeveless, off-shoulder dress, collared dress, green dress, double-breasted, buttons, frills, garter straps, frilled thighhighs, purple thighhighs, white footwear, high heels

  • Cheerleader :

    • ear covers, official alternate costume, side ponytail, hairband, long sleeves, armband, white gloves, white jacket, buttons, double-breasted, pleated skirt, white dress, layered dress, blue skirt, long skirt, black socks, brown footwear

  • SSR Power :

    • official alternate costume, ear covers, red headwear, hat feather, bare shoulders, off shoulder, necklace, jewelry, wide sleeves, long sleeves, black gloves, lace gloves, red dress, off-shoulder dress, frilled dress, layered dress, lace trim, frills, fishnets, lace-trimmed legwear, black thighhighs

  • Casual :

    • casual, ear covers, green bow, hair bow, one side up, short sleeves, bracelet, vertical stripes, striped dress, black dress, see-through, high heels, black footwear

    • Monster(drink) like clothes can not reproduce well. :(

  • Tracen summer uniform :

    • tracen school uniform, summer uniform, serafuku, puffy short sleeves, purple bowtie, horseshoe ornament, sailor collar, sailor shirt, purple shirt, white skirt, pleated skirt, frilled skirt, miniskirt, zettai ryouiki, white thighhighs, brown footwear, loafers

  • Tracen winter uniform :

    • tracen school uniform, winter uniform, puffy long sleeves, sailor collar, sailor shirt, purple shirt, white bowtie, horseshoe ornament, purple skirt, frills, brown footwear, loafers

Use a model derived from or mixed with animefull model.

• Settings

  • Stable Diffusion WebUI by Automatic1111

  • DPM++ Series(SDE Karras, 2M Karras, etc.)

  • about 20 steps, CFG scale 3.5~6.5

  • I recommend kl-f8-anime2.vae

  • CLIP skip = 1

  • Use Hires.fix to get higher quality image

    • Upscaler

      • Latent, Latent (nearest-exact), Latent (bicubic antialiased) or other Latent series.

    • Denoising strength 0.50~0.65

LoRA training info

• trained on sd-scripts by kohya_ss.

  • LyCORIS by KohakuBlueleaf

  • Thank you all a lot!

• based on Animefull-pruned model

• 275 images consist of

  • 4x low quality 96 images

  • 8x medium quality 91 images

  • 16x high quality 88 images

  • => 1 epoch = 2520 images

• replaced character feature tags with king halo \(umamusume\)

  • 1girl, red eyes, brown eyes, purple eyes, bangs, pink eyes, long hair, blue eyes, blonde hair, horse tail, simple background, solo, orange eyes, brown hair, black hair, horse ears, parted bangs, yellow eyes, medium hair, horse girl, animal ears

  • I think these tags are the essentials of King Halo or

  • unnecessary for training because of duplication.

• resolution : 768x768 with aspect ratio bucketing.

• other training settings is included in training_data.zip or metadata of LoRA file.

And, special thanks to my best friend. DH.

It's a experimental work,which purpose is to add dots to the image.

This is a LoCon model that trained by 200 img.

weight 1.0 maybe generates better img,trigger is Cinerite.

Recommend prompts: watercolor \(medium\),

Hope you found it usefulミ(・・)ミ

Storm / Ororo Munroe.

No, i'm not hiding comic book stuff from people I just forget people want this stuff.

PUUHHLEEEZE:

JOIN THE DISCORD AND DEMAND THINGS OF US: https://discord.gg/Da7s8d3KJ7

Listen to the music that we've made that goes with our art:
https://open.spotify.com/playlist/00R8x00YktB4u541imdSSf?si=b60d209385a74b38

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

This is the SAME only a LYCORIS EDITION of the Art style Lora we've created. This means it infers the style A TON MORE without it losing color. Also funny noting that certain models of ours ACTUALLY have our art in them (Epic Mix, And a few others) - this is because Duskfall AI and Duskfall test were TRAINED models on HF (Huggingface) with our art in them -- both are indeed in Epic Mix and many merges with Epic Mix in it!

THIS PRODUCES A TON OF FLOWERS FOR NO REASON AT ALL (lies we know why) - PLEASE USE "FLOWER" and "FLORAL" IN YOUR NEGATIVE PROMPTS TO REMOVE ANY PANSIES FROM YOUR BRITISH GARDEN (sorry bad joke i'm gonna run off now and spray myself in the face)

JOIN THE DISCORD AND DEMAND THINGS OF US: https://discord.gg/Da7s8d3KJ7

Listen to the music that we've made that goes with our art:
https://open.spotify.com/playlist/00R8x00YktB4u541imdSSf?si=b60d209385a74b38

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

Anyways, this is based on our ART (See Visual Art 1.5 and Visual Slime Tutorial Checkpoint models) - this is NOT an extraction, we still dont' know how to do that - this was trained.

The Training data for the original LORA editions: https://huggingface.co/datasets/Duskfallcrew/DuskfallCrewArtStyle_Lora

ORIGINAL LORA SETS: https://civitai.com/models/14846/duskfallcrews-art-style-lora

We ask that you do NOT RESELL The training data but use the training data to better your AI for better or for B movie laughing status!

真正的木雕/木制艺术风格，也可以出人物，具体效果可以参考图片

The wood carving/wooden art style, chinese quintessence, can also be out of character, specific effects can refer to the picture

新人练手，不喜勿喷，谢谢

Rookie practice. Don't abuse if you don't like it. Thank you

recommend weight 0.8-1.2

you can support my work at patreon.com/goofyai

commission open

all my models are available at https://pixai.art/@goofy-ai/artworks/models

This model can do both her maid outfit and bunny outfit, just prompt accordingly.

lora weight of 0.7-1

highres fix recommended

drop your review. it helps a lot.

It's an another adafactor LoCon model, trained from 50 various sources images about hanafuda. The tests found that clothes&decoration&background performed better,and which tone is where the name comes from.

You can use the LoCon to design figures in the style of hanafuda

Add KoiKoi to improve better.

Weight 0.9-1.0 will perform better

We will comply with the "Cyber ​​​​Girl YouTuber Shiro", "Idol Club" and "Merry Milk" License Terms.

lowleg leotard ver.3.0

recommend prompt:lowleo ballet leotard

lycoris weight <lyco:lowleo v3.0:0.5:0.5:16:BODY>

lora block weight preset:0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0

LyCORIS of Eva Green

LyCORIS of Emily Blunt

LyCORIS of Diane Kruger

LyCORIS of Courteney Cox

This is a remake of the old dreambooth of her, I removed some of the pictures where she is older so there should be more consistency towards the younger age.

LyCORIS of Cara Delevingne

LyCORIS of Allison Mack

LyCORIS of Alexandra Daddario

因为之前有人提了一句所以就顺手做了，效果一般抛砖引玉。打电话这个姿势使用tag也可以直接生成，而此lora是让打电话这个特征更稳定,，同时如果你的tag合适它也可以生成一些姿势。我提供了某些极端场景下使用简例，你也可以自行发挥。从背后视角的素材较少所以效果不如侧面。

Because someone mentioned it before, I did it smoothly, and the effect is generally to attract jade.The gesture of making a phone call can also be generated directly using tags, and this lora is to make the feature of making a phone call more stable, and it can also generate some poses if your tag is suitable. I have provided some simple examples of use in some extreme scenarios, and you can also use it yourself. The material from the rear view is less so the effect is not as good as the side view.

It's a experimental work,which purpose is to challenge more concepts.

This is a LoCon model that trained by 200 Tea-Time img.

weight 1.0 maybe generates better img,trigger is dunch,and you can try a lot of different foods&drinks.

+++++++++++++++++++++++++++++++++++++++++++++++++

Add small animals to prompt,there are also surprisesᶘ ᵒᴥᵒᶅ

This is a LoCon of a favorite japanese artist of mine called Unasaka Ryou, he's known for mostly drawing girls in school uniforms (joshi kosei) and worked on a lot of H visual novel games, doujins, mangas and some animes too. His style is really gorgeous and adorable.

真正的青花瓷艺术风格，华夏国粹，也可以出人物，具体效果可以参考图片

The real blue and white porcelain art style, chinese quintessence, can also be out of character, specific effects can refer to the picture

新人练手，不喜勿喷，谢谢

Rookie practice. Don't abuse if you don't like it. Thank you

1girl, bailu \(Honkai Star Rail\), woolen skirt, swim, solo, smile, purple hair, plaits, open mouth, long hair, green eyes, dragon horns, detached sleeves, cute, clothes, braids, yellow tassels, blue and white clothes, black skirt, bare legs, blue background, ocean, masterpiece, best quality, water, detail,

尾巴的prompt是 dragon tail, purple tail 但是这个效果不太好，等我后期优化优化试试

还有prompt：elf ear, gourd 但我展示的图忘记用了，被白露可爱的傻掉了

权重0.75左右

训练依赖的大模型：mixProV3_v3

白露好可爱啊

v2: Optimized training data set, slightly improved goggle generation, improved usage with custom outfits, added a trigger word(lol

Mei Hatsume from the My Hero Academia series

Trained on YesMix 1.6.

Play around with the weights yourself, personally I liked 0.8-0.9 the most.

Use keywords like: pink hair, yellow eyes, symbol-shaped pupils for best results.

A very healthy model

Sample image is generated by AnyMiniPasky

• 欢迎加入AI绘画模型博物馆QQ群

• ①群：791222249

• ②群：599962550

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。

by a famous artist ，Adolpyh_Menzel ，I make it.However, please note that the actual effect is still far from the world's top celebrities. Works of art need appreciation, I see few works, good or bad is difficult to distinguish。

Different models have different roles. I recommend these：chilloutmix_NiPrunedFp32Fix.safetensors [fc2511737a], dalcefoPainting_3rd.ckpt [5efb10a9e1], ligneClaireAnime_v1.safetensors [672977e447], rpg_V4.safetensors [e04b020012]

通过有名的艺术大师menzel的50幅作品，我制作出来这个lycoris。但是请注意，实际效果仍与世界顶级名家相差甚远。艺术作品是需要鉴赏力的，我看的作品少，好坏难分。

不同的模型有不同的作用，我推荐的有：chilloutmix_NiPrunedFp32Fix.safetensors [fc2511737a], dalcefoPainting_3rd.ckpt [5efb10a9e1], ligneClaireAnime_v1.safetensors [672977e447], rpg_V4.safetensors [e04b020012]

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

• Multiple_girls to trigger, and you can add 3girls / 4girls / 5girls / 6+girls

Compositions:

• Group_picture: common group photo composition.

• Lineup: similar to group_picture, but tends to be arranged in horizontal lines.

• Harem: similar to group_picture, but with more erotic poses.

Although I try to classify the training images accordingly, the actual results may vary, so you may switch between them as needed.

Actions:

• Kneeling

• Lying

• Standing

• Sitting

Please note that as the number of people increases, the AI may tend to mix various actions.

This LyCORIS cannot perfectly generate images with multiple girls, but it can improve and increase the chances of generating such images. You still need some patience and luck, thank you.

One of the main heroines from the hit romantic drama VN game, White Album 2. Kazusa Touma! I didn't see a single lora model of her, and so I decided to change that. She's easily my favorite VN heroine.

(Beta) This LyCORIS is trained from many outputs from my mix Diffusion-Brush, no trigger words needed, simply add the lycoris into your prompt and it will do magic! Works well with Dreamshaper and Diffusion-Brush

( WARNING: USE :0.5 as a starting point 1 is too strong ) For example: <lyco:nucleardiff:0.5> . Nuclear Brushes are toxic so use them lightly.

Please take a second to consider...

Support my work: https://www.patreon.com/nucleardiffusion or a one off easy tip here: https://ko-fi.com/nucleardiffusion

Notice: This free work is intended for art and illustration purposes of emergent technology and not intended nor' trained for making illicit images, as the author I've taken care to not use any private data in the training process or designed it for such purposes. By downloading this model you are agreeing with the aforementioned intent and that you will not use my work to create / share images of a potentially immoral and / or illegal nature.

This is a LoCon of a japanese artist called Tetto (Onnoveltet). He's known for drawing OCs in jirai kei fashion in daily life situations in meticulous detail, one of them is called Hinamizawa Hinami. Back then he drew other kinds of characters such as an office lady, but now his focus is exclusively towards jirai kei girls (which I lately developed an extreme fetish for after gothic lolita fashion, don't ask...)

This is a LoCon I trained of one of my favorite japanese artists called Igarashi Kyouhei, he's known for mostly drawing his OCs but sometimes draws copyrighted characters as well, he rarely draws full nudity since he really hasn't done much hentai stuff at all AFAIK (of the light kind that is), but with this LoCon you can finally rectify that, even seeing your favorite character/waifu in his artstyle!

When I tried out a LoRA of an artist called Doyora (one of my favorite artists) I got from Catto's LoRA archive on GitGud, I noticed it wasn't that great with recent models such as AURORA (I highly reccomend you download and use it since it's the best anime model I've ever used in terms of versatility and prompt accuracy), surely because it could be lots of months old, so I decided to train one my own but as a LoCon after following a guide here made by Hollow/Holostrawberry.

Since this artist is known for his (kind of) oil painting style, depending on the model you use, you may not get the exact style of his (more or less) from the images of the dataset I trained it with, like for me with AURORA, I mostly get semi-realistic style by default on my end, let me know if using another model does give you his exact style or not.

Fuzichoco (藤ちょこ/藤原) style

極彩色の魔術師 | 极彩色的魔法师

An Artstyle LoHa

Intoduction:

Fuzichoco is an artist I greatly admire. Fuzichoco's artwork is renowned for its exquisite lines, vibrant and colorful tones, and intricate and elaborate character and scene designs. Although my model cannot fully replicate Fuzichoco's style, attempting to imitate it still produces some surprising results.

Fuzichoco 是我非常喜欢的一位画师。fuzichoco老师的作画以精美的线条，明艳绚丽的色彩和精致繁复的人物场景设计闻名。我的模型并不能百分百还原老师的风格，但是只是这些模仿也能达到一些让人意想不到的效果。

Update Note:

I am introducing a new v2a version model. In this update, I have used new training parameters and improved the quality of the tag. This updated model greatly improves the quality of the generated images, including but not limited to the clothing details of characters in the images, the way characters are colored, and the style and color of the background. Below, I will share some comparison images for your reference.

我推出了新的v2a版本模型。在此次更新中，我使用了新的训练参数并提高了标签的质量。这个更新后的模型极大地提高了生成图像的质量，包括但不限于图片中人物的服饰细节、人物的上色方式，以及背景的风格和色彩。下面我将分享一些对比图供大家参考。

Comparison（对比图）:

lize helesta, sebastian piyodore, 1girl, aqua hair, blue flower, blue thighhighs, boots, chair, colored inner hair, couch, cup, day, flower, grey skirt, hair flaps, hair ornament, holding, indoors, jacket, knee boots, long hair, long sleeves, looking at viewer, miniskirt, multicolored hair, on couch, plant, pleated skirt, potted plant, purple eyes, rubber duck, saucer, sitting, skirt, smile, sunlight, teacup, thigh strap, thighhighs, two-tone hair, watermark, white flower, white footwear, white hair, white jacket, window, zettai ryouiki, virtual youtuber, blue hair, solo, holding cup, bangs, monocle, duck, duckling, frilled skirt, frills, closed mouth, tea, shirt, blunt bangs, white shirt, holding saucer, animal, <lora:fuzi_loha_v1j:1>
Negative prompt: verybadimagenegative_v1.3, EasyNegative
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5.5, Seed: 1825595788, Size: 512x1024, Model hash: c870e984cd, Model: anythingv5.5, Clip skip: 2, ENSD: 31337

1girl, animal ears, bird, breasts, cherry blossoms, doorway, flower, fox ears, fox girl, fox tail, hair ornament, hand fan, hill, holding, holding fan, japanese clothes, koi, lake, landscape, painting (object), patterned clothing, pond, porch, red eyes, red tassel, smile, tail, tree, veil, water, waterfall, wedding dress, white hair, uchikake, kimono, solo, white kimono, looking at viewer, crane (animal), bride, hair flower, standing, bangs, long sleeves, closed mouth, animal ear fluff, hood, wataboushi, animal, sidelocks, wide sleeves, origami, bridal veil, branch, nail polish, from side, tassel, floral print, animal print, <lora:fuzi_loha_v1j:1>
Negative prompt: verybadimagenegative_v1.3, EasyNegative
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5.5, Seed: 2622988727, Size: 512x1024, Model hash: c870e984cd, Model: anythingv5.5, Clip skip: 2, ENSD: 31337

• Recommeneded prompts: japanese clothes,kimono,hair ornament,bow,white hair, animal ears (Please check my sample work for more details)

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

• Recommeneded Weight: 0.7-1 (You can adjust it by your perference)

• CFG scale: 5.5-7

• Check point recommendation: Anything V5

• 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。

It's an another adafactor LoCon model, trained from 50 various sources images,mainly blue theme flowers. The tests found that flowers&background performed better,and which tone is where the name comes from.

You can use the LoHa to design clothes & face,don't forget flowers

Add blue theme&Ultramarine to improve better.

Weight 0.9-1.0 will perform better

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

the regular creativeml-openrail-m license isn't applicable for this work

FIRST 4/25/23

Lemme take u in deep short backstory of my inspiration

Blue Moon🥰 from Cyberpunk 2077 the breathtaking character as game as well.

When had i first heard and learnt about US CRACKS..... oy boy(long sigh). I instantly wanted to meet them in the game (ngl i love every single quest that i completed). And eventually i have been taken a chance to meet them that was abrupt+awesome (as now remembering *incoming call Kerry*

Kerry: drop whatever u doin' meet me at the marina, don't make me wait )more spoilers ahead so u better play it by yourself)

Well...(these warm memories still chasing me) lemme tell u of model a little bit.

Fist thing first all the data that i could scrap and i did were game engine screenshots(intensively waving to the nsfw enjoyers) made by me+from net.

(do not frustrating now there's always is a solution see about that version and +nsfw capabilities also depends of models)

Second as per of data lack this training was a challenge one (as any other train in kohya-ss). I've spent literally a whole(~26h.) day only on trainings (in the name of science i suppose...yeah)

To start🤔:

(masterpiece, best quality, detailed:1.4), Blue Moon cyberpunk, 1girl, eyes, lips, spiked collar, blue hair, long hair, black jacket, yellow collar, solo focus, pink baseball cap, leather jacket, denim, looking at viewer, realistic, cyberpunk, head tilt, outdoors, grey eyes, upper body

(masterpiece, best quality, detailed:1.4), Blue Moon cyberpunk, 1girl, \\m/, bangs, simple background, blue hair, blunt bangs, bracelet, breasts, yellow choker, crop top, double bun, double v, double w, jewelry, lips, looking at viewer, see-through skirt, sport bra, small breasts, solo, standing, tank top, v, v over eye, w, wristband, grey eyes

Important note:

issue with

Kiroshi optics

(needs further investigation)

TL;DR SD sometimes can be a real bitch.

When u trying to make pupils of characters more natural and live then SD will always blur them out and contrary when u gonna make them double on each eye he's mostly just giving up. I recommend whether INPAINT(sketch) or Photoshop to make it properly).

LyCORIS of Alison Brie

LyCORIS of Billie Eilish

LyCORIS of Anna Chlumsky

LyCORIS of Bella Hadid

LyCORIS of David Tennant

This is a costume LoHa. Not character LoHa.

weight: 0.6-

What does it do you ask? Why have angelic waifu when you can have one that comes to you straight from the gates of hell? Evil and deadly, change your waifu into a demon and who isn't that nice, cute waifu you used to know. Use it either by itself or with any other character Lora you have. I left the base version with additional prompts to not change her overly dramatically but add my recommended tags to do the changes in ways you'd like. I didn't add violence or blood in my shown pictures, as it was preventing me from submitting pictures, but add it for yourself!

How to use it:

(I use Sampling Method DPM++ SDE Karras and Upscaler R-ESRGAN 4x+ Anime6B).

0.7 to 1 depending on model/lora/etc.

Main Triggers

DemonicT- Use this for a baseline change. Won't change everything about your waifu without the next part but some people might like that so left it like that.

Highly Recommended

blue skin, black sclera, colored sclera, colored skin, yellow eyes, demon wings, demon horns, slit pupils - To get that true evil waifu.

Other Fun Ones

Whatever you want! Preferably of the evil type.

And then whatever else you want!

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

Picture Descriptions:

1: RevAnimated using just this Lora.

2-3: AnythingV3 with just this Lora. One more demonified.

4-5: AbyssOrange using just this Lora. One more demonified.

6: AnythingV3 combo'd with released Dominatrix Locon.

7: AnythingV3 combo'd with my in progress Tied till Orgasm Locon.

8: AnythingV3 combo'd with my in progress Momo (My Hero) Lora.

9: AnythingV3 combo'd with my released Yunaka (Fire Emblem) Lora.

10: AnythingV3 combo'd with my released Sheik Lora.

How was it trained?

It was trained on animefull_final with 30 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's next for this?

I think it's fine? LMK.

What's next that isn't this?

More character Loras. (Rupee from Nikke, Spectre Tenebria from E7, Zephia from Fire Emblem Engage, Ivy from Soul Calibur 5, Heine from Combatants Will Be Dispatched, Conqueror Lilias from E7, Evelynn from LoL (both base and All Out version), Tharja from Fire Emblem, Laharl-Chan from Disgaea 5, Miku 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, Lucoa from Dragon Maid, Laharl-chan from Disgaea, 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. I decided to list them as well, so if people have preferences on which to fix up first let me know. Specifically, Bound Yuri (two girls tied up, breast to breast and likely kissing), Sultry Succubus Transformation, Tied up with vibrators (basically the waifu tied up with vibrators and being overwhelmed with pleasure), Beta Censorship (Automatically censors any images you create to make them BetaSafe), being stepped on, Robotification, Defeated (decently NSFW defeat scenario for the waifu getting abused), Petrified (I'm unsure if I should bother with this one anymore), Infection (this one is VERY nsfw, as the waifu is completely infected), Dominatrix (waifu becoming a dominatrix), Tentacles, Pillory, Masked Evil (A mask that possess' the waifu), Latex Corruption, Combat Drone (hard to describe but basically making them a mindless sci-fi combat warrior), JAV Supervillain, Mind Flayer Defeat, Blacked, Hypnotized Harem, Hourglass Figure (just to make your waifu curvy), and Holding a Leash.

And, uh, I want to make a few SFW Locon's so I have a Goth Chick, Delinquent (The delinquent girls in japanese schools specifically), GoodCop/BadCop (Locon with outfits for being a good and for being a bad cop), Maid (both normal and naughty variants), Biker Chick (ride a motorcycle and look like you should), Race Queen (Chiling sexily on a car), forest fairy, Anime Officer, barbarian.

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

I've also added a couple of my extreme NSFW preview images to my twitter (https://twitter.com/TwoMoreTimes89). Let me know if you think I should work on them more for a submission here.

Trained on 79 images drawn by SeeDark (Jeong Hoon Park), the art director and original artist of PangYa. He passed away from leukemia in 2009. He was 29 years old. This was made out of love and respect for his art. I can't read/speak/write Korean, sorry. Please use an online translator.

パンヤの元公式絵のアートディレクターとお絵かきさん「SeeDark」様の絵79枚を参考としてこれを作りました。シードアーク様は2009年で白血病からいなくなりました。彼はたった29歳でした。彼の作品にの愛と尊敬としてこれを作りました。

Training settings at the bottom of description.

トレーニング設定はこの説明の下に書いてありました。

Recommended weight（おすすめウエイト）:

Results vary, but 0.5-1.0 seem to be good. This mostly affects characters. I haven't tested with other models or male characters. May be difficult to use with other LoRAs, but you may be able to get around it by using detailed prompts, IMG2IMG, and/or ControlNet.

結果がいろいろ変わると思うけど、0.5-1.0は良さそうです。キャラぐらいだけ影響されます。他のモデルや男キャラには試したことないんです。他のLoRAと同時に使うのが難しいかもしれないが、詳細を説明するタグやIMG2IMGやコントロールネットを使ったら行けるかもです。

Models used（使用したモデル）：

• anyloraCheckpoint_bakedvaeFtmseFp16NOT

Since the moment I first saw LoRAs, I've always wanted to try making this. It's my first style LoRA. It took 3 days of trial and error. LoRA/LoCon failed many times: grotesque body parts, image changing too much, background becoming ugly, etc. I was about to give up and tried LoHa once. It was a success.

初めてLoRAを見た時、これを作ってみたかったんです。初めてのスタイルLoRAです。3日間ずっと実験してしました。LoRA/LoConなら何回失敗：グロい身体の部分、イメージ変わりすぎた、背景がゴミになった、とか。諦めるところでLoHaを一回試してみたら、成功でした。

I used images from Danbooru, Google, and also some I saved when I played PangYa 20 years ago. I also used IMG2IMG to resize and clean some images. I did not include images that were too different from his Pangya art style.

借りたイメージはDanbooru, グーグル、そして20年前パンヤをやった時に保存したイメージを使いました。IMG2IMGでイメージをリサイズと綺麗にもしました。パンヤの絵よりスタイル違いすぎるイメージを使いませんでした。

Training settings used（トレーニング設定）:

• 79 images

• 5 Repeats

• Batch 2

• Epoch 20

• bf16

• Seed: 31337

• Learning rate: 0.0005

• Cosine with restarts

• 5% warmup

• Text Encoder learning rate: 1e-4

• Unet learning rate: 5e-4

• Network Rank (Dimension): 8

• Network Alpha: 4

• Convolution Rank (Dimension): 4

• Convolution Alpha: 1

• Max res: 576, 576

• Min SNR gamma: 5

• All other settings should be default.

Pela (Honkai Star Rail)

All uploaded images are generated by YesMix

options

LoRA weight 0.7~0.8

Sampling steps 30 or Higher

Recommend Hires

Trigger words

• Default

  beret, pela, honkai star rail, black gloves, black headwear, black legwear, blue eyes, dress, glasses, gloves, hat, long hair, looking at viewer, pantyhose, red-framed_eyewear, semi-rimless eyewear,under-rim eyewear

斥罪这个角色不太好炼，那个盔甲基本只能在非常拟合或者说过拟合的情况下才比较能还原。所以这个Loha可能会过拟合比较多，建议权重0.8及以下。底模是NAI 7G版。

出的图全部是用DPM ADAPTIVE跑的，最近发现如果是直接出高清修复后的图的话dpm adaptive简直算得上是又快又好，自适应步数还是很省心的。

It's a experimental work,which purpose is to change the tone of the theme.

This is a LoCon model that trained by 60 img.

weight 1.0 maybe generates better img,trigger is noctiluca.

Recommend prompts such as scenery, XX theme, cityscape and so on,lol

• 768*768 or similar resolution is recommended.

weight: 0.6-0.8

真正的青铜艺术风格，青铜重器，也可以出人物，具体效果可以参考图片

The real bronze art style, bronze heavy objects, can also be out of character, specific effects can refer to the picture

新人练手，不喜勿喷，谢谢

Rookie practice. Don't abuse if you don't like it. Thank you

weight: 0.5-

option: jacket, head-mounted display, thigh strap, sandals, energy sword, dual wielding

「I'm your believer.」- Ciel

• CFG Scale: 5.5 - 7

• Weight: 0.8 - 1

• Trigger word not required

• Recommend prompts (WIP): See sample images for now

Other tags: long hair, white hair, standing, large breasts, hair over one eye,

Some of the sample images have other Lycoris added (check my profile),

The Scroll Of Taiwus a role-playing game known for its popularity as a Chinese indie game. The game is based on Wuxia and Chinese mythology.

The Scroll of Taiwu is a role-playing game set in the vast and immersive "Taiwu" Universe. Players take on the role of a "successor of Taiwu," where they must navigate a world filled with choices and consequences. Whether they choose to be virtuous or malicious, their actions will shape the outcome of their journey.

Meet them at ： https://www.conchship.com.cn/.

太吾绘卷，能找来的，懂得都懂。不解释了，我慢慢做。

A ufotable screencap, i recommend you to use the LoHA if you wanted better results

Recommended weight for the LoCon and the LoHA: 0.3~0.45 if you didn't fate artstyle to take over the character you want, keep it in your mind that most of Dataset was gathered from fate, i recommend 0.3 the most. It works good, for upscaling use ultrasharp-4x with denoising 0.3 & Tiling [Multidifussion] Method

0.5 and above if you want to do a chara in fate style

Mika Pikazo Style

An Artstyle LoCon

V3l LoCon is out. Please check it out!

Update Note:
For the V3L version, I have enhanced the tag quality of the training set and extended the training time, resulting in improved prompt obedience. However, I have noticed that the visual effects of this version are slightly inferior to the previous one. I am currently working on resolving this issue, and I have trained 11 versions before releasing V3L. I will release a new version as soon as I find a better solution to the problem.

In conclusion, I still recommend you download the new version since I think it is more important for a model to have a strong prompt obedience.

对于 V3L 版本，我增强了训练集标签质量并延长了训练时间，导致提示服从度有所提高。然而，我注意到这个版本的画面效果略逊于之前的版本。我正在努力解决这个问题，并在发布 V3L 之前进行了 11 个版本的训练。我将在找到更好的解决方案后尽快发布新版本。

总的来说，我仍然建议您下载新版本，因为我认为对于一个模型来说，具备强大的提示服从度更为重要。

Here are some comparisons between v2a and v3l (两个版本之间的对比) :

Please ignore the broken hand... （请无视损毁的手部。。。。）

parameters

1girl, sei shounagon (fate), visor cap, headphones, headphones around neck, shorts, swimsuit, bikini, breasts, yellow eyes, one eye closed, navel, twintails, multicolored hair, jacket, blue hair, thigh strap, chewing gum, red hair, green headwear, black hair, open clothes, bubble blowing, open jacket, looking at viewer, white bikini, short shorts, solo, long hair, black shorts, thighs, bikini top only, bare shoulders, paint splatter, off shoulder, medium breasts, large breasts, skateboard, white jacket, sidelocks, long sleeves, cleavage, blush, colorful, multicolored background, splashing, <lora:mika_locon_v2a:1>

Negative prompt: verybadimagenegative_v1.3, EasyNegative

Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5.5, Seed: 140136480, Size: 512x1024, Model hash: c870e984cd, Model: anythingv5.5, Clip skip: 2, ENSD: 31337, Hashes: {"vae": "f921fb3f29", "embed:EasyNegative": "66a7279a88", "embed:verybadimagenegative_v1.3": "d70463f870", "lora:mika_locon_v2a": "4f03e1877d", "model": "c870e984cd"}

parameters

kaguya luna, 1girl, :o, bangs, bare shoulders, black dress, blue background, blue eyes, blush, cleavage cutout, close-up, closed mouth, d-pad, d-pad hair ornament, dress, eyebrows visible through hair, finger to mouth, grey hair, hair ornament, hairclip, looking at viewer, colorful, multicolored background, multicolored eyes, multicolored hair, open mouth, paint, paint on body, paint on clothes, paint on face, paint splatter, pink eyes, pink hair, red background, short hair, short twintails, sleeveless, sleeveless dress, smile, twintails, upper body, v-shaped eyebrows, virtual youtuber, x hair ornament, yellow background, clothing cutout, solo, breasts, paint splatter on face, white hair, cleavage, character name, x, portrait, print shirt, print hair, <lora:mika_locon_v2a:1>

Negative prompt: verybadimagenegative_v1.3, EasyNegative

Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5.5, Seed: 2946328660, Size: 512x1024, Model hash: c870e984cd, Model: anythingv5.5, Clip skip: 2, ENSD: 31337, Hashes: {"vae": "f921fb3f29", "embed:EasyNegative": "66a7279a88", "embed:verybadimagenegative_v1.3": "d70463f870", "lora:mika_locon_v2a": "4f03e1877d", "model": "c870e984cd"}

• Recommeneded prompts: multicolored hair,long hair,looking at viewer,smile,teeth,jacket,grin, (Please check sample pictures for more details)

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

• Recommeneded Weight: 0.7-1 (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。

mimikaki concept

for better result you can add :pov from below, from below, looking down, ,mimikaki ,looking at viewer,lap pillow invitation,lap pillow,sitting,underboob

you can find me on :https://www.pixiv.net/en/users/34515535

or can support me on :https://www.buymeacoffee.com/bessu216z

my tw:https://twitter.com/phucbessu216

recommend weight:0.4~0.9

recommend model:Anything4.5,Anything5

characters promt:

shirosaki hana, long hair, hair flower, black hair, blue eyes, child

hoshino hinata, brown hair, red eyes, one side up, side ponytail, short hair, child

himesaka noa, :3, blonde hair, long hair, green eyes, hair bow, child

tanemura koyori, twintails, long hair, red hair, x hair ornament, purple eyes,child

konomori kanon, ahoge, brown eyes, short hair, brown hair, child

hoshino miyako, hair over one eye, brown hair, red eyes, short hair, large breasts

没什么特别说明，这是R18吗？不算吧大概

Nothing special, is this R18? Not probably

也许可以使用“hand on own stomach”让出图更稳定

Maybe you can use "hand on own stomach" to make the output more stable

偶尔会出两个肚脐，而且hiresfix非常容易把它修成pussy，真的

Occasionally two navels appear.Hiresfix is super easy to inpaint it to pussy, really

LoHa of Mejiro Palmer (umamusume). Please install extension on WebUI.

After installing extension, put LoHa file in your sd/models/lora folder and write LoRA notation in prompt to applying LoHa. <lora:loha_name:weight>

All uploaded images are generated by AOM3A1

[2023-04-26 UTC 08:00] fixed Racing tags.

Recommended options

• LoRA weight 0.7 ~ 0.8

• Trigger words

  • Base

    • mejiro palmer \(umamusume\)

  • Racing competition costume :

    • ear ornament, fur trim, fur-trimmed jacket, white jacket, yellow shirt, necklace, jewelry, open jacket, open clothes, long sleeves, white gloves, single glove, fingerless gloves, crop top, midriff, cropped jacket, navel, belt, green skirt, miniskirt, shoes, sneakers, asymmetrical footwear, mismatched footwear, black footwear, white footwear

  • Halloween :

    • halloween costume, official alternate costume, ear ornament, demon horns, jewelry, crop top, bare shoulders, sleeveless, cleavage, navel, midriff, single glove, elbow gloves, black gloves, asymmetrical gloves, fur trim, bracelet, nail polish, multicolored nails, fingerless gloves, demon wings, brown belt, black shorts, short shorts, thigh strap

  • Casual :

    • ear ornament, casual, green shirt, green bowtie, long sleeves, belt, shirt tucked in, grey pants, plaid, black footwear, high heels

  • Tracen summer uniform :

    • tracen school uniform, summer uniform, serafuku, puffy short sleeves, purple bowtie, horseshoe ornament, sailor collar, sailor shirt, purple shirt, white skirt, pleated skirt, frilled skirt, miniskirt, zettai ryouiki, white thighhighs, brown footwear, loafers

  • Tracen winter uniform :

    • tracen school uniform, winter uniform, puffy long sleeves, sailor collar, sailor shirt, purple shirt, white bowtie, horseshoe ornament, purple skirt, frills, brown footwear, loafers

Use a model derived from or mixed with animefull model.

• Settings

  • Stable Diffusion WebUI by Automatic1111

  • DPM++ Series(SDE Karras, 2M Karras, etc.)

  • about 20 steps, CFG scale 3.5~6.5

  • I recommend kl-f8-anime2.vae

  • CLIP skip = 1

  • Use Hires.fix to get higher quality image

    • Upscaler

      • Latent, Latent (nearest-exact), Latent (bicubic antialiased) or other Latent series.

    • Denoising strength 0.50~0.65

LoRA training info

• trained on sd-scripts by kohya_ss.

  • LyCORIS by KohakuBlueleaf

  • Thank you all a lot!

• based on Animefull-pruned model

• 127 images consist of

  • 4x low quality 54 images

  • 8x medium quality 48 images

  • 16x high quality 25 images

  • => 1 epoch = 1000 images

• replaced character feature tags with mejiro palmer \(umamusume\)

  • 1girl, aqua eyes, bangs, long hair, green eyes, horse tail, simple background, 1girl, wavy hair, grey eyes, streaked hair, white hair, blue eyes, hair between eyes, multicolored hair, brown hair, medium hair, two-tone hair, horse girl, horse ears, ponytail, solo, animal ears, very long hair, sidelocks, ahoge, parted bangs

  • I think these tags are the essentials of Mejiro Palmer or

  • unnecessary for training because of duplication.

• resolution : 768x768 with aspect ratio bucketing.

• other training settings is included in training_data.zip or metadata of LoRA file.

And, special thanks to my best friend. DH.

Trigger word

• mitsuha

• apron

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.

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。

「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!

Transform any character in a beautiful stone sculpture.

The first image was create using Anything v4.5 and the other 4 was created using AOM3.

How to get good images?

Use the next words:

• nude can improve a lot your results, and can prevent pieces of clothes over your "statue".

• white hair prevent statues with colorful hair.

• (twintails:1.4) if your character not has twin tails, is good idea put this as negative prompts. Try highest values if not work.

Other important things:

• Indoors background usually are better than outdoors backgrounds. Outdoors backgrounds usually has a lot of stone around.

• The poses generated usually are so similar, I recommend use open pose for variety.

• Horizontal images could not work so well. You can get good results with a horizontal images, but is not the usual.

This is my first published LoRA, if so many people like it, I will propose produce a better version that solve this problems.

LoHa of Ikuno Dictus (umamusume). Please install extension on WebUI.

After installing extension, put LoHa file in your sd/models/lora folder and write LoRA notation in prompt to applying LoHa. <lora:loha_name:weight>

All uploaded images are generated by AOM3A1

Image of Team Canopus is created with Tiled Diffusion and my LoRAs of team members. If you interested in, see my uploads.

Recommended options

• LoRA weight 0.7 ~ 0.8

• Trigger words

  • Base

    • ikuno dictus \(umamusume\)

    • Eyewear is not include!

  • Eyewear

    • round eyewear

  • Racing competition costume :

    • ear ornament, white shirt, collared shirt, red necktie, long sleeves, dress shirt, two-tone vest, green vest, white vest, armband, open clothes, green coat, belt, tailcoat, coattails, white pants, thighhighs, thigh boots, high heel boots, black footwear

  • SSR Guts :

    • long sleeves, turtleneck shirt, white shirt, ribbed shirt, suspenders, shirt stuck in pants, blue pants, jeans, denim

  • Casual :

    • She has no official casual clothes. I suggest some,

      • white shirt, long sleeves, green jacket, open clothes

      • long sleeves, black sweater, hands in pockets, blue shorts, black socks, sneakers, white footwear

  • Tracen summer uniform :

    • tracen school uniform, summer uniform, serafuku, puffy short sleeves, purple bowtie, horseshoe ornament, sailor collar, sailor shirt, purple shirt, white skirt, pleated skirt, frilled skirt, miniskirt, zettai ryouiki, white thighhighs, brown footwear, loafers

Use a model derived from or mixed with animefull model.

• Settings

  • Stable Diffusion WebUI by Automatic1111

  • DPM++ Series(SDE Karras, 2M Karras, etc.)

  • about 20 steps, CFG scale 3.5~6.5

  • I recommend kl-f8-anime2.vae

  • CLIP skip = 1

  • Use Hires.fix to get higher quality image

    • Upscaler

      • Latent, Latent (nearest-exact), Latent (bicubic antialiased) or other Latent series.

    • Denoising strength 0.50~0.65

LoRA training info

• trained on sd-scripts by kohya_ss.

  • LyCORIS by KohakuBlueleaf

  • Thank you all a lot!

• based on Animefull-pruned model

• 59 images consist of

  • 4x low quality 20 images

  • 8x medium quality 12 images

  • 16x high quality 27 images

  • => 1 epoch = 608 images

• replaced character feature tags with ikuno dictus \(umamusume\)

  • 1girl, streaked hair, red eyes, horse ears, hair between eyes, white hair, very long hair, horse girl, brown hair, long hair, green eyes, two-tone hair, simple background, bangs, purple eyes, multicolored hair, horse tail, short hair, brown eyes, grey eyes, yellow eyes, animal ears, orange hair, braided ponytail, single braid, solo, blue eyes

  • I think these tags are the essentials of Ikuno Dictus.

  • or unnecessary for training because of duplication.

• resolution : 768x768 with aspect ratio bucketing.

• other training settings is included in training_data.zip or metadata of LoRA file.

And, special thanks to my best friend. DH.

Sakigake Rei / 魁 令 / レイ from wixoss Series.
来自wixoss系列的令。

***LOCON file***

I also use the LoRA Block Weight, picture above shows the results of different parameters. If you don't want to install the plugin, you can just read the 'ALL' part.

我还使用了 LoRA Block Weight 插件，上图表示出不同参数下的结果。如果不打算安装这个插件，只看“ALL”的部分即可。

Weight can be 0.9-1.0, But I do like to use OUTD and OUTALL with the plugin.

权重大概在0.9-1.0这里，但我通常结合插件使用 OUTD 或者 OUTALL ，模式。

problems/已知问题

Too bright if you use other lora at the same time. Maybe prompt can help this.

与其他lora模型同时使用会变得非常亮眼。或许配合一些提示词可以改善。

If too close to the character, nipples will be obvious under cloth.

如果镜头离角色太近，乳头突起会很明显。

ラーメンを手づかみではなく箸で正しく食べるようになります。

相変わらず麺をこぼしがちですが・・・

If you'd like to support me and do more:

https://ko-fi.com/edgonkofi

White Mage Fashion

This one is dedicated to all the healers, all of those that keep the team alive and cope with overly aggressive tanks trying to aggro half the planet (yes, I'm one of them).

Thank you healers, enjoy this model which was made for you.

For all of you, I hope you will enjoy the white mage aesthetics of this LoRA. Shiny spells, magic staves, beautiful hats and so on!

very strong LoRA, stay under .3

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

这是一个，基本上可以把你想得到的场景都变成微缩的模型，经过测试可以很好的匹配大多数常见模型出图。出图非常简单，但是在这之前你需要安装一下插件，因为LyCORIS还没有被内置到webui.

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!!!

强烈建议先使用不含微缩的模型如chilloutmix测试，如果没有跑出微缩效果代表插件没有生效！

使用方法：

• 建议权重0.2-0.9之间。当权重在0.6-0.9之间可以跑出梦幻的效果，和微缩场景质感，但是会改变更多的真实建筑特征，如果你想要一个和原始建筑更像的模型，建议权重在0.2-0.5。请参考下面的XYZ图的巴黎圣母院了解。

  Recommended weight 0.2-0.9, when the weight is at 0.6-0.9 you will get more fantastic result and more feeling for miniature, but it will change a lot from the real building. if you are using a realistic SD model and prefer the looks like model, adjust the weight to 0.2-0.5 to get better "like" model.

  

• 使用mini\(ttp\)触发整体效果，使用miniature, lanpscape触发模型效果，如果这个模型本身有类似微距效果，那会大幅增强这个lora的特效，当然没有也不重要

• 使用容器的名字放在最后，比如plate, cup, bowl等可以把微缩城市装进去

• 使用isometric这种单词来触发方块的效果，其他角度你可以自己尝试

• 使用类似(statue of Liberty)的单词来触发主建筑物，使用(one man in the bathroom)这样的词来触发人物场景，没错这个微缩模型可以把小人放进去。满足你奇怪的xp。

• 其他没什么了，你们可以自由发挥了，希望你们能喜欢，一定要给我支持。

• 大场景需要高分辨率支持，显卡差的跑小场景有利于出图。

If you want to support me, please contact me through WeChat @tangtuanzhu and support me. This way, I'll have the funds for electricity and graphics cards to continue creating models. You can find my other models here: ttplanet Creator Profile | Civitai

This is a tool that can basically turn any scene you want into a miniature model. It has been tested and works well with most common models. Generating images is very simple, but you need to install a plugin first because LyCORIS hasn't been integrated into the webui yet.

Please note that since LyCORIS is not built into the webui, you must remember to install the plugin before using it.

The most updated lora block weight or additional networks plugins have bugs, causing the lora prompt to be ineffective. Please install with caution!

The most updated lora block weight or additional networks plugin are with bugs, it will make the lora command in prompt inactive, please wait for me to delete this line to update, otherwise you can't use Lycoris!

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

Then use the traditional method to call lora, for example: <lora:miniatureWorldStyle_v10:0.6>

Usage:

• Recommended weight is between 0.6 and 0.9.

• Use minimini\(ttp\) to trigger the overall effect, and use miniature, landscape to trigger the model effect. If the model itself has a macro-like effect, it will greatly enhance the lora's special effect, but it's not essential if it doesn't.

• Place the name of the container at the end, such as plate, cup, bowl, etc., to put the miniature city inside.

• Use words like "isometric" to trigger the block effect, and you can try other angles yourself.

• Use words like (statue of Liberty) to trigger the main building, and use phrases like (one man in the bathroom) to trigger character scenes. Yes, this miniature model can put small figures inside to satisfy your strange xp.

• There's not much else to say, so feel free to be creative! I hope you like it, and please support me.

「多くの言葉、僅かな煌き。どれほど遠く穢れても、私は星を探すのです」

• This is a Lycoris model, please install the Locon extension first to use

• Trigger words: artoria caster (fate),

• Use these to specific ascension stages (ignore for other costumes): artoria caster (first ascension) (fate), artoria caster (second ascension) (fate), artoria caster (third ascension) (fate),

• CFG Scale: 5.5 - 7

• Weight: 0.6 - 1.0

• Recommends prompts: blonde hair, long hair, green eyes, very long hair, hat, long sleeves, holding staff, twintails, black gloves, black pantyhose, black bow, belt, blue headwear, blue cape, bowtie, buttons, cape, puffy sleeves, pleated skirt

• Example images are generated on DanMix2.3

Redjuice Style

An Artstyle LoHa

You can use this LoHa generate Inori with Redjuice Style!

• Recommeneded prompts: 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

• 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。

If you'd like to support me and do more:

https://ko-fi.com/edgonkofi

Black Mage Clothing

With this model, enter the world of magic and fantasy. Give your character the magic robes it needs to stand in the magical universe you would like to generate.

This is a very powerful LoRA that will generate you beautiful outfits, but that you should keep between 0.2 and 0.6. Venture above at your own risks!

Black magic is after all a dangerous thing!

Get this LoRA to embark into high fantasy!

Duskfall's Own Vtuber Lycoris

Models used to test are in the image gens.

Come hang with us and yell at us for our insanity: https://discord.gg/Da7s8d3KJ7

"DEPICTS AN ACTUAL PERSON" is ticked only because it's an avatar of us, and it's sure not OUR FACE literally, but it's based on us. XD

Yea ok so we're not hololive, and we're unknown as hell, but yea, you've seen us around - we're duskfallcrew on twitch and duskfallmusic on youtube. This is a lycoris ENTIRELY based on our vroid avatar LOL.

No, it's not perfect, we're waiting on our peers to come back and go "Do we have to hold your hand to make sure you CAPTION IT CORRECTLY?" us going: NO CAUSE IT WAS DONE WHILE WE STREAMED ON A WHIM.

We blame Project AI for the dare.

"WE"? - We have Dissociative identity disorder, ADHD, Autism and CPTSD - "WE" as in we're a system of over 200 alters, and we're not ashamed about it. We believe that AI can break down barriers in some aspects of mental health, but we also believe that AI can hinder aspects of it.

PERSONAL USE IS A OK - BEANS WEBSITE BETTER NOT YOINK THIS because y'know we can't stand you NYEH!

MERGE MIXERS: Go ahead and merge this, we dare you.

Everyone else: We KNOW this won't get a ton of downloads but we did it on a whim as a joke.

Trained on the artworks of Pixiv user くろうめ, or Kuroume (link).

I really love their art style and wanted to make some similar stuff.

This is the first time I ever train a model and I don't have a lot of experience with Stable Diffusion in general so the results are a bit funky. It's somewhat working though, thus why I am posting the model here anyway.

If you have any insights on ways I could improve it please let me know. Right below are the details on how I trained it.

Following this guide, I first downloaded all of their art pieces in 250x250 (low res but otherwise downloading in bulk would've been harder.) I then upscaled them to 512x512 with R-ESRGAN 4x+ Anime6B. Lastly I labeled the images with DeepDanBooru which I had to curate to remove bad tags like "blurry" or various character tags (made a simple Python script for that).

Then for the actual training I just used the notebook they linked in the guide, making a few tweaks to the parameters: set the training model to this nice mix I found, set keep_tokens to 0 because I didn't use activation tags (to be honest I just don't really know what it is or what it does), set the steps to 5 repeats, 10 epochs (well actually 9), and 2 batch size, using 57 images resulted in 1430 steps minus a whole epoch since I'm using the 9th version, so 1290 total steps. Didn't touch the training section, and set the LoRA type to LoCon LyCORIS. And that's it.

To use it in Automatic1111's webui you first need to install this extension. Then just like any LoRA add <lora:kuroume-style:1> to your prompt (you can mess around with the 1 to set how much the model will influence the final image).

For the example images, I used NyanMix, found it worked pretty well. But honestly most models work pretty much the same.

Disclaimer

As I said previously, I'm very inexperienced. But I also did not put a lot of effort into this version, I was just trying to get it working. Think of it as a POC for my own ability at training LoRAs. I will definitely upload a new version where I'll go more thoroughly through the whole process to hopefully get better results.

(and btw if you're wondering the name of the LoRA is the artist's romanized name)

A Character LoCon for C.C.

• Recommeneded prompts: Please check my sample pics.

• Trigger word: c.c.,green hair,yellow eyes

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

• CFG scale: 5.5-7

• Check point recommendation: Anything V4.5, DanMix

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

• 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。

May Nasu and Takeuchi be blessed for creating this character.

My first lora/loha.

This model is for the character "Jeanne D'Arc Alter" from "Fate/Grand Order" and trained on >1500 images. Recommended weights are: 0.6 -0.8 . It should work well on most checkpoints.

General trigger word combination for the character: "jalter,1girl" .

Costumes require in general a higher weight >1.

Experiment with all parameters (e.g. sample steps, denoising steps, denoising strength, cfg scale, weights, etc.) for potentially better results!

Includes the following costumes:

1. Her first ascension outfit (use a combination of the tags: "jal_og_ver, headpiece, cape, short_hair")

2. Her final ascension outfit (use a combination of the tags: "jal_fin_asc, long_hair, very_long_hair, headpiece")

3. Her Shinjuku outfit (use a combination of the tags: "shinj_ver, short_hair")

4. Her Luluhawa swimsuit (use a combination of the tags: "lulu_swimsuit, red_tighhigh, black_jacket, long_hair, very_long_hair")

5. Her purple dress (use a combination of the tags: "jal_pur_dre, long_hair, very_long_hair")

6. Her heroic spirit festival swimsuit (use a combination of the tags: "alt_swi, long_hair, very_long_hair")

7. Nightingale's Trick-or-Treatment costume (Do NOT use the "1girl" tag here because of other characters bleeding into the image. Use a combination of the tags: "tri_o_tre, nurse cap, green tighhighs". Also add "syringe" to the negative prompt.)

8. Artoria's Royal Icing costume (use a combination of the tags: "roy_ici, long_hair, very_long_hair")

9. Bunny costumes (use the tag "jal_bu")

10. Maid costumes (use the tag "jal_ma")

Due to the incosistency or low amount of the available data, the costumes 7. - 10. are not that good (I only found like 7 images for Royal Icing and had to use other characters for Trick-or-Treatment).

Model of ANnie Lennox from the years 1985-1987

trigger word : melania

recommended strength : 0.85 -ish

thank you for visiting my post <3
you need this extension to run this file

then load like a normal lora file

if you like my stuff pls consider supporting me on Kofi

LyCORIS of Anna Mucha.

Weights around 0.8 - 1.4 depending on the model used.

LyCORIS version of Gillian Anderson.

Weights between 0.8 and 1.5 depending on the model.

LyCORIS of Aubrey Plaza.

Weight depends on the model used so do experiment. Usually, the sweet spot is between 0.8 and 1.4.

This is my first LyCORIS.

I believe that the LyCORIS quality is superior to that of a regular LORA.

In general, my technique has not changed. I still train a dreambooth model and then extract the result. Previously I've been extracting it as a LORA, now it will be a LyCORIS.

I still keep all the trained Dreambooths but I feel like it is no longer necessary to upload them as the LyCORIS quality is "good enough".

Samples were done on various models. You need to test the weight on each model as different checkpoint actually prefers different weights.

RealisticVision tends to work with 1.0 (probably because that was the base of the dreambooth training) but others need between 0.8 and 1.4 usually.

Realistic models will give us realistic outputs, stylized models will give us stylized results :)

Sticks a penis on a girl (not as if there isn't a ton of these already), All demo images use AnythingV3.0, from 2 sets of 8 gens.

Retrained with LoCon using Hollowstrawberry Colab

• Smaller size 28.9mb

• Have her Tonfa weapon

• Better generated images in my opinion.

I think it is decent enough

Trigger word is : makoto nanaya

Your review is greatly appreciated.

Recomended Model: https://civitai.com/models/5706/anyhentai

Negative prompt: (worst quality, low quality:1.4)

Hires steps: 20, Hires upscale: 1.8, Hires upscaler: R-ESRGAN 4x+ Anime6B, Denoising strength: 0.45, Discard penultimate sigma: True

The following settings are recommended
Checkpoint : anyhentai_19
Positive prompt: 1girl,dress,clownpiece
Negative prompt: EyasNegative,black background,simple background
weight: 0.8~1.0, CFG Scale: 4~8

If you want to use latest model, you must install this extention

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

This is Loha works in most models, and I think 0.5 or more is the best.

This model was trained by AOM3. Keyword is "Yoisaki Kanade"

Generate images of pokemon-like evolution

Welcome to my bilibili to see more pictures ~

【【SD模型】首个怪物进化模型！快去制作属于你的宝可梦吧！】

the base model is jimeido

example:

((monsterevolution)),(evolution:1.2),(giant),8k, high quality,highres,best quality, masterpiece,high detail, ((silavra art style)),3d render ultra detailed of a beautiful animal, ethereal, light rays, particles, beautiful studio soft light, rim light, vibrant detail painting, hell
Negative prompt: easynegative, humans, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, bad hands, normal quality, ((monochrome)), ((grayscale:1.2), breasts, nude, multiple tails

Steps: 30, Sampler: DDIM, CFG scale: 7, Seed: 650843989, Size: 1024x512, Model hash: 18fd1c8516, Clip skip: 2, ENSD: 31337, AddNet Enabled: True, AddNet Module 2: LoRA, AddNet Model 2: evilmonMonster_v10(402d84503538), AddNet Weight A 2: 0.7, AddNet Weight B 2: 0.7

• Weight: 0.9-1.1

• CFG Scale: 5.5-7

• Trigger words (important): fully body, chibi

• Negative prompt: card \(medium\)

• Some links to models:

  • Yuel

  • Morgan

  • Nia

  • Shallistera

• Have fun!

米夏·涅库罗/ ミーシャ・ネクロン/ Misha Necron

魔王学院的不适任者/ 魔王学院の不適合者/ The Misfit of Demon King Academy

建议权重/ おすすめ重さ/ Suggested weight: <0.6

建议开启Hires. fix以取得较好效果/ Hires.fixをオンにするとより効果的/ Recommended to turn on Hires. fix for better results

因为可用的图太少了，所以效果不是特别好/ 使えるイラストが少ないので、よい結果じゃないと思う /The effect is not good because few pictures can be used

This is LoCon grabbing a boob from behind.
<lora:nike-breast_grab_from_behind-v1:0.9>,breast,breast grab, 1boy,

Feel free to use it. If you like it and would like to support me, please purchase my works

https://www.dlsite.com/maniax/circle/profile/=/maker_id/RG73851.html


後ろからおっぱいを掴むLoConです。

<lora:nike-breast_grab_from_behind-v1:0.9>,breast,breast grab, 1boy、

ご自由にお使いください。気に入った方、応援してくださる方は、私の作品をご購入ください

https://www.dlsite.com/maniax/circle/profile/=/maker_id/RG73851.html

What is Top Artist Challenge？

i always heared “AIGC have not soul.”

so,i just want to try this style lora series .

i will use many ways to control best effect on AIGC.

you can find tac000 in here

For certain reasons, I have decided to hide the name of the artist and use a code name instead.

Who is 001？↓

i dont know,maybe ....

DIFF form LoRA Version：

advance version LoRA,don't need Improved quality prompt such as 'master piece,best quality',

it can show train data quality.

it can change to other model such as 3D real and works well.

BUT

LoHA cant load with Additional Networks

you must use

<lora:[TAC]tac001_style_LoHA_V1:0.8>

to trigger model

you can see more info in LyCORIS

Trigger words：

1. tac001_style

2. blurry

3. chromatic_aberration

4. depth_of_field

5. floating_hair

6. liquid

7. see-through

001's painting style has four 2 characteristics, the most obvious are the Dynamic and Lens feeling usage.

you can try to get Dynamic with these trigger words:

1. floating_hair

2. liquid

you can try to get Lens feeling with these trigger words:

1. chromatic_aberration

2. blurry

3. depth_of_field

I made some adjustments specifically for transparent scenes, and the next version will differentiate between close-up facial training and full-body training.

Extra

if you want to get more color use,you can try

1. vivid_color

2. neno_color

I did not specifically train for these two prompts, but I still achieved good results.

and use monochrome as negative prompt.

i upload train data caption txt,you can use it with wildcard for test random effects.

wildcard

just put it to extension/wildcard dir and use __tac001__ to try

LoHa of Haru Urara (umamusume). Please install extension on WebUI.

After installing extension, put LoHa file in your sd/models/lora folder and write LoRA notation in prompt to applying LoHa. <lora:loha_name:weight>

All uploaded images are generated by AOM3A1

Recommended options

• LoRA weight 0.6 ~ 0.7

• Trigger words

  • Base

    • haru urara \(umamusume\)

    • ponytail, ear covers, hair bow, long hair

  • 'Sakura Eye'

    • symbol in eye, symbol-shaped pupils

    • This eye needs enough space to draw. On too small or large space, you can not get Urara with 'Sakura Eye'.

    • I guess 100x100 ~ 200x200 for each eye is good. For example,

      • Portrait : 512x512 ~ 640x640

      • Cowboy shot : 1024x1536 with High.res 2x from 512x768.

  • Racing competition costume :

    • red headband, fingerless gloves, red gloves, gym uniform, white shirt, track jacket, open jacket, red buruma, bandaid on leg, striped socks, kneehighs, white footwear

  • New Year :

    • official alternate costume, hair bun, hair ornament, ear covers, wide sleeves, long sleeves, floral print, japanese clothes, red kimono, sash, obi, pink hakama, hakama skirt, frills

    • I think this costume reproduce not well.

  • Casual :

    • red headband, pink vest, puffy short sleeves, white sleeves, neck ribbon, red ribbon, collared shirt, overall shorts, blue overalls, sneakers, pink footwear

    • (optional) ice cream cone

  • Tracen stage uniform :

    • red headband, official alternate costume, choker, long sleeves, wristband, cropped jacket, white shirt, white jacket, red vest, midriff, overskirt, red shorts, short shorts, garter straps, purple thighhighs, white footwear

  • Tracen summer uniform :

    • red headband, tracen school uniform, summer uniform, serafuku, puffy short sleeves, purple bowtie, horseshoe ornament, sailor collar, sailor shirt, purple shirt, white skirt, pleated skirt, frilled skirt, miniskirt, zettai ryouiki, white thighhighs, brown footwear, loafers

Use a model derived from or mixed with animefull model.

• Settings

  • Stable Diffusion WebUI by Automatic1111

  • DPM++ Series(SDE Karras, 2M Karras, etc.)

  • about 20 steps, CFG scale 3.5~6.5

  • I recommend kl-f8-anime2.vae

  • CLIP skip = 1

  • Use Hires.fix to get higher quality image

    • Upscaler

      • Latent, Latent (nearest-exact), Latent (bicubic antialiased) or other Latent series.

    • Denoising strength 0.50~0.65

LoRA training info

• trained on sd-scripts by kohya_ss.

  • LyCORIS by KohakuBlueleaf

  • Thank you all a lot!

• based on Animefull-pruned model

• 115 images consist of

  • 4x low quality 34 images

  • 8x medium quality 27 images

  • 16x high quality 41 images

  • 16x close-up 13 images for reproducing 'Sakura Eye'

  • => 1 epoch = 1216 images

• replaced character feature tags with haru urara \(umamusume\)

  • 1girl, horse tail, pink eyes, horse girl, symbol-shaped pupils, pink hair, bangs, solo, purple eyes, simple background, hair intakes, flower-shaped pupils, animal ears, red eyes, horse ears

  • I think these tags are the essentials of Haru Urara.

  • or unnecessary for training because of duplication.

• resolution : 768x768 with aspect ratio bucketing.

• other training settings is included in training_data.zip or metadata of LoRA file.

And, special thanks to my best friend. DH.

THIS IS A LOCON.

May not be updated, after so many attempts.

Trained on NAI.

All images from AOM2nsfw. (I use MATURE FEMALE, otherwise this location/sample image may be removed soon, lol)

Weight = 1 work.

This LoHA is trained based on chilloutmix. If you want to join a community of AI drawing, and to learn more about LoRa training tricks, you can scan the QR code below to join our WeCaht group chat.

这个LoHA模型是基于chilloutmix训练的。 如果你想加入ai绘画团体，并且对LoRa训练小技巧感兴趣，你可以扫描下方的二维码加入我们的“C站炼丹术分享讨论群”，欢迎大家加入并且交流讨论！！！

捏Ta QQ频道同步欢迎你的加入（复制到qq频道内搜索）

t0o1vs01m0

Hey, i would appreciate if you leave a review <3

Release v4 Update! Better control. Trained with new experimental features(and more coming in later versions!).

I worked hard on it, and here it is! Overall, improved everywhere, except NSFW, dataset is still 100% SFW. Also no more latex-y clothes everywhere, that's been fixed. Hats shouldn't appear nearly as much, though, i didn't have that issue personally. General quality is quite a bit higher, since training was completely switched to Adan with optimized parameters special to ranks im using. Features like separate weight decay for unet and tenc were used, tenc weight clipping too.
v4 features expanded dataset, which led this lora to better generalization, and now it follow composition of original model better, which means you're not losing much while using it. Stability at weight 1 improved too.

Well, don't just believe my words, since they might just be lies for your model xD I have not released my own mix im training on yet, and i don't think it'll happen in near future, so, please, test yourself, and don't just believe my words. v4 should generalized styles better, so if it worked in v3, it should work better in v4.

Hey, it's my first time actually posting something that i trained, though, this is not my first attempt or anything, im doing that for quite a while, just finally deciding to share.

Im not most experienced person in this, and i can't experiment more than i already do, since my hardware is not allowing me to even do Dreambooth.
But you can be certain that i really worked hard on getting to what i train today.

LoRA you'll see and read about today is not my only project, i have multiple others, including NSFW, though, im not sure i'll ever release that. But it is likely that i will release more SFW in future, so i really would love to hear your feedback on using this one.

You can contact me about this LoRA through discord, Anzhc#5269, please, leave message, since im not gonna add someone to friends without knowing who they are.
If you'd like to support my endeavours and experiments with money - there will be a Patreon link at the bottom of text, since i will feel bad posting it at the top, just blatantly asking for money without you even reading about my stuff xD

If you're interested in detail about my training processes, you could join D8ahazard Dreambooth discord channel. im active, and share my findings and parameters there. At the bottom part of text i am providing general description of what i was doing.

Well, to actual description down there!

This is a personal training of style, or, averaged representation of an amalgamation of style concepts, if you will.
It's nothing special(wrong, now it's special, at least to me :D -- v4 update), since i can't do much on my GPU, but it works and i personally love it.
---Couple comparisons before we start wall of text---

P.S. v4 1.0 released, so, please, look at it's images for better representation of capabilities of last version, also - no more latex clothes everywhere.

---Text wall starts here---
S12 does not require trigger words, as it was trained as a fine-tune, but ohwx is there if you want to make it more powerful(mostly not needed).
Main features of that lora is abstraction and abstract power of subjects, which im very bad at describing, but i think you'll understand what i mean with example. Base models are usually already capable of understanding that concept, but my goal was to enchance it and make more tangible.

Style itself in base is anime-based, but with huge representation of 3D-like imagery.

Prompt for this: masterpiece, best quality, 1girl, upper body, powerful, abstract <lora:S12 ohwx Adan_18000:0.75>
negative:(worst quality, low quality:1.4)
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 3914799726, Size: 512x768, Denoising strength: 0.7, Clip skip: 2, ENSD: 4537, Hires upscale: 1.7, Hires upscaler: Latent (nearest-exact)

It is very simple to use. Preferrable weight of S12 LoRA is ~0.5-0.75, since 1 give too much of an effect, and will make image too dark for my liking, since it was trained with offset noise. In combo with other LoRAs it can be put down to 0.25 even. (And it does combine with some of them quite easily, without introducing visible errors, but i suspect it will not be the case, if other LoRAs are trained with offset noise too, since that'll increase effect)
Note: image below are made with the use of S12, but it is not a solo LoRA in generation and upscale.

(I don't remember prompt, and it was lost in upscaling one, sorry. I was deleting thousands of images cuz no space ;-;)
This image uses LoRA that i called "Effect"(i know im so f'ing original), it is meant to enchance composition a bit. S12 here is at 0.25

Dataset is fairly small, but enough for style purposes, 360 images as of v3.3.2(Will be expanded to ~580 for v4.0, and further for itterations). At the time of version 3.3.2 they were generated on Midjourney v4(one specific sub-style of their, but i don't remember which, probably 4b) with special style prompt, which is called very original, "style12".

I wanted to separate more nice knowledge, like "split theme" for example, since it's not very known by models i use, but dataset representation was not enough to overwrite that token knowledge, so it worse quite poorly and mostly with high weight, if ever.

---Capabilities---
Generally anything, since it's style LoRA. +Special concepts that i already described, + expasion of them in future.

NSFW?
Yes, but it was not trained on NSFW imagery. Dataset is 100% safe for work, so don't expect anything new in that regard and rely on your model knowledge. It will apply style to NSFW, and i quite like it sometimes.

---Example Images---
Since im using custom deep mix that i never posted on the internet, i will be including some images from one of base models of it - ACertainModel, which results you can apply to generally any other anime model, like Anything3. You'll be able to discern those generations by more 2d-ish look, or you won't be able, in that case consider me doing a good job and this LoRA is applicable to most anime models without much concern.

---Training quirks and techniques---



Lower info is valid for versions below v4 0.5

Dataset consists of 360 images generated in MJ, but with special style prompt that does not represent specific style. You can expect overall look to be quite MJ-ish, but... Not that MJ-ish, i guess.

It is trained with Adan D-Adaptation optimizer, which is decently new and experimental, so, expect quite a few rough spots.

Train type is fine-tune, so it will affect whole model for the most part.

Some images were tagged with custom quality tags, different from standard quality tags like "masterpiece" and so on, so they are not affected. They are not represented enough to matter in v3.3.2, so they will not be disclosed.

Techniques like offset noise were used, so it also improves contrast. Preferred weight of lora in that matter is 0.75, as it is too strong at 1.0. Noise offset was at probably ~0.12-0.15.
As it was trained with D-Adaptation, LR is not relevant much, as well as it's not available in main/dev D8 at the time of writing, so i don't know(rip).

Lora ranks i prefer 96/384, but for Adan training i had to lower it, since D-Adapt optimizers are heavy on VRAM. I think it's in range of 48-64/256.

Scheduler used was DEIS im pretty sure.

Since there were bugs that were making training be worse if you generated sample images in middle of training(should be fixed already), i would also specify that it is one-shot, and had no sample generation or saving checkpoints in process.

Note: It is trained on custom deep mix, which is a combo of ACM+AR in step 1, and around 1/3 of AOM2 in step 2, with selective high usage of certain blocks, which turned out to work awesome, and in lots of cases better for my usage. It will work with any other anime model, because of similarities, but best would be described above, specifically AOM2, since style is closer to it.

Images at that point are almost completely manually tagged(pain...), so, lora should be quite responsive to wide arrangement of them, not bleed from concept to concept, and be highly editable.

As i said at the start, here's link to my patreon - https://www.patreon.com/anzhc
It is not centered around AI, but i don't mind sharing my experiments there in WIP state, if that ever would interest patrons.
Just for genuine support, if you want.

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.😉

GitHub - 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。

GitHub - 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。

Recommended parameters:

推荐参数：

Model: MIX-Pro-V3

Vae: kl-f8-anime2

Clip skip: 2

weight:0.6

Sampler: DPM++ 2M Karras

Steps: 20

Highres fix: Enabled

Upscaler: SwinIR 4x

Upscale: 2

Highres steps: 15-25

Denoising strength: 0.3

Size: 512x512|448x640|640x448

CFG Scale: 7

Preferred Prompts:

推荐提示词：

Prompt:

1girl,solo,floating,otcg, <lora:Style_FantasyEffects_BlazeandBreeze (OTCG)_v1.0:0.6>

Negative Prompt:

EasyNegative, badhandv4

EasyNegative | Stable Diffusion TextualInversion | Civitai

badhandv4 - AnimeIllustDiffusion | Stable Diffusion TextualInversion | Civitai

Instructions:

说明：

This model supports most cel-shaded and flat-colored base models; it may produce weird results with thick-painted or 2.5D base models; it does not support real-life images.

支持大部分赛璐璐、平涂底模；用厚涂系、2.5d底模，有概率出奇怪的东西；不支持三次元。

If the face is distorted when drawing a full-body image, you can import img2img fix it.

画全身如果脸崩，可以导进图to图重绘一下。

The default color will be reddish-golden, if you want other colors, add color prompt.

默认颜色会偏红金，想要其他颜色加提示词。

Adding too many prompt words requires increasing the model weight accordingly, but too high weight will make you behold the eldritch horrors and lose your grip on reality. Like this↓Beware of the unspeakable things that lurk in the shadows.😱

加太多提示词需要相应提高模型权重，但权重过高会导致san值狂掉。如比这样↓

If you want an abstract background that looks like you had a trip on magic mushrooms, you may try using a higher weight.

如果你想要一个像吃了毒蘑菇一样的抽象背景也许可以尝试使用更高的权重。

Of course, you can also use it to create an atmosphere for Legendary creatures, like this↓

当然也可以用它做神兽的氛围图，比如这样↓

trigger word : h0mel@nder

recommended strength : 0.85 -ish

thank you for visiting my post, you people are the real heroes :wink:
you need this extension to run this file

then load like a normal lora file

if you like my stuff pls consider supporting me on Kofi

I recommend the weight to be 0.8

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<<

Valkyries

Golden wings, the sky enveloping, warrior maidens landing.

The fight is beginning.

Get your valkyrie!