IF YOU ARE NEW TO STABLE DIFFUSION OR YOU HAVE PROBLEMS WITH MY MODELS, GO HERE: Guide

How to use this

If you are new to the world of Stable Diffusion, you can learn how to run it by yourself here:

Tutorial

You'll also need a VAE to use this. For the examples, I've used the Grapefruit one:

https://huggingface.co/iZELX1/Grapefruit/resolve/main/Grapefruit.vae.pt

What's new in v2?

I wasn't satisfied by the overall performance of the checkpoint. It often used the same colors and required a low CFG level. With the newer model you can use a higher CFG level without over saturation and distortion. It's less Disney-ish and more a crossover of vintage western and modern anime style (which is what I wanted to create in the first place).

What's this model

Mistoon_Amethyst is a variant of my SD model Mistoon_Sapphire that tries to achieve a Disney pinup style.

The starting point for this model was my blend. From there, I then added a few of my custom LoRAs in the mix to change the esthetic (mostly Pear and Contact).

Try also the other flavors:

The style

It creates realistic and expressive characters with a "disney" twist.

All the examples have been created using this version of the model using my workflow. Check the guide if you want to learn more about my workflow.

The model can also easily do NSFW stuff (check the last examples).

Right now the model is biased over sea backgrounds and cyan clothes (for whatever reason), so I highly recommend addind those to the negative prompt.

Guides

Learn how to create pictures like mine with my step-by-step tutorials:

• Beginner's Guide

• Prompting

• My Workflow

• Inpainting

Support

Do you want a custom LoRA or picture? Check out my commission page:

LoRA and Picture Commissions

Do you want to get access to my LoRAs before everyone, and contribute to my next models? Check out my supporter tiers on Deviant Art:

If you want to support my work for free, you can also check out my music/art here:

• Bandcamp

• Instagram

• DeviantArt

Hi,

This is an anime/cartoon model.

If you like my work and want to buy me a coffe, you can do it here: https://ko-fi.com/alpharadix patreon.com/Radix382

The model is trained on 15K NSFW anime images that features two person interaction.

The dataset is so restricted to the point that it becomes an one trick pony in NSFW generation.

The base model is Sweet-mix 1.5, so prompting rules apply, especially on negative textual inversions. The only difference is that you should not apply any vae, otherwise the result will be too bright.

Due to the one trick nature of this model, you may have to generate NSFW content, controlnet to lock, then use another model to fill in.

Many thanks to the creator of sweet-mix

https://civitai.com/models/18927/sweet-mix

• Recommended Settings

1. VAE : kl-f8-anime2

2. Steps : More than 25 steps

3. CFG scale : 6-8

4. Sampler : DPM++ 2M Karras

5. Hires upscaler : R-ESRGAN 4x+ Anime6B or Latent (nearest-exact)

6. Negative prompts : easynegative, (worst quality, low quality:1.4), (multiple views:1.2), (variations:1.2), (monochrome:1.1), watermark

• All the samples I uploaded were used with highres + SD upscale. There are more samples in my pixiv. https://www.pixiv.net/users/86784786

Trained on immunofluorescence images of human cytomegalovirus infection in fibroblasts at x40 magnification. Nucleus stained blue (DAPI), early endosomes red (EEA1), golgi in green (GM130).

※このモデルにはいくつかのバリエーションがあることを見逃さないでください

　Note:Don't miss that there are several variations of this model!

,

caulkinumV2 series for Anime characters.

Note: If you like it, please give it a STAR MARKS !

,

The caulkinumV2_AR series is a model designed to beautifully portray anime characters and light.

This checkpoint was spawned accidentally through multiple iterations of hierarchical merging.

Therefore, the detailed recipe is already lost.

,

However, there is no doubt that Silicon28 is at the root of it and that it has inherited the influence of ElysiumAnime_V2.

If someone likes it I will be very happy.

,

The caulkinumV2 series has several variations.

"AR" is the first model, which is a modern anime/game style.

• https://civitai.com/models/32537?modelVersionId=38959

"AR2" is the normal evolution of the first model, and it can create better illustrations.

• https://civitai.com/models/32537?modelVersionId=48531

"A4FS" has a dramatic style, which is more in line with recent RPG games.

• https://civitai.com/models/32537?modelVersionId=48630

"ARCT" is an anime-style painting that makes the characters look young.

• https://civitai.com/models/32537?modelVersionId=48791

"ARc" has the same anime style, but is a bit more difficult to handle and a bit outlandish.

• https://civitai.com/models/32537?modelVersionId=48898

"NCA" is a little bit special and unique, but it was created to depict a slime girl.

• https://civitai.com/models/32537?modelVersionId=50350

"NLPS" is a very special model, and if you master it, you will experience a completely different world.

• https://civitai.com/models/32537?modelVersionId=51205

"FGA" will be a model for drawing fluffy beasties

• https://civitai.com/models/32537?modelVersionId=56569

,

A sample comparison is provided below.

It does not always work, but I am generally satisfied with the results.

Note: Underscores at the beginning of checkpoint names are assigned by AUTOMATIC1111 for sorting purposes and have no particular meaning.

,

If the samples are too hard to see and judge, you can dl and try them out.

Then post and show us your great results!

Or you can use ctrl+wheel.

,

,

caulkinumV2（アニメキャラクター用）

※ 気に入っていただけたならぜひ★をつけてください！

,

caulkinumV2_ARシリーズは、アニメのキャラクターと光の表現を美しく描くために考案したモデルです。

このチェックポイントは複数回の階層マージを繰り返して偶発的に産まれました。

そのため細かいレシピはもう失われています。

,

しかしSilicon28が根底にあること、ElysiumAnime_V2の影響を受け継いでいることは間違いありません。

誰かが気に入ってくれたなら私はとても幸せでしょう。

,

caulkinumV2シリーズはいくつかのバリエーションがあります。

「AR」が最初のモデルでモダンなアニメやゲームの画風です

• https://civitai.com/models/32537?modelVersionId=48531

「AR2」は正常進化版でより良いイラストを錬成できます

• https://civitai.com/models/32537?modelVersionId=48531

「A4FS」は劇画調で近年のRPGゲームらしさが表現できます

• https://civitai.com/models/32537?modelVersionId=48630

「ARCT」はアニメ風の塗りでキャラクターは幼くなるでしょう

• https://civitai.com/models/32537?modelVersionId=48791

「ARc」も同様にアニメの画風ですが、やや扱いづらく突飛なところがあります

• https://civitai.com/models/32537?modelVersionId=48898

「NCA」は少し特殊で独特の画風ですが、スライム娘を描くために作りました。

• https://civitai.com/models/32537?modelVersionId=50350

「NLPS」はかなり特殊なモデルで、使いこなせば今までとはまったく異なる世界を味わえるでしょう。

• https://civitai.com/models/32537?modelVersionId=51205

「FGA」はもふもふな獣人を描くためのモデルとなります

• https://civitai.com/models/32537?modelVersionId=56569

,

下記に比較サンプルを掲載しておきます。

必ずしもうまくいくとは限りませんが、私はおおむね満足しています。

注釈 ) チェックポイント名先頭のアンダースコアはAUTOMATIC1111で並べ替えのために付与したもので特に意味はありません

,

サンプルが見えづらく判断に困る場合、あなたはDLして試用することが可能です。

そして投稿してあなたの素晴らしい成果を見せてください！

もしくは ctrl+wheel を使う方法もあります。

,

※ You can download the original image at HuggingFace

　(HuggingFaceで元画像をダウンロードできます)

- https://huggingface.co/Shiyaku/caulkinumCheckPoint/blob/main/README.md

,

,

,

caulkinum Series Checkpoint (caulkinumシリーズ チェックポイント)

caulkinumV2 for Anime Style

• https://civitai.com/models/32537/caulkinumv2ar-for-anime-style

caulkinumV1_RL for semi Realistic Style

• https://civitai.com/models/32958/caulkinumv1rl-for-25d-style

caulkinumV1_S for Realistic Style

• https://civitai.com/models/45689/caulkinumv1s-series-for-realistic-style

,

,

Donation (寄付)

Coffee will increase work efficiency and motivation.

コーヒーは作業効率を高めモチベーションをアップさせるでしょう。

https://ko-fi.com/489489

,

Thank you for reading to the end.

I hope one of you will use it.

,

最後まで読んでくれてありがとうございます。

使用してくれるか方が一人でもいることを祈っています。

This model is a hierarchical merge model based on Defacta or Dexmia.
I, the model’s author, created it for study and personal use.
Note that VAE is not included, so please download it separately.

We've released a new model, V3.5, as an update to the V3 model. At the time of creating the V3 model, a new base model was released, so we updated it to version 3.5.
The V3.5 model leans towards a more realistic style, but we're confident that it's also a versatile performer that can produce anime-style results. It particularly excels when using art-related prompts. We believe it can generate high-quality outputs, so please feel free to experiment with it. For detailed usage instructions, please refer to Hugging Face.

Hugging Face: https://huggingface.co/teasan/endlessMix

author:
https://twitter.com/wims_Tea

Disclaimer:

• The creation of SFW and NSFW images is at the discretion of individual creators. The model maker is not responsible.

• This model was not created for the purpose of publicly displaying NSFW content, such as in a public setting.

Check out my other model! Chronos

About Analog Madness

A very versatile model, the more powerfull prompts you give, the better results.

Capable of creating both NSFW and SFW images but also great scenery,
both in landscape and portrait.

Example prompt:

hyper realistic gopro action photo of a beautiful 20yo Dutch girl with small breasts (looking at camera:1.3), windy, wearing old trashy worn torn out clothes, in a themepark, analog style, masterpiece, exposed breasts

negative prompt:
3d max, grotesque, desaturated

tnx to Hoblin for this negative prompt!

Euler a, 25 to 35 steps, CFG7

Disclaimer:
All characters on my images are 18+. No likeness to real people intended. No pixels were hurt during the production of this images.

This model is a merged model that utilizes three models and is designed to maximize hand gestures and posture expressions. In addition, by applying positive prompts such as "cute girl mix" and "pure eros face," it is possible to create a different feel of facial expressions.

http://youtube.com/@sanglook

https://www.patreon.com/VisualEmotion

Son Gohan SSJ2

female

Brainime

Will produce almost anything (as long as it's anime) without too much of a prompt, see Notes below.

This model prefers natural language than long comma-separated chunks. It will still work with it, just not as well!

Your feedback is essential to improve the future updates, please consider leaving a comment or a review 🙏

Recommended Settings:

• Sampler

  • DDIM

  • DPM++ variants

• CFG Scale

  • 3~7 (I leave mine around 5 usually)

• Steps

  • With Hi-Res: 20~30

  • Without Hi-Res: >=50

• Hi-Res highly recommended

Recipe

This is my first merge, I've sort of lost the recipe at this point but it's using mostly

• CoffeeMix

• BluePencil

Notes

• Highly recommended to use EasyNegative

I'm still new to this so I'll add more details on here as I get feedback from the community.

This is a Mix Mash of Training and merging. First model... Use your own vae... you should have one? Selections based upon Euler a 40 steps no hirez 512x512 profile.

This should be good for general use.

50/50 Split of Eris and color Fusion models

A Basic merge of Eris and Color Fusion - This one should generate some unique mixes, this is mostly aimed at those who want a color infusied, but 3d rendered cartoon type model.

Also perfect for mixing into other models. This is a seperate upload from the mega model series. - May produce the odd nsfw shot so if u dont want those use a negative prompt

Color Fusion - Color Fusion - ver 1.o | Stable Diffusion Checkpoint | Civitai

Eris - Civitai | Share your models

2023.4.27 Change model to safetensor.

Click here if you want ckpt version.

This model is created by merging many LoRa and other models to produce pictures of cute furry girls. I hope you like it. (*^▽^*)

融合了一些模型和Lora得到的模型，可以生产可爱福瑞图。希望各位用得开心。(*^▽^*)

Recommend settings:

512×768, DPM++ 2M Karras, Steps: 20-30, and add “furry, animal ears, tail, " to prompt.

Used models:

NineKeyMIX5

Mistoon_Anime

MIX-HIRES
Used Loras:

Squeezer - Experimental

Rikose Style

Nukunuku Nigirimeshi
Furry girl

Anime smooth and shiny

XSAchi-InteriorDesginV5ForCN

Suggestions for use:

• Lora, TI, VAE and other documents for indoor use

• This data model has a corresponding LoRa model. If you wish to use it, please download it:

• XSarchitectural-7Modern interior | Stable Diffusion LORA | Civitai

• XSarchitectural-38InteriorForBedroom | Stable Diffusion LORA | Civitai

• 禁止商业用途，模型已获多国所有权（CN、JA、US...）

• 商業利用は禁止されており、モデルは複数の国で所有されています(CN、JA、US…)

• Commercial use is prohibited, the model has been granted multi-country ownership (CN, JA, US...)

• If you think it's good, give me a five-star review

• If you have any questions, please contact me at xsarchitecturalstudio@gmail.com

Simple but interesting mix I made for use in my own works. V1.1 is a 55:45 weighted sum mix of MeinaMixV9 and Aurora, with a 0.3 weighted sum of NeverEnding Dream with that. The result is pretty nice IMO, giving a stylized yet highly detailed set of images. I don't take credit for doing anything beyond pressing the merge button, so go like and support the others.

Source Models:
MeinaMixV9
Aurora
NeverEnding Dream

In the interest of honesty I will disclose that many of these pictures here have been cherry picked, hand-edited and re-generated. That's because the majority are working pieces of concept art for a story I'm working on. I don't want to give the impression that this model can't give good first-generation images - it can - but when using it you should just kind of expect to have to hi-res fix and/or inpaint at least once. None of the pictures posted here use a Lora for style, but as part of the typical prompt I use they do use EasyNegative and BadHandV4 at usually reduced weights (0.6 and 0.4 respectively). The ones that do use Lora are only for character specific features (Moth Girls, Obsidian Skin, Black Sclera, etc.). All images were post-process upscaled with 4x_NMKD-Superscale-SP_178000_G to give a painted look. If you're wondering why all my pictures have extremely high CFG scales, its because I use Dynamic Thresholding. If you're doing anything where you want a very specific result based on prompts alone, I highly recommend it. The parameters I use for all images are:
Mimic CFG Scale - 7 (rarely 3)
Mimic/CFG Scale Scheduler - Half Cosine Up
Min. Value of Mimic/CFG Scale Scheduler - 3

Below are recommended gen info based on the stuff I typically use, combined with legal stuff inherited from NED, Aurora and MeinaMix, who thankfully use the same license parameters. This mix uses CreativeML OpenRAIL-M as the two previous models do, and since Meina disallows the use of the model on generation services or paid services without her consent, this model is similarly banned for use in similar contexts.

Recommendations:

Enable Quantization in K samplers.

Hires.fix in some form is needed for high quality images. You can either use the t2i hires fix, which I'd recommend running at 1.5-2x scale at 0.3 denoising for 15 steps. I do this manually by running the best finished images through img2img

Recommended parameters:
Sampler: DPM++ 2M Karras: 25-35 steps.
CFG Scale: 7+. See note on Dynamic Thresholding above.
Resolutions:
-Default: 512x512 txt2img -> 1024x1024 img2img, 0.3-0.65 Denoising
-512x768, 512x1024 for Portrait
-768x432, 1024x576 for Landscape (16:9)
Hires.fix: 4x_NMKD-Superscale-SP_178000_G, with 15 steps at 0.1 up to 0.3 denoising.
Clip Skip: 2.
Negatives: ' (worst quality:2, low quality:2), (zombie, sketch, interlocked fingers, comic), '
--------------------------------------------------------------------------------

License

This embedding is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:

1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content 
2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
3. You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully)

Please read the full license here

Disclaimer

The use of this learning model is entirely at the discretion of the user, and they have the freedom to choose whether or not to create NSFW content. 
This is important to note that the model itself does not contain any explicit or inappropriate imagery that can be easily accessed with a single click. 
The purpose of sharing this model is not to showcase obscene material in a public forum, but rather to provide a tool for users to utilize as they see fit. 
The decision of whether to engage with SFW or NSFW content lies with the user and their own personal preferences.

I have poured my soul into this materpiece. This is a mix of over 50 MODELS. If you appreciate my work, kindly consider supporting me:

• https://ko-fi.com/lyloguumy

As a creator, I have always been passionate about creating digital art that blends realism with a touch of anime. That's why I'm excited to introduce you to my latest project: IceRealistic, a unique simulation model that generates stunningly realistic images of ice and snow, with a twist - my model features beautifully rendered anime girls that bring a new level of charm and elegance to the art.

IceRealistic is perfect for anyone looking to create game assets, architectural visualizations, or scientific research projects that require realistic renders. With my model, you can create stunningly detailed images that will capture the imagination of your audience, with the added bonus of a touch of anime charm that will make your work stand out from the crowd.

But to continue creating and improving my model, I need your help. Donating to my ko-fi page will allow me to invest more time and resources into my work, allowing me to create even more stunning and realistic images for you to enjoy. As a token of my appreciation, I will provide exclusive content and perks to all of my donors, including early access to new updates and personalized tutorials.

So if you're a fan of anime and love seeing digital art come to life, please consider donating to my ko-fi page. With your support, I can continue to create amazing art that captures the beauty of the natural world, with a touch of anime magic. Thank you for your support!

The beggining of a great project !

Providence - Prelude 

• A precise chosen noise offset setting

• A large variety of subjects, I took care to to keep the flavour at maximum from genesis seeds feel

• No baked VAE (works fine without any tbh)

Hyper Realism

For good results please use the forAnalogPortrait negative embedding.

Digital painting

For my digital painting finetune, please use “character portrait” along “digital painting” for enhance the quality

Analog

use the forAnalogPortrait negative embedding aswell.

Troubleshoot :

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

A 2.1 model require either --xformers and / or --no-half start argument from the webui-user.bat

set COMMANDLINE_ARGS= --no-half --xformers

check also if your config file is named correctly alongside the model

Happy tuning !

romerorz.art - Twitter - DeviantArt - Gumroad

https://discord.gg/MDpfe3JBQX

https://ko-fi.com/bgmodel

Please read the description for the models i put out. There is some good information that may increase the quality of your output when using my models.

test ckpt trained on Mike Mignola

To activate the token, please use the word without the class . If the character loses prominence, consider adding parentheses to the token to emphasize the character more, like this: ((elxchavo)).

El Chavo del 8 - Realistic Vision V2 & Deliberate V2 Models Introduction This repository contains two models trained on a dataset of 61 images from the popular Latin American TV show "El Chavo del 8". The models are based on Realistic Vision V2 and Deliberate V2 architectures.

Dataset The dataset consists of 61 images from the "El Chavo del 8" show, with manual captions added. The images were collected from various episodes to ensure diversity.

Training Process Both models were trained using the Dreambooth repository. Checkpoints were conducted every 1000 steps to monitor the performance and behavior of the dataset with these models. The training process aimed to generate realistic and deliberate visual representations based on the images and captions provided.

Please feel free to contribute to this repository or use these models in your projects. For any questions or feedback, reach out to us.

the other models to try are in this link:

aaronamortegui/elxchavo · Hugging Face

All samples are raw txt2img generations with no ControlNet, img2img or other editing applied!

To be done still:

• Finish this description

• Update Huggingface page

• Add more samples

• Upload safetensors, etc

Char-v2.0 by AI_Characters

A 512 resolution model trained (not merged) on SD v1.5 for various high-quality artstyles and established characters.

If you would like to support my model creation (I have spent the past 5 months and over 1000€ on model creation to reach this level of quality today) you can donate to me on KoFi: https://ko-fi.com/aicharacters

In case you generate some beautiful artwork using my model, feel free to tag me with it on Twitter and/or Instagram!

https://www.instagram.com/ai_characters/

https://twitter.com/ai_characters/

Changes from v1.0 to v2.0

Everything. It is a completely new and different model. Most of the issues from v1.0 are gone and the quality and likeness is on a whole different level. Forget that v1.0 ever existed. Delete it from your system. V2.0 is the model v1.0 was always meant to be.

V2.0 was trained on a 512 resolution instead of a 1024 resolution, so for optimal interference you have to generate at 512 and high-res fix/img2img to 1024 res. Still, the quality is vastly superior to v1.0 despite the lower trained res. This is because higher res training resulting in higher quality seems to have been a fluke, and the real issue were my training settings and data.

Issues with this model and how to fix them

• Some prompts and artstyles are still biased towards certain ethnicities, character looks, or other things, use longer and more detailed positive prompts or negative prompts in this case to get what you want!

• Negative prompts are still finicky and I only recommend to use them sparingly and only when needed, they can make a generation worse!

• Just like with negative prompts, use prompt weights sparingly

• Some prompts may still feature incoherences with the body proportions or other aspects of the image

• Some SD v1.5 stuff like Emma Watson and such seem to have vanished from the model, despite the training shouldn't cause that, unfortunately I cannot reverse that at this moment without destroying my model's likeness, use img2img or ControlNet if you really want to do something with that in my model

• Not all characters are 100% flexible in the outfits and hairstyles they can wear as well as the artstyles they can be portrayed in, however, you generally can still portray every character in at least some different outfits and artstyles

• Not all prompts work well in all artstyles, e.g. giant people seem to work only in renders and digital art pieces, not for example photos

Generally I highly recommend using ControlNet with my model as it will provide the coherency and consistency, while my model can provide the quality and different artstyles.

Recommended settings for interference

These are my recommended settings for quick high quality results. These are also the settings on which I tested the model.

- Sampler: Euler_a

- Steps: 20

- CFG: 7

- Resolution: 512x512 + using high-res fix to 1024x1024

Recommend way to prompt this model

This model was NOT trained using danbooru tags, so danbooru tags are generally not recommend to be used with this model. This model also generally shouldn't need these huge and long positive and negative prompts to work correctly, and those in fact may make the model perform worse at times!

Generally, the model was captioned the following way:

"(beautiful lighting), POV, (from behind), style, image description"

Examples:

• full-length, photo, woman

• beautiful lighting, medium-shot, photo, woman wearing tshirt, standing in city street

• gh artstyle screencap, fantasy castle upon a hill next to a river

Trained tags

Not all artstyle and outfit tags will have perfect likeness, but they will get close.

Artstyles

• photo

• digital art

• lk artstyle screencap (Legend of Korra artstyle)

• hg artstyle render (realistic 3d renders)

• cw artstyle render (StarWars: The Clone Wars artstyle)

• gh artstyle screencap (Ghibli movies artstyle)

• ms artstyle screencap (generic anime artstyle influenced by Makoto Shinkai and SAO)

• disney artstyle render (3D Disney artstyle)

• hq artstyle digital art (high-quality detailed realistic digital art)

• dd artstyle digital art (Darkest Dungeon artstyle)

• tc artstyle digital art (Turtle of Canada artstyle (with their permission))

• anime screencap (generic animated artstyle)

POV

• full-length

• medium-shot

• closeup

• longshot

• headshot

Misc

• beautiful lighting

• from behind

• silhouette of

• street art of

• humanoid animal (e.g. humanoid fox)

• half-x half-y hybrid person/animal (e.g. half-turtle half-duck hybrid animal)

• anthro animal person (e.g. anthropomorphic fox)

• female/male x furry (e.g. female fox furry)

• dragon

• person with feathered/dragon/insect/butterfly/fairy wings

• giant woman

• miniature woman

Characters

• kr (Korra)

• ns (Nausicaä)

• sn (San)

• kk (Kiki)

• vd (Videl)

• al (Aloy)

• ah (Ahsoka)

Outfits

• sweatjacket

• sg outfit with cape (Supergirl outfit)

• sy outfit with cape and helmet (Great Saiyaman 2 outfit)

• df outfit (Korra season 1/3 outfit)

• sc outfit (Korra season 1 winter outfit)

• pb outfit with helmet (Korra probending outfit)

• fs outfit (Korra season 4 outfit)

• ek outfit (Korra season 4 earth kingdom outfit)

• hz outfit (Aloy default Nora outfit)

• nz outfit (Aloy generic game outfit)

• sw outfit (Ahsoka The Clone Wars season 7 outfit)

• bl outfit (Nausicaä)

• mk outfit (San)

• kd outfit (Kiki)

Hairstyles

• sidecut hairstyle

• ponytail hairstyle

• sd hairstyle (Korra season 1-3)

• sh hairstyle (Korra season 4)

• nr hairstyle (Aloy)

• tl hairstyle (Ahsoka lekku)

• rh hairstyle (Nausicaä)

• sb hairstyle (Kiki)

• jf hairstyle (San)

FTU-NoiseOffset0.5fp16 was just the temporary name of colorful version when I tested the model.

I don't know how to change the model name in gereration data ....

Dark Sushi 2.5D has been released.

If you like 2.5D style,you can try it.

*****************

VAE is vae-ft-mse-840000-ema.

起名废玩烂梗系列，事后想想起的不错。

首先暗图效果比较好，dark合适，自己融的时候差不多也是寿司那样什么都塞点，dark 寿司没毛病。

光效比较好，能生成比较暗的图，正常亮度的图片也有不错的光影表现。

融了群友私模和一大堆我也记不住组成的以前的自融模

能确定包含的是自己之前的五仁Five Nuts Mixed 五仁月饼Mix | Stable Diffusion Checkpoint | Civitai 和Unrule | Stable Diffusion Checkpoint | Civitai

以及群友的TmndMix ，绝对的杰作，推荐大家都去试试 https://civitai.com/models/27259

darker版本里还加了noiseoffset的lora，生图要比brighter版本暗一些，光影表现上要强一点，但缺点就是正常生图也会比较暗。大家一般用brighter版再自己调用noiseoffset的lora也能达到类似效果。

Intro:

named from a dark soul's meme in Chinese community.

This model has good light effects, can generate darker images, and also has good light and shadow performance on normal brightness images.

merged with my friend's pravite model and some other models merged by myself.

Including Five Nuts Mixed 五仁月饼Mix | Stable Diffusion Checkpoint | Civitai ,

Unrule | Stable Diffusion Checkpoint | Civitai

and my friend's TmndMix https://civitai.com/models/27259 .it's definitely a masterpiece.

if you like dark sushi ,you should also try tmnd.

noiseoffset lora is merged into the darker version. So the imgs generated by the darker version will be darker ,and have better light and shadow performance. you can use brighter version and noiseoffset lora to get similar effect.

第四版是重新融合的，第三版有我个人觉得比较失败的色彩问题，但有些朋友用着也不错所以留着。

The forth edition is a remake, not particular better than previous editions but is good to try. The third one has some failure in color, but it works for some people so I will just leave it here.

在第三版我继续融合了改善身体的模型，以及加强了光影。

很抱歉在上一版的例图中我使用了分层的lora让大家困惑，也让大家复刻我的例图变得困难。所以新一版的例图我没有使用任何lora。想了解lora分层的可以参考：GitHub - hako-mikan/sd-webui-lora-block-weight

In the third edition, I continued to merge to improve body and lighting.

I apologize for using lora block weight in the example images of the previous edition, which confused most of you and made it difficult for you to replicate my examples. Therefore, in the new edition, I did not use any lora in my showcase. If you would like to learn about lora block weight, please refer to: https://github.com/hako-mikan/sd-webui-lora-block-weight

我的分层参数 My lora block setting:

身体

BODY:1,1,1,1,1,1,1,1,0,0,0,1,1,1,1,1,1

BODY0.5:1,1,1,1,1,1,0.2,1,0.2,0,0,0.8,1,1,1,1,1

脸部（脸型、发型、眼型、瞳色等）

FACE:1,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0

FACE0.5:1,0,0,0,0,0,0,0,0.8,1,1,0.2,0,0,0,0,0

FACE0.2:1,0,0,0,0,0,0,0,0.2,0.6,0.8,0.2,0,0,0,0,0

修手专用

HAND:1,0,1,1,0.2,0,0,0,0,0,0,0,0,0,0,0,0

服装（搭配tag使用）

CLOTHING:1,1,1,1,1,0,0.2,0,0.8,1,1,0.2,0,0,0,0,0

动作（搭配tag使用）

POSE:1,0,0,0,0,0,0.2,1,1,1,0,0,0,0,0,0,0

上色风格（搭配tag使用）

PALETTE:1,0,0,0,0,0,0,0,0,0,0,0.8,1,1,1,1,1

角色（去风格化）

KEEPCHAR:1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0

背景（去风格化）

KEEPBG:1,1,1,1,1,1,0.2,1,0.2,0,0,0.8,1,1,1,0,0

减弱过拟合（等同于OUTALL）

REDUCEFIT:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1

第二版我改善了一下特定成分的权重，也改善了光影

In second edition I adjust the weight of certain element so that the hands and body can be better. I also adjust the lighting slightly.

新的展示图用了胶片风 In show case I used 胶片风女孩LoRA

融合了多种模型，可以生成好看的脸部，也能有效应对暗部处理。远距离脸部需要inpaint以达成最好效果。

A good looking model, suitable for NSFW and dark scene (because I added noiseoffset). long-range facial detail require inpainting to achieve the best results.

听我一句劝，不要开脸部修复！不要开脸部修复！

Please don't use Face Restoration! Please don't use Face Restoration!

推荐关键词 recommended positive prompts: Best quality, masterpiece, ultra high res, (photorealistic:1.4), 1girl

如果想要更暗的图像 if you want darker picture, add: in the dark, deep shadow, low key, etc.

负面关键词 use ng_deepnegative_v1_75t and badhandv4 in negative prompt

I've used a bug-fixed version of DPM++ 2M Karras, you can check this out: https://civitai.com/models/35966/dpm-2m-alt-karras-sampler

推荐参数Recommended Parameters:

Sampler: DPM++ 2M Karras (bug-fixed) or DPM++ SDE Karras

Steps: 20~40

Hires upscaler: R-ESRGAN 4x+ or 4x-UltraSharp

Hires upscale: 2

Hires steps: 15

Denoising strength: 0.2~0.5

CFG scale: 6-8

clip skip 2

脸部修复的方法 to inpaint the face: inpaint-->only masked-->set to 512x512-->Denoising strength:0.2~0.5

basic formula:
KanPiroMix + XSMix + ChikMix

关注我的TG频道看更多例图：https://t.me/majic_NSFW

This is a three way mix of Refinedv10, EpicRealism, and Chilloutmix. It's been pretty good at generating Asian, Middle Eastern, and Indian figures.

CTUniji0.3fp16 was just the temporary name of v2.0 when I tested the model.

I don't know how to change the model name in gereration data ....

using MBW to merge Chillout and Dark Sushi

and also add a little NijiExpressV2

* Version 3.0 Update: Add Noise offset And ClearVAE for 2.5-3D Model style

Prompt

• (General using prompt)

Negative

• (EasyNegative, badhandv4, ng_deepnegative_v1_75t,)

Default Setting.

DPM++ SDE KARRAS

Sampling: Steps 20-30

Upscaler: 4x-UltraSharp

Hires Steps: 10-20

Denoising Strength: 0.4-0.6

CFG Scale: 7-10 (CFG More = More Contrast)

Mimic CFG Scale: 7.0

Clip Skip: 2

Recommended VAE: ClearVAE

这个模型是根据中国诡异风格的图片集训练而来，主要是生成一些诡异风格的图片，目前是试用版本，正在做优化。

经过测试对半身的效果更好，对一些中式服装和饰品表现良好，但是全身的角色容易出现身体错乱，如果你需要这样的错乱增加画面的诡异感，也是一个不错的选择，后续还将继续调整更新。

[Some example images have gone through SDupscale process]

V2 update

Little more realistic texture and more natural face shape

It's semi-realistic model and Good for making a female character with sexy body

Can generate NSFW image, male characters and Landscape

▼Please check settings and prompts▼

Merge model

ZemiHR

dalcefo_v4

revAnimated_v11

cetusMix_Version35

Hipoly 3D Model LoRA V2

Pastel-Mix LoRA

Process

0.7(ZemiHR) + 0.3(revAnimated_v11) = A

0.85(A) + 0.15(dalcefo_v4 =B

B + 0.15(Pastel-Mix LoRA) + 0.2(Hipoly 3D Model LoRA V2) = C

0.9(C) + 0.1(cetusMix_Version35) = ZemiHR V2

Recommended setting

Sampling method : DPM++ SDE Karras, (DPM++ 2M Karras, DDIM)

Clip skip : 2 

Hires steps : 13

Hires.fix upscaler : R-ESRGAN Genaeral WDN 4xV3, (R-ESRGAN 4x+Anime6B) 

CFG Scale : 7~10

VAE : vae-tf-mse-840000

Recommended prompt

Prompt : (masterpiece:1.2, best quality) 

Negative : (worst quality, low quality:1.4), EasyNegative, badhandv4 + (blush, see through:1.2)

Textual Inversion : EasyNegative, Badhandv4

Something that I've been throwing together over time and I think I've come to something stable enough to share with others. I can't remember everything I've put into it.

Has some good performance as low as 27 steps and I personally prefer DPM++ 2S a Karras and DPM++ 2M Karras, though it performs fairly well with all samplers.

The components in no particular order are: AnyHentai1.7, Bastard v4 Anime, Clockwork Oranges, Corneos, DHClassic, Digitals4wed0ff, DH SuperCute, ExpMix, Kazuki Merge, Moyu, Hassaku, DeGradeRev, Lazy Amateur Mix, URPM 1.5, SirenMix

A merge checkpoint of KizamiSakura and momoko.

刻樱+桃子=樱桃。

简单易上手的动画风格，不过低分辨率下，眼睛里的高光会逐渐消失，记得开高分辨率修复。

a dark model for dark fantasies

Thank you so much for the feedback and examples of your work! It's very motivating.

Do you like what I do? Feel free to buy me a coffee ☕

My models:

• Colorful

• Darkful

• Animatrix

New model created by me

setting

• - Sampler: DPM++ 2M Karras

• - Step: 20 ~ 50

• - CFG scale: 7～15

• - Denoising strength: 0.55 ~ 0.6

• - Clip skip: 2

• - Hires upscale: 2

• - Hires steps: 10 ~ 13 or 0

• - Hires upscaler: Latent (nearest)

• - VAE: kl-f8-anime2

author:

https://twitter.com/wims_Tea

Disclaimer:

• The creation of SFW and NSFW images is at the discretion of individual creators. The model maker is not responsible.

• This model was not created for the purpose of publicly displaying NSFW content, such as in a public setting.

'Creativity is not just creating works of art, but the need to express one's uniqueness and leave a mark on the history of mankind.'

\(￣︶￣*\))

Model for Portrait Photography

// Update V4 //

Realistic
Sharpen
Beautiful Face

// Recommend //

Step 30
Text Guidance 7.5
DPM 2M Karras

***Large Image Size for High Quality

リアル系マージモデルです。日本人を始めとするアジア系の再現ができるように調整しています。特にjapanese doll likenessとの親和性を意識しています。

このマージモデルを公開するにあたり、使用したモデルの製作者の皆様に感謝申し上げます。

This is a realistic merge model. It is adjusted to reproduce Japanese and other Asian descent. We are especially conscious of the affinity with japanese doll likeness.

We would like to thank the creators of the models we used for making this merge model available to the public.

マージ使用モデル(Merge use model)

MUSE_v1

https://civitai.com/models/13564/musev1

Colorful

https://civitai.com/models/7279/colorful

majicMIX realistic

https://civitai.com/models/43331/majicmix-realistic

V1.1より以下に変更

XXMix_9

https://civitai.com/models/47274/xxmix9

本モデルも使用したモデルの利用条件に従う形になりますが、以下に関しては厳に使用を禁止いたします。

・暴力的な表現

・児童ポルノ

・未成年者の性的な表現

本モデルは『CreativeML Open RAIL-M』の範囲でラインセンスされます。 本モデルを使用した上での問題に関しては、当方は一切責任を持ちません。ご了承の上ご使用ください。 また、マージモデルのライセンス変更に伴い、公開を停止することがあります。

This model is also subject to the terms and conditions of use of the model used, but the use of the following is strictly prohibited.

・Violent expressions

・Child pornography

・Sexual expression of minors

This model is licensed within the scope of "CreativeML Open RAIL-M". We are not responsible for any problems caused by the use of this model. Please use this model with your understanding. We reserve the right to suspend publication of the merged model due to changes in its license.

VivCharMix - Effect of transforming ACG characters into real-life people is outstanding

(The English translation can be found below the text)

這個模組將ACG角色轉換成現實人物的效果非常出色

如同從圖畫中活生生過來，

避免像Coser(真人對應不足、外表太過真實)、人偶(眼過大、小鼻子)、手辦(表情僵硬、塑膠感)，使生成的真人角色更貼近原作

必備：

EasyNegative

veryBadImageNegative

Oekaki negative TI

Negative Embedding for Realistic Vision v2.0

建議負面詞： (EasyNegative verybadimagenegative_v1.3, sd1-failed-oekaki:0.8), ( realisticvision-negative-embedding:1.1)

使用Hires Fix整數縮放(最接近) Nearest可以得到最像真人結果

這個是經過微調的比例，可以有更好的寬容性和結果
EasyNegative 越高臉型越符合，但畫面會越銳利

verybadimagenegative_v1.3 越高衣物細節越正確，但像油畫

sd1-failed-oekaki 越高背景越好，但越像小孩而且嬌小

可以視需求微調

因為所有微調都是建立在這組負面詞上

若不使用這組負面詞，無法達到我煉製和分享的模組預期

不需要VAE

優點：
(1) 臉部好看的生成模型，適合生成ACG人物真人化，保留氣質和神韻和衣物細節

(2) 對ACG的Loras有高適應性和寬容度(建議權重0.3~0.8，用0.05當作一個區間調整)

(3) 鮮明飽滿的顏色和動態的表情

(4) 可以相對正確的生成NSFW內容，以及可正確表現體液

(5) 對畫風style類Loras敏感，大概0.3就足以改變輸出結果，0.6會大幅影響畫面，增加可玩性。 對姿勢類、服裝類Loras也有相當好支援性

(6) 可以產生ACG中出現的裝飾元素，例如spoken heart、emphasis lines、speech bubble...

--> 使用新的負面詞後，甚至可出現紅眼、綠眼

(7) 適合當作ACG轉真人圖生圖的模型

負面詞：EasyNegative, verybadimagenegative_v1.3, sd1-failed-oekaki、修改幅度0.45

經過兩次img to img

缺點：

(1) 場景單一，特別容易出現日常或素色背景，建議用Loras或Prompts調整

解決，加上Oekaki negative TI負面詞後，會擁有更出色的品質和印象深刻的背景

(2) 沒有對ACG男性角色微調，可能導致娘化

好看的阿福

(3) masterpiece關鍵字被錯誤綁定，容易生成nsfw內容

(4) 使用新關鍵字後，身材過瘦、過於平坦，我覺得如果更凹凸有致
並且有點肌肉會更好看。 或著可以用prompt或Loras調整

目前找不到更好的相同的model可以merge，在我能學會mwb之前這個就是完成版了

特別感謝：
KanPiroMix

3moon anime_cosplay

3moon REAL_Photo

DDosmix

Lofi V2

與其他真人化效果

原圖 Original image

加上新的負面詞Oekaki negative TI重新比較

原圖2 Original image 2

(Translated by Chatgpt)

VivCharMix

This module's main purpose is to transform ACG characters into real-life people.

It brings characters to life from drawings while avoiding issues like cosplayers (insufficient real-life representation, appearance too REALISTIC), dolls (oversized eyes, small noses), and figurines (stiff expressions, plastic feel).

Requirement:

EasyNegative

veryBadImageNegative

Oekaki negative TI

Negative Embedding for Realistic Vision v2.0

(EasyNegative verybadimagenegative_v1.3, sd1-failed-oekaki:0.8), ( realisticvision-negative-embedding:1.1)

Using Hires Fix-Nearest can achieve the most realistic results.

This is a fine-tuned proportion that offers better tolerance and results.

The higher the EasyNegative, the more accurate the face shape, but the image will be sharper.

The higher the verybadimagenegative_v1.3, the more accurate the clothing details, but the output will resemble an oil painting.

The higher the sd1-failed-oekaki, the better the background, but the result will look more childlike and petite.

You can fine-tune according to your needs.

All fine-tuning is based on these negative Prompts. Without using them, it's impossible to achieve the expected results of the module I created and shared.

No need for VAE

Advantages:

(1) Generates good-looking faces, suitable for turning ACG characters into real-life people, preserving their looks, charm, and clothing details

(2) Highly adaptive and tolerant to ACG Loras (recommended weight: 0.3~0.8, adjust in 0.05 intervals)

(3) Vivid, full colors and dynamic expressions

(4) NSFW Okay!! Visable Body Fluid!!

(5) Sensitive to style-related Loras, roughly 0.3 is enough to change output results, 0.6 will significantly affect the image, increasing playability. Also offers good support for pose and clothing Loras

(6) Can generate ACG decorative elements, such as spoken hearts, emphasis lines, speech bubble, etc.

--> After using the new negative terms, even red eyes and green eyes can appear.

(7) Suitable as a model for creating real-life images from ACG characters

Recommended negative Prompts: (EasyNegative:0.9), (verybadimagenegative_v1.3:0.9), Amplitude: 0.45

-

After two img-to-img adjustments,

-

Disadvantages:

(1) Limited scenes, especially prone to everyday or simple backgrounds, recommend using Loras or Prompts to improve

Solved. By adding "Oekaki negative TI" negative terms, the quality and backgrounds will be even more outstanding.

(2) No fine-tuning for ACG male characters, maybe looks girly

Good-looking Astolfo Face

-

(3) The "masterpiece" keyword is mistakenly bound, easily generating NSFW content

(4) After using the new keywords, the body is too thin and overly flat. I think it would look better if it were more curvy and had some muscle. Alternatively, you can adjust it using prompts or Loras

Currently, no better, similarly purposed models can be merged. This is the completed version until I learn MWB.

Special thanks to:

KanPiroMix

3moon anime_cosplay

3moon REAL_Photo

DDosmix

Lofi V2

Compared to other real-life transformation effects

欢迎大家来评论

在copper模型的基础上使用了mbw插件继续融合

这个模型很容易画出看起来娇小的角色

使用推荐

正面tag建议不要使用medium breast，small breast的效果更好

负面tag推荐使用Easynegative（一个负面pt模型）：https://huggingface.co/embed/EasyNegative/tree/main

例如：nsfw,EasyNegative,nsfw,(low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2),bad composition, inaccurate eyes, extra digit,fewer digits,(extra arms:1.2),2girl,lanjiao5,text

VAE推荐挂载通用VAE(即Animate.vae)，使用vae-ft-mse-840000-ema-pruned这种写实风格的vae也可以，不过偶尔会有颜色过重的情况。

v1.5

相比于1.0变化不大，解决了一些颜色过重的问题，1.5版本的颜色会更浅一些

如果觉得颜色不够鲜艳推荐使用clearvae:ClearVAE | Stable Diffusion Checkpoint | Civitai

同时也改善了一些nsfw场景，但不多（

v2.0

在尽可能保留模型特色的基础上，改善了一些人物细节，增强了背景效果

v3.0

与2.0相比有了一定的变化，但不多，主要是丰富背景

I really like Mix-Pro and Mix-Pro Beta. They're some of my favorite models because of how expertly they capture their style. But I also like landscapes, and Mix-Pro is not really a landscape focused model.

You can see where I'm going with this.

This model is an attempt to strike the middle balance between the expert character model that Mix-Pro is and give it some more ability to work in stunning landscapes besides the excellent character art. It... sort of accomplishes that, but at the cost of being neither the best landscape model, nor the best character model, but splitting the difference. It retains the style of Mix-Pro pretty well but mainly tends to zoom out a bit from the character, which makes them less prominent.

Give it a shot if you're interested. I'll probably keep tinkering with this on and off.

HuggingFace

铭刻，漫步于樱花丛中的时光。这里是来自于galgame《樱之刻》的原画级大模型。

Time of walking among the cherry blossoms. Here is the original painting-level model from galgame Sakuranotoki.

特点是干净又卫生通透，眼眸与头发丝的高光非常细腻。

This model features clean, transparent, very accurate highlights on the eyes and hair.

使用时，以下配置可能比较有用（其他的用法待开发，欢迎尝试和分享）：

When using it, the following configuration may be more useful (other usage methods are to be developed, welcome to try and share):

DPM++ 2M Karras/DPM++ SDE Karras

Hires.fix/BSRGAN

CFG scale=6~9

完美女孩，perfect girl

注意：本模型禁止用于涩涩，任何相关nsfw内容与我本人无关！

Attention：Do not use this model to generate nsfw pictures.And all relevant uses have nothing to do with me

• 插播：

  

  如果想要更好的q版人物图，可以看看这个模型：

  QteaMix 通用Q版模型 - gamma | Stable Diffusion Checkpoint | Civitai

说明 | specification：

• 本模型为幼态特化模型，但并非不能出成女等体型角色，如果达不到想要的效果，请加入“small breast”等提示词。

• It is not impossible to produce adult female and other body types. If the desired effect is not achieved, please add "small breast" and other prompt words.

  

• 本模型在Anything2.1H的基础上进行训练，修改了如“Small chest”、“kawaii”等与模型主题相关提示词的权重。

  This model was adjusted using the full Anything2.1N model to modify the weights of prompts such as "Small chest", "kawaii", etc. related to the model topic.

• 【PS：】请不要添加“masterpiece”等起手式，除非画面出现各种问题

  Please do not add "masterpiece" or other prompt words

• 本模型的VAE没有任何问题，无需使用其他VAE

  There is no problem to VAE of this model, there is no need to use other VAE

• 【两个版本】

  ICE：最推荐下载的版本，适合正常使用

  Merge：适合融合的版本，如果你想要把这个模型融合到其他模型里面，推荐下载这个版本

  之前那个模型因为一些zzcw的原因被干掉了，如有需要的话我再把原来的版本发上来（PS：上传太慢了）

其他说明 | other：

当然了，跟所有的模型都一样，这个说明放在这里只是为了叠甲，该怎么玩怎么玩就是了。

Of course, as with all models, this instruction is just there in case something goes wrong.

• 你可以随意将本模型融合到其他地方，但如果你共享该融合模型，别忘了标注一下

  We allow you to merge with another model, but if you share that merge model, don't forget to add me to the credits.

• 除此之外允许任何人复制和修改模型，但是请遵守CreativeML Open RAIL-这里是M.CreativeML Open RAIL-M相关内容：

  Anyone else is allowed to copy and modify the model, but please comply with CreativeML Open RAIL-M.You can learn more about the CreativeML Open RAIL-M here：

  License - a Hugging Face Space by CompVis

• 模型可以和其他模型一样随意使用，但是请遵守所在地区的法律法规，以免造成麻烦（我们并不负责）

  The model can be used as freely as other models, but please abide by the local laws and regulations, so as not to cause trouble (we are not responsible for it).

特别鸣谢：

• GitHub - bmaltais/kohya_ss：训练UI

• GitHub - XavierXiao/Dreambooth

• Automatic1111: SD-WebUI

• Danbooru：打标器，帮助标准训练集

• 秋叶甜品店：想法！

• QteaMix：通过研究这个模型得到的思路！

祝大家玩得开心＜3

Updates on V2:

• replaced Anything V3 with pikasNewGenerationV1_v10

• replaced HARDBlend with edgeOfRealism_eorV20Fp16NoVae

• Safetensors version uploaded too. It's on the box down to details.

Hi guys!

This is my first SD 1.5 Mix. I trying to give a little bit of comic, anime and semirrealism style.

This is the models contained in this mix:

Comic Diffusion

Complex Lineart

Copeseethemaid

Mtompunk

AnythingV3

HARDBlend

I recomend this VAE for your experiments.

Greetings! :8)

Abstract,

Anime Draw

Paradise,

Woman

My Ckpt

https://civitai.com/models/48668/mix-realistic-fantasy

https://civitai.com/models/48265/dream-real-comic

I don't have a pickletensor (ckpt)

raised safetensors or ckpt

This is a merge of many many realistic checkpoints

一个简单的融合模型，使用了WolfBoys_2D和NijiV5style为主加上一些其他融合模型，并融合了一些成熟男性的lora，推荐使用muscular和facial hair为提示词。

A simple merge model, using WolfBoys_2D and NijiV5style as the main plus some other merge models, and incorporating some mature male lora, with muscular and facial hair recommended as prompt words.

A well trained checkpoint (using JP's Dreambooth implementation) of Jena Malone's character from the cult class "Donnie Darko" where she played Jake Gyllenhaal girlfriend. 12images / 2000 steps. Use "GretchenRoss Person" as a trigger in your prompt.

This is a rather old model of mine, so it might be faulty in some aspects. Most notably, it sometimes produces nonsensical faces and creates textures where they shouldn't be, but you should be fine as long as you use recommended settings.

Sampler: DPM++ 2M Karras, DPM++ SDE Karras

Steps: 20+

CFG Scale: Anywhere between 4 to 11, beyond that I did not test, but I doubt you'd be using values outside of that bracket anyway. Lower scale values seem to be more "creative" at the cost of producing more muted colors. Also, higher CFG scale values tend to produce aforementioned errors more often (for instance, on Euler a with 20 steps and no hires.fix)

Hires.fix: on, upscaler either any of the Latents or R-ESRGAN 4x of your taste. Latent is more creative, R-ESRGAN is less prone to nonsense generations. Use denoise strength in the range of 0.55 to 0.75 for Latent and 0.3 to 0.75 for R-ESRGAN.

Clip Skip: 1 or 2, the effect is similar to CFG Scale with 1 being more creative and 2 having deeper contrast and more popping colors (all example images made with Clip Skip 2).

VAE: the one from the base SD 1.5 model. I didn't bake it in just in case you want to use something else with the model.

A merge of the Fluffyrock AES V2.1 and Fluffyrock YS V2.1 models at 50/50.
Because why not?

Vae ☻ ↓

https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt

https://huggingface.co/AIARTCHAN/aichan_blend/tree/main/vae

https://civitai.com/models/6844/ema-pruned-vae

https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors

This model is a simple merge of 60% Corneo's 7th Heaven Mix and 40% Abyss Orange Mix 3. Full credit goes to their respective creators.

This is perfect for people who like the anime style, but would also like to tap into the advanced lighting and lewdness of AOM3, without struggling with the softer look.

There is also a separate Vtubers version, great for Hololive and to a lesser extent Nijisanji and independent vtubers. This is not the default version as it doesn't play as nicely with Loras (but there are some examples anyway).

Technically more like V10 but its renamed as V3 to keep in line with my previous versions.

This is a merge of many many checkpoints , including the few dreambooth models i created myself

All credit goes to the other creators in helping me create this version.

It can and will create both SFW and NSFW images with the correct prompts

V4.0 is live

Based on my own taste, this model is fine-tuned by training and fine-tuning about more than 100 pictures of my favorite style.

The base model is a base model based on the original sd v1.5.

Made by merging some excellent light and shadow animation models through merge block weighted.

VAE is the same as Anything v3.0

Most of the parameters of the previewed pictures are based on

• size: 512x768

• sampler: DPM++ SDE Karras

• step: 28

• highfix: true

• Latent 1.5x ~ 2x

• highfix step: 10-20

• highfix denoise: 0.6~0.7

About the name:

This is the model used by a chatbot called alu.

四元融合Q版通用模型，具有更少细节与更扁平风格，适应表情素材、立牌、人物Q版化，较为泛用。通常加“chibi”就能简单的达成，甚至不加也可以。推荐原始分辨率不超过960

The following content is from machine translation: the general model of the quaternary fusion chibi version, with less detail and a flatter style, adapted to expression materials, standing cards, and character chibi version, more general. Usually it can be achieved simply by adding "chibi", or even without it. The recommended native resolution does not exceed 960

两个版本没有差别，如果不希望占用太多存储空间，建议使用半精度版。

There is no difference between the two versions, if you do not want to take up too much storage space, it is recommended to use the half-precision version.

你也可以在抱脸上找到这个模型

You can also find this model on the hugging face

chenxluo/QteaMix · Hugging Face

Model is trained on 40 pictures of hare from soviet series "Nu, pogodi!" (main 16 episodes) at 512*512 resolution for 500 epochs.
Examples are NOT cherrypicked, generated each 5 by Euler a, Euler, DPM++ 2M Karras and UniPC respectively.

Trigger word is "waj" due to it is a "rare token", what means it is very unlikely to interfere with the "main model" itself.

Please consider Buying Me A Coffee if you like my contributions to the community: https://www.buymeacoffee.com/Rjox

Differences between versions:

v5 - more flexible and well-rounded, so I recommend it over v6

v6 - has had some custom training done to it, and while it still works well, it was created more for my personal work than for public release. It leans more towards studio backdrops and my custom training, so it's a little less flexible than v5

About RyanBlend:

The RyanBlend models are designed primarily to create photorealistic images of women with a high-end glamour/boudoir photography style, with a very subtle hyper-realistic anime style for facial features and body proportions. It's NSFW friendly, but doesn't have to be used that way.

It's excellent at creating closeup skin texture and high quality faces without using a face restoration model.

I've been a professional photographer since the mid-2000's and specialize in a high-end retouched look for glamour and erotic photography. So I try to make my RyanBlend models close to my own photography and editing style.

The base of the merge consists of:

Deliberate, phDMix, and Realistic Vision v2

...plus a little:

Analog Madness, HARDblend, URPM, and Perfect World

v6 also has 10,000-15,000 training steps of my own custom appearances for faces and bodies.

I recommend using VAE ft mse 840000

I have created my first Checkpoint model that is suitable for mostly realistic styles, but I added some of my own trained cartoon style models for extra flares. I focused on refining the hand and feet issues as much as possible. Based on the images I have generated so far, it seems to be fine most of the time. I avoid using the word "dynamic" as it can cause the hand to go wild.

For the example images, I used the same prompts, and all the displayed images are completely raw without any additional support from Loras or Embeddings. However, using negative embeddings like "bad artist" or "bad hands" can greatly improve the result.

If you have any suggestions for improving my model, please leave them in the comments below. Thank you.

Prompts:

Pos: (European), beach, sea, (hand well), (finger well), (long braided hair), (bangs), with in frame, ultra-detailed, 8k, masterpiece, beautiful detailed face, (volumetric lighting), (beautiful detailed eyes), (ambient light), realistic shadows, (fullbody:1), (1girl), (cowboy shot), pattern shirts, pants, looking at viewer, perfect body, body well:1, leg well:1, arm well:1, dynamic pose

Neg: ((hands poor)), (fingers poor:1), (badquality), (bad anatomy), (nsfw), (legs poor), (inaccurate limb:1.2), (2girls), bad composition, inaccurate eyes, (extra hands:1.2), (inaccurate limb:1.2), (extra arms:1.2), (bad quality), bad face, (bad eyes), deformed, ((deformed fingers:1.2)), ((extra fingers, extra limbs)), bad hands, bad fingers, Asian

Pos: beach, sea, (hand well), (finger well), (long braided hair), (bangs), with in frame, ultra-detailed, 8k, masterpiece, beautiful detailed face, (volumetric lighting), (beautiful detailed eyes), (ambient light), realistic shadows, (fullbody:1), (1girl), (close up shot), pattern shirts, pants, looking at viewer, perfect body, body well:1, leg well:1, arm well:1

Neg: ((hands poor)), (fingers poor:1), (badquality), (bad anatomy), (nsfw), (legs poor), (inaccurate limb:1.2), (2girls), bad composition, inaccurate eyes, (extra hands:1.2), (inaccurate limb:1.2), (extra arms:1.2), (bad quality), bad face, (bad eyes), deformed, ((deformed fingers:1.2)), ((extra fingers, extra limbs)), bad hands, bad fingers

Pos: ((cartoon)), anime style, beach, sea, (hand well), (finger well), with in frame, ultra-detailed, 8k, masterpiece, beautiful detailed face, (volumetric lighting), (beautiful detailed eyes), (ambient light), realistic shadows, (fullbody:1), (1girl), (close up shot), pattern shirts, pants, looking at viewer, perfect body, body well:1, leg well:1, arm well:1, hand pose

Neg: ((hands poor)), (fingers poor:1), (badquality), (bad anatomy), (nsfw), (legs poor), (inaccurate limb:1.2), (2girls), bad composition, inaccurate eyes, (extra hands:1.2), (inaccurate limb:1.2), (extra arms:1.2), (bad quality), bad face, (bad eyes), deformed, ((deformed fingers:1.2)), ((extra fingers, extra limbs)), bad hands, bad fingers

Read the description for more information.

It is highly suggested to inpaint eyes after a generation.

Suggested Prompts:
((high quality, masterpiece:1.4)), 1girl, ((upper body)), side view, oil painting style, oil painting, medieval portrait, portrait, medieval theme, fantasy, beautiful face, pretty face, symmetrical face, symmetrical eyes, symmetry, absurdres, realistic proportions, dynamic pose, high details, intricate, intricate details, realistic eyes, realistic eye proportions, sharp focus, 32k, realistic lighting, extreme details, realistic pupils, realistic proportion eyes, realistic proportions pupils, realistic shadows, rembrandt, rembrandt style

Suggested Negatives:
(worst quality:1.5), (low quality:1.5), (normal quality:1.5), lowres, bad anatomy, bad hands, merged bodyparts, extra limbs, blurry, ugly hands, extra arm, extra hand, split arm, missing finger, extra finger, three fingers, four fingers, six fingers, disfigured, unclear, indistinct, merged fingers, misplaced hand, misplaced foot, signature, watermark, text, artist, distorted, cropped, multiple people, 2girls, 3girls, multiple personas, character sheet, character concept, multiple characters, crop, split screen, animal ears, cat ears, EasyNegative, asian, japanese, chinese, korean, frame, border, painting frame

Textual Inversions Used:

• EasyNegative

You do not need a VAE as it has already been pre-baked.

Suggested settings:

• Sampling Method: Euler A

• Sampling Steps: 25

• Hires. Fix Upscaler: R-ESGAN 4x+

• Hires Steps: 20

• CFG Scale: 6

• Width: 512

• Height: 712

• Denoise Strength: 0.3

If you enjoy my model, consider joining our discord server for more content: discord.gg/dreamlabs

The model merge has many costs besides electricity. With your support, we can continue to develop them.

モデルマージは、電気代以外にも多くのコストがかかります。皆さんの応援があれば、これからも開発を続けることができます。

https://ko-fi.com/breakd

We have gone back to the drawing board and redone the merge to remove the Creator Credit Required license from BreakDro. this is a new model that is finished to the same level as BreakDro.

BreakDroからCreator Credit Requiredライセンスを取り除くために、初心に戻ってマージをやりなおしました。BreakDroと同等に仕上げた新作モデルです。

Future updates will be made to this model.

今後の更新はこのモデルを行っていきます。

中文/English

这是一个混合模型，如果有需要改进和做得不对的地方请指出来（如果你能说出纠正的方法最好，因为我可能不是很清楚）。

这个模型的方向是尽可能忠于提示词,能生成更多的东西(人物是主要的,但不代表其他事物的效果不重要),舒服的光影。

从AOM3:https://huggingface.co/WarriorMama777/OrangeMixs 和Anything:https://civitai.com/models/9409/or-anything-v5 知道了很多模型和学习了很多混合方面的经验,十分感谢。

使用建议:

在脸部不是焦点的画面中(非正面也是同样),脸部可能出现崩坏的情况,使用Hires. fix能够获得更好的面部,使用其他随机种或者工具也是不错的办法。

过高的,用于画面质量(像best quality这些)的正面和负面提示词是不必要的,会使画面趋于一种风格,也可能减少细节,以及无法对提示词做出反应。

你可以尝试细致地描述,使画面更加贴近你的想象。

v2.2即将到来

This is a merge model, so please point out if there is something that needs to be improved and done wrong (If you can say how to correct the best, because I may not be very clear).

The direction of this model is to be as faithful as possible to the prompt words, to be able to generate more things (the character is the main, but does not mean that the effect of other things is not important), comfortable light and shadow.

I know a lot of models and learn a lot of mixing experience from AOM3:https://huggingface.co/WarriorMama777/OrangeMixs and Anything:https://civitai.com/models/9409/or-anything-v5, thank you very much

(To be added)

你可以从版本信息中知道使用了哪些模型/You can know which models are used from about the version

我使用这个VAE/I use this VAE:https://huggingface.co/stabilityai/sd-vae-ft-mse-original

配方在/Recipe in → huggingface:https://huggingface.co/Fre2C/DreaMirror-Mix

Introducing my versatile photorealistic model - the result of a rigorous testing process that blends various models to achieve the desired output. While I cannot recall all of the individual components used in its creation, I am immensely satisfied with the end result. This model incorporates several custom elements, adding an extra layer of uniqueness to its output.

One of the model's key strengths lies in its ability to effectively process textual inversions and LORA, providing accurate and detailed outputs. Additionally, the model requires minimal prompts, making it incredibly user-friendly and accessible.

VAE recommended: sd-vae-ft-mse-original.

I apologize if my English is not fluent, as it is not my native language.

This model was initially intended to produce cute anime characters, but it has also been discovered that it has the capability to produce Softcore NSFW content.

here is a link to access the model on Hugging Face: https://huggingface.co/Kybalico/CalicoMix.

If you like this model, you can support me at https://www.buymeacoffee.com/kybalico.

Recommended parameters:

Sampler: DPM++ SDE Karras
CFG Scale: 5-7
Hires.fix: R-ESRGAN 4x+Anime6b
Negatives: EasyNegative, (worst quality, low quality:1.4), (malformed hands:1.4),(poorly drawn hands:1.4),(mutated fingers, missing fingers:1.4),(extra limbs:1.35),(poorly drawn face:1.4)

CFG Scale: lower is cuter, recommend around 5.0-7.0

For the VAE, you have the freedom to use any model you prefer. As an example, ClearVAE is recommended for producing soft color images, along with Anything and OrangeMix. If you're looking for a bit more vivid colors, consider using vae-ft-mse-840000-ema.

If you need a detailed background, SnailVAE is a good option to consider.

Read Description

Important: Having multiple models uploaded here on civitai has made it difficult for me to respond to each and every comment. One of the ways I plan on addressing this is via the creation of a pdf guide for each and every model (think of it as the models documentation). This will take a while though. So in the meantime, if you have any questions or feedback -

visit my thread of the Unstable Diffusion Discord

VAE Required

Download VAE - https://huggingface.co/stabilityai/sd-vae-ft-mse-original/tree/main

|──sd

|____|──stable-diffusion-webui

|____|____|──models

|____|____|____|──VAE

|____|____|____|____|──<Put your VAE file here>

Cine Diffusion - Fine tuned on cinematic stills and 35mm film.

Note: Description is limited atm. Full PDF documentation is in the works.

Have you heard of the high elves?
If you're attempting to generate an image of a elf and aren't seeing pointy ears:

• Set your CFG to 7+

• Make sure elf is closer towards the beginning of the prompt.

Warning - This model is a bit horny at times. I tried to alleviate this by fine tuning the text-encoder using the class nsfw and sfw. Although this solution is not perfect. More experimentation is needed. I've had most success with adding sfw or nsfw within the first 75 tokens of the prompt. Adding sfw or nsfw to the negative prompt can also help.

Settings

Clipskip : 1

ETA Noise Seed Delta : 31337

CFG : 6 - 8

Sampler recommendations :

• DPM++ SDE Karras, 18-40 steps

• Euler a, 20-32 steps

• DPM++ 2M Karras, 30 - 70 steps

• Heun, 20 -30 steps

• Euler, 20 - 70 steps, Sigma Churn = 1

These parameters are not strictly required, experiment around with other samplers and parameter values. You might find something that works better for you.

Textual Inversion (Negative) used in sample images :

• Bad_prompts_version2 - https://huggingface.co/datasets/Nerfgun3/bad_prompt/blob/main/bad_prompt_version2.pt

  • Install location - stable-diffusion-webui -> embeddings -> <place .pt embedding here>

Check out my other models

• Project AIO - https://civitai.com/models/18428/project-aio

• Eris - https://civitai.com/models/21952/eris

• WonderMix - https://civitai.com/models/15666/wondermix

• Refined - https://civitai.com/models/8392/refined

• Experience - https://civitai.com/models/5952/experience

• Elegance - https://civitai.com/models/5564/elegance

• Clarity - https://civitai.com/models/5062/clarity

• VisionGen - Realism Reborn -https://civitai.com/models/4834/visiongen-realism

LoRA

• Pant Pull Down - https://civitai.com/models/11126/pant-pull-down-lora

Yet another anime model. Produces a different aesthetic to most other anime models.

I'd appreciate any feedback you can give. NAI VAE is baked, but feel free to use an external VAE, though your results may vary. R2.0 preview images use ClearVAE; this is my recommended VAE for R2.0.
I used easynegative for the preview images.

If you like my work, consider supporting me: https://ko-fi.com/mira6

“Through the process of constantly learning, my work has been evolving with it as well. This model is the culmination of all my experience with creating models.”

This started out as a simple idea of going back to the drawing board and re-creating DNW from scratch and seeing what would come from it. With the Initial intention of releasing as 2.5 or 3rd version of DNW.

What came out of my endeavour was so far ahead, and so far removed from any of the original models that I decided it deserved to become its own separate model.

ALL THE MAIN EXAMPLES ARE 1:1, NO LORAS USED, NEGATIVES EMBEDS ONLY!

(The famous people seen in the examples were previously trained into the models and are not done with LORAS. Very little inpainting to no inpainting done in examples & only eyes.)

This model can do pretty much any style fantasy art at a very high level. The main concepts:

✔️Illustrated
✔️concept art
✔️surrealism
✔️oil paint
✔️photorealistic
✔️cinematic

(all of the above list can be used as triggers and weights)

It knows most D&D races and it will produce some interesting results when prompted correctly.

It’s hard to nail this model down with simple description. But I can tell you that I’ve tested it for some time now and it produces results that are as good or equal to some of the absolute best models on this site.

I will add more information in the coming days. For now… Create, Explore, & Enjoy the Adventure 💗

HOW TO USE THIS MODEL EFFECTIVELY

There is no embedded VAE but it was designed to be used with this:

☞ vae-ft-mse-840000-ema-pruned.ckpt

I recommend using the following negatives:

EasyNegative (The all around catch all. Always good in a bind)
badhandv4_anime (absolutely your go to neg for anime hands)
bad-hands-5 (good for the more realistic images but will not work amazingly for anime)
bad-artist \ bad-artist-anime (I’ve had good and bad luck)
bad_prompt \ bad_prompt_version2 (works very good sometimes)

Note: There is a full & a pruned version of this model available. This is a complex merge so even the pruned is large as well.

Here are some optimal example settings for you to start with, but please keep in mind YMMV!:

Steps 30 = Sweet Spot / 15-150
Sampler DPM++ 2M Karras, DPM++ SDE Karras, Euler A, DDIM,
CFG scale 7-9.5 = Sweet Spot / 4.5-15
Size 512x768, 768x512, 640x840, 840x640, 600x896, 896x600, (any size)
Clip skip 1
Hires Fix Enabled
Upscale by 1.5-2
Hires Steps Equal to Half or Exact amount of Steps or higher
Upscaler Remacri Upscaler, Lollypop Upscaler, 4x_NMKD_DigArt_10k|

Duskfall's Osenayan Mix

Recipe Card: https://rentry.co/Osenayan-Mix

1.5 Beta is dedicated to @Luna because we were always inspired by Kenshi. Also to @Meina because Meina mix is a great stabiliser in the merge race!

"OH-SEH-NA-YAHN" - An unknown race beyond space and time.

Also PSST: This weirdly has a VERY Gender Diverse focus on accident, so we tip our hats and wave our flags - because sadly no it can't do programmer socks, but sure its' gender diverse in other ways! - Illustration and Ani mixes both make very diverse things!

NEW: TEST IT OUT ON HUGGING FACE: https://huggingface.co/spaces/Duskfallcrew/Osenayan_Mix (ONLY FOR THE 1.2 STABLE EDITION- HAVENT UPDATED IT YET)

DEMO SPACE FOR ILLUSTRATION BETA: https://huggingface.co/spaces/Duskfallcrew/OsenayanMix-Illustration-BasicAPI
(IF EITHER SPACE QUEUES TOO MUCH PLEASE MAKE A PULL REQUEST OR JOIN THE DISCORD FOR SUPPORT ON GETTING A SPACE RESTART)

BETA V2: https://huggingface.co/Duskfallcrew/my_merged_models/resolve/main/OseaIllus15bv2.safetensors / https://huggingface.co/Duskfallcrew/my_merged_models/resolve/main/Osenayan1-5BetaStable.safetensors

HOW TO SUPPORT US:

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

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

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

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

As for who's fronting when we made this merge AND who wrote the card: The dumbass himself: Matoya'iivi Na'agora - yep, there's a lycoris made of me, wtf did you think we were doing training those JUST to make people boy thirsty? LOL.

MoggleMix is a model that makes delicate and beautiful illustrations.
It makes both male and female characters beautiful.

Do not use this model for commercial purposes

MeinaUnreal objetive is to be able to do anime art with a 2.5d feeling.
( the VAE is already baked in the model )

I created a discord server where you can post images that you generated, discuss prompt and/or ask for help. https://discord.gg/F6yeQtEQ98 <3

If you like one of my models and want to support their updates:
I have a ko-fi and Patreon page where you can support me or buy me a coffee <3 , it will be very much appreciated: https://ko-fi.com/meina and https://www.patreon.com/MeinaMix

MeinaUnreal and the other of Meinas will ALWAYS be FREE.

The images in the samples aren't cherry picked, if i had spent 2h to cherry pick each of the images it would hide how the model really performs and i don't think thats fair.

Recommendations of use:
--------------------------------------------------------------------------------
Enable Quantization in K samplers.

Add ' Realistic ' and/or ' Photorealistic ' as first token of the prompt to have 2.5d feeling in the images.
---------------------------------------------
Recommendations:
Sampler: Euler a: 40 to 60 steps.
Sampler: DPM++ SDE Karras: 30 to 60 steps.
CFG Scale: 7.
Resolutions: 512x768, 512x1024 for Portrait!
Resolutions: 768x512, 1024x512, 1536x512 for Landscape!
Hires.fix: R-ESRGAN 4x , with 10 to 15 steps at 0.1 up to 0.3 denoising.
Clip Skip: 2.
Negatives: ' (worst quality:1.4, low quality), monochrome, zombie, (interlocked fingers), '
---------------------------------------------

--------------------------------------------------------------------------------
This is an Attempt to update MeinaMix V5 into its own branch according to feedback. so in the merge list its MeinaMix V1~5.1, MeinaHentai V2.1.

A just for fun checkpoint trained on 18 interesting Midjourney V5 images shared on their Discord/FB pages recently. Trained for 3k steps using Dreambooth (JP's model) on SD V1.5.

Use "Topnotch Artstyle" in your prompt to trigger the training. Basic prompts like "Topnotch Artstyle, egg", for example, will often produce pretty cool results. I do a lot of these random DB trainings for fun. All trained local w/ a good GPU. If there's interest I'll share some other random models.

融合了多个模型，加入了一些自练的模型，拥有更好的背光效果，以及更强烈的光影质感，风格特征更明显，在暗处的表现惊人，线条上更利落

Combining multiple models and adding some self-trained models, it has a better backlighting effect, as well as a stronger light and shadow texture, and the style characteristics are more obvious, the performance in the dark is amazing, and the lines are more neat

Solstice

Solstice is my attempt to create Manhwa artsyle model. Different from the usual manga art style used in some of the models trained in stable diffusion.

Legal & Risk

In principle, I do not support the use of this model to generate any adult content. When using this model, users must abide by laws and regulations, and do not infringe other people's reputation, privacy or portrait rights.

Changelogs

• v1.0-Areum: First release, "forking" my Eclipse checkpoint, reworked most of it's datasets to Manhwa artstyle.

• v1.0b-Areum: added safetensors format, eveything is same as ckpt format, just in safetensors format

Who find this model perfect

• LoRA/LoCon/LyCORIS user who are looking for a compatible checkpoint

• Creative prompters

• People who needs Manhwa model

Usage

• Almost compatible with all sampling methods available, but Euler a and DPM++ SDE Karras are my recommendation

• Use 20-40 sampling steps, more than 40 is not giving any improvement

• CFG: 4.5-7

• Less complicated, short negative prompts are better

• adding textual inversion like Pure Eros Face or Ulzzang-6500 does enhance generated images

If using Hires:

• Hires step: 10-20 steps

• R-ESRGAN 4x+ Anime6b, if you're prefer Latent upscaler, use Latent (nearest-exact)

• Denoising: >0.5. This model needs a lot of denoising.
  

VAE: https://huggingface.co/stabilityai/sd-vae-ft-mse-original

After download this model, try to share with us about your findings.

Please give star reviews with your generated picture if you like it.

p.s. It's not necessary but If you like my creations please consider to support me to create more projects by donate in here:

https://ko-fi.com/gorillamonsooniii

I'm also accept commisioning for LoRA/Ckpt

The model also available at HuggingFace: https://huggingface.co/kayfahaarukku/AingDiffusion

If you like my model and wanna support me more, you can tip me on Ko-fi. (You don't have to but it will help me a lot).

===

Model comparison

( https://huggingface.co/kayfahaarukku/AingDiffusion/blob/main/comparison.png for high-res version of the comparison )

Note: the comparison result is not cherry-picked, not upscaled, and only using minimal negative prompt (literally just EasyNegative).

===
AingDiffusion (read: Ah-eeng Diffusion) is a merge of a bunch of anime models. This model is capable of generating high-quality anime images.

The word "aing" came from informal Sundanese; it means "I" or "My". The name represents that this model basically produces images that are relevant to my taste.

Guide to generate good images with this model

From my research, this model is compatible with NovelAI's animevae and pastel-mix-vae. I highly recommend pastel-mix-vae because it produces more vibrant color than animevae.

UPDATE [25/04/2023]: I found an alternative VAE that you can use, ClearVAE Variant. It generates clearer images without upscaler like pastel-mix-vae did. Hi-res fix is still a good thing to do, though.

Important: Hi-res fix is a must if you want to generate high-quality and high-resolution images. For the upscaler, I highly recommend SwinIR 4x. All the samples I uploaded are using SwinIR 4x.

I also recommend using EasyNegative for negative prompts.

Set Clip skip to 2, ENSD (eta noise seed delta) to 31337 and eta (noise multiplier) for ancestral samplers to 0.667.

balbala

AngrA

- Model -Checkpoint Type merge

If you feel like supporting my work, please consider visiting my BuyMeaCoffe

Trying balance between photorealistic, realism and flexibility to make a cool image, a semi-realistic model aimed to achieve beautiful realistic scenes.

It works to making general or NSFW images of models of high quality.

Inspiration

----

photorealistic

Anime

Futa

Uberreal

----

Recommend: DPM++2M Karras, Steps: 20+, Hires fix 0.2 - 0.5, CFG 6.5 - 9.

Model is highly stylized and character centric. It does really well with portraits, and it works best when the width is at most equal to the height. Beyond that, characters will start to lose detail if you want both landscapes and characters.

But it has an aesthetic I really enjoy so ¯\_(ツ)_/¯ maybe I'll do more iterations and see if I can get better wide shots. We have the technology.

Model Bias

The model is very stylized because of its ingredients:

• DalcefoOilpainting V1

• BetterPastelMix

• Silicon29Dark

• MossaCannibalis_E8 (Lora)

• Krekk0vV2 (Lora)

Dalcefo'sOilPaintingv1 model was the base and it is a portrait style model. Pastel and Dalcefoil's styles had great synergy, but both have an issue with highly detailed characters at times. So the trouble has been made double.

Mossa is a brilliant artist (that's just science don't get mad at me) whose style mixed really well with the Dalcefoil style. And I included Krekk0v because. Both are pretty lewd artists so the model trends towards risque outfits, but this bias can be controlled as per the example pics.

Final Notes

• Occasionally the model will render blue eyes like spice eyes from dune. This should be regarded as a feature rather than a bug.

• It is called "FajitaMix" because everyone's naming their models after food and Mossa's Aztec stuff is high culture. Me gusta mucho.

This is my first model so please let me know if you have any feedback or simply wish me bodily harm in the reviews. I wish you many healthy hands and a full set of fingers.

My first model merge, 5 models in total, used with supermerge. NO VAE BAKED. I use 84000
Has a heavy pastel focus, prompt (((digital photo:1.2))) to help offset it. Saved as fp16.
(I recommend you use hires fix and tile vae/diffusion)

04-24

This version has changed mixing ratios compared to the previous v2.2 and includes some experimental attempts. As always, extensive testing has not been conducted.

I uploaded the pruned version separately, but it is currently downloading only the full version maybe there is a problem with Civitai. I will solve the problem and upload the pruned version within the 25th

Upload pruned model finished

04-10

I attempted to improve the skin tones and lighting and shadows, which were previously awkward. I have noticed a difference in the image, but I have not had the opportunity to thoroughly test the changes due to my limited skills.

VAE : vae-ft-mse-840000-ema-pruned

Steps: 50, Sampler: DPM++ SDE Karras, CFG scale: 20, Size: 512x768, Model hash: a8907e32c0, Denoising strength: 0.35, Clip skip: 2, Hires upscale: 2, Hires steps: 20, Hires upscaler: R-ESRGAN 4x+, Dynamic thresholding enabled: True, Mimic scale: 10, Threshold percentile: 99.95, Mimic mode: Cosine Up, Mimic scale minimum: 0, CFG mode: Half Cosine Down, CFG scale minimum: 15, Eta: 0.2, DDetailer prompt: "", DDetailer neg prompt: "", DDetailer model a: "bbox\mmdet_anime-face_yolov3.pth [51e1af4a]", DDetailer conf a: 30, DDetailer dilation a: 4, DDetailer offset x a: 0, DDetailer offset y a: 0, DDetailer preprocess b: False, DDetailer bitwise: None, DDetailer model b: "None", DDetailer conf b: 30, DDetailer dilation b: 4, DDetailer offset x b: 0, DDetailer offset y b: 0, DDetailer mask blur: 4, DDetailer denoising: 0.35, DDetailer inpaint full: True, DDetailer inpaint padding: 32, DDetailer cfg: 11

04-01

As always, I didn't conduct many tests. However, I did notice a slight improvement in prompt recognition.

I'm not entirely certain if it's an actual upgrade from the v1 version or not.

Steps: 40

Sampler: DPM++ SDE Karras

CFG scale: 20

HIRES. FIX

R-ESRGAN 4X+

STEP : 20

DENOISING : 0.35

UPSCALE : 2

Dynamic Thresholding 

MIMIC CFG SCALE : 10

PERCENTILE : 99.95 OR 99.9

MIMIC SCHEDULER : COSINE UP

MINIMUM VALUE OF MIMIC : 0

CFG SCHEDULER : HALF COSINE UP

MINIMUM VALUE OF CFG : 0

Ddetailer

bbox\mmdet_anime-face_yolov3.pth

Denoising strength(Inpaint) : 0.35

Mimic CFG Scale : 15 (custom ddetailer setting)

----------------------------------------------------------------------------------

03-17

I have updated the model with the extension "safetensors" upon the request of someone else, while keeping it the same as the existing model.

----------------------------------------------------------------------------------

This is a realistic model that I personally merged.
There is limited information due to a small number of tests conducted.

This is an example of a prompt used

ultra realistic details, sharp focus, detailed skin, (21 years old pretty korean girl:1.2), (kpop idol:1.2), (korean beauty:0.8), (korean mixed),

(large breasts),

Negative prompt: Drawings, abstract art, cartoons, surrealist painting, conceptual drawing, graphics, (low resolution:1.4), (blurry:1.3), (strabismus:1.1),

(worst quality:1.3), (low quality:1.3), (thick thighs:1.2), collage, (makeup:0.8), nsfw, bad proportions, earings, floral print, loli, big eyes, (watermark:1.2), letter, (abs:1.2),

Steps: 40, Sampler: DPM++ SDE Karras, CFG scale: 7.5, Seed: 3203463706, Size: 768x512, Model hash: 1d8d6e1204, Denoising strength: 0.35, Clip skip: 2, Eta: 0.2

HIRES. FIX

R-ESRGAN 4X+

STEP : 20

DENOISING : 0.35

UPSCALE : 2

Dynamic Thresholding 

CFG : 20

MIMIC CFG SCALE : 5

PERCENTILE : 99.95 OR 99.9

MIMIC SCHEDULER : COSINE UP

MINIMUM VALUE OF MIMIC : 0

CFG SCHEDULER : HALF COSINE DOWN

MINIMUM VALUE OF CFG : 10

Merge of Chronos and Blossom Models

this is an altered model only for mario bros from the movie 2023, i tried mario bros lore and it doesn't work to my liking, that's why i created this ckpt model for better result use "Mario_Bross" prompts for better result

credits: https://huggingface.co/DrBob2142/Midnight_Mixes/blob/main/Midnight%20Mixer%20Melt.safetensors

This model was created as a universal anime-oriented drawing model. Both characters and landscapes turn out well. It was created for the visual novel Harley Quinn Trainer.

It works best if you use DPM2 Sampling method

If you are interested, this novel is currently at the creation stage, it will be released on https://boosty.to/harleypro

Visit Civitai-Crystal team on discord! -updates and news, development, beta program

Join RunDiffusion on discord! -Nvidia systems running Auto1111 or InvokeAI

Crystal Clear 2

Now you can generate images as high as 2048x696 with no fix or upscaling required.

Works best for landscapes and panoramic views.

Unleash your creativity by using high CFGs to create new types of art.

Examples in the gallery.

This model is the accumulated result of many testing hours, training, feedback and minute detail improvements, upon the Crystal framework, done by an entire team of people.

__________________________________________________________________________________________________

Try it on RunDiffusion where you can generate your favorite CIVITAI models or train new.

The model was created to produce clean, noiseless and unrivaled image quality, while improving the colors and sharpness.

Features and changes:

• clean and noiseless renders

• ability to produce raw images at high resolutions (no fix or upscaling)

• adjustments to color vibrancy

• adjustments to shadows and exposure

• designed to output images at high CFGs without quality degradation

• lessened the anime style influence of the model (compared to Crystal-Clear 1)

• smaller file size for the full version (compared to Crystal-Clear 1)

• ability to produce a wide range of styles per seed while maintaining consistency

Note: This is not a photorealistic model, it will not produce real looking elements.

There is a Crystal-Photorealistic version in development.

Naked Pokimane

A mix of Deliberate, RealDosMix and Lyriel

Update 1.2:

As announced in version 1, Juggernaut is evolving more towards NSFW. It should now be easier to create images in this direction. However, the model is still very versatile and you can cover different styles with it. Example images are, as always, the raw output. With all the tweaks available to you, you will surely be able to create significantly better images than I ever could :D

I would appreciate a review. And I'm also looking forward to seeing your images using this model :)

By the way, you can follow me on Instagram

Or you can buy me a Coffee/Beer (but it's really not necessary, I would still appreciate it though :D)

Today, I'm presenting to you "Juggernaut," a side project that emerged during my work. In the past, I was always annoyed when I had to switch models to achieve different styles, so I had been working on another model (StaVis) for some time. From that, this side project emerged, which didn't fit my original vision but was still too good to be thrown away.

Juggernaut is an extremely versatile model that can serve several style directions (paintings, digital, realism, anime, NSFW) and delivers absolutely sharp quality. The sample images are the raw output from Stable Diffusion. The standard VAE was used, so you can experiment with your own VAEs.

It started a couple of weeks ago when i started training my own model (with round about 5k images (50 % Hand tagged/50 % AI Tagged). It was still not the model i searched for so I merged a couple of models into this (Dont ask me which models, i simple didnt write it down) . This Model is more "NSFW" as the Original Project (StaVis) and prob will be updated more into nsfw in the future. But in the End you guys decide ;)

The model works very well with weighted prompts but has some difficulties with shorter prompts.

I hope you have as much fun with "Juggernaut" as I do :) I would be happy to see some of your images in the gallery below.

Edit: Added 2GB Pruned Version

It's interesing.

随手融合的一个模型，比例已经忘记了，只是把之前炼制的lora融合到新模型里：

所用的底模为 revAnimated 1.2.2

增加对比度使用了 Ally's mix

替换亚洲脸使用了 BRA

烘焙了 clearVae

最后加上自己早期训练的仙侠服装Lora以及水墨风格Lora,

未使用 BWM, 就成了这么一个模型。

可以使用两个触发词：inkwash style (水墨风格) | xianxia style (仙侠风格)。

现在还有点瑕疵，等有空了再调整....自己有基础的，也可以按照上面的逻辑自己融合，比如自己替换其他风格衣服等....

本人郑重声明：

本模型禁止用于训练基于明星、公众人物肖像的风格模型训练，因为这会带来争议，对AI社区的发展造成不良的负面影响。

machine translation:

A model that was fused casually, the proportions have been forgotten, just integrate the previously refined lora into the new model:

The counter used is revAnimated 1.2.2

Increased contrast using Ally's mix

Replaced Asian face with BRA

baked clearVae

Finally, add Xianxia Lora I trained in the early days and the Ink Wash Lora,

Without using BWM, it becomes such a model.

Two trigger words can be used: inkwash style (ink wash style) | xianxia style (Xianxia style).

There are still some flaws, and I will adjust it when I have time... If you have a foundation, you can also integrate it according to the above logic, such as replacing other styles of clothes by yourself

solemnly declare:

This model is prohibited from being used to train style models based on the portraits of celebrities and public figures, because it will bring controversy and have a negative impact on the development of the AI ​​community.

V4.1 NEW MODEL OUT : Works with a higher resolution, up to 850x850, has a better fur and NSFW content and without VAE.

Please consider supporting me so I can continue to make more models --> https://linktr.ee/bumblebee95_

This model is a mix of various furry models.

It's doing well on generating photorealistic male and female anthro, SFW and NSFW.

I HIGHLY recommend to use Hires Fix to have better results

Below is an example prompt for the v4/v3/v2.

Positive:

anthro (white wolf), male, adult, muscular, veiny muscles, shorts, tail, (realistic fur, detailed fur texture:1.2), detailed background, outside background, photorealistic, hyperrealistic, ultradetailed, professional photography, hdr, dslr, 3d, lifelike

Negative:

I recommend to use bad-artist, boring_e621 and EasyNegative

Settings :

Steps: 30-150

Sampler: DDIM or UniPC

CFG scale: 7-14

Size: from 512x512 to 850x850 (only v4)

Denoising strength: 0.6

Clip skip: 1

Hires upscale: 2

Hires steps: 30-150

Hires upscaler: Latent (nearest)

CONTABY MODEL

FOR V3 ONLY: Suggested to use more than 30 steps, to have a modern, clean quality.

Where are the differences between v1, v2 & v3?

Contaby is an anime (Hassaku-based) model, unlocked for nsfw.

I (PlayaEmblem) don't own the base of the model.

Contaby is trained to have a smoother, warmer light with even more details.

Contaby vs Hassaku (Girl, City, Rainy, Blue Hair)

Contaby vs Hassaku (Girl, Flying, Sunset)

This Model is better for girls than for boys. Try it yourself and download Contaby.

Any questions? DM me on Discord: ! PlayaEmblem#9877

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

❗ March 26th - NOISE OFFSET v2 UPDATE - Darker & Better quality

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

DIABOLIQUE MERGES - Version Preview

❗ ORIGINAL - First / original merge. Semi-realistic to realistic. All the other versions below are its babies. This will stay as it is for reference.

❗ NOISE OFFSET - Merge with noise offset added to it. Recipe is photoreal + fantasy.

❗ PHOTOREAL - Further merged for photo realism.

❗ FANTASY - Strong fantasy prompting capabilities. Can do semi-realistic / digital / anime. Offspring of original merge above.

❗ ANIME - Anime merge is an offspring of the fantasy merge. Aimed towards anime, does digital arts / fantasy as well, decent at it, with right prompts.

❗ MANGA - Manga merge is an offspring of anime merge. Primarily aimed for manga / illustrations, anime and such.

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

ℹ️ PROMPT BOOK v2 (Download the PDF) or CHECK ONLINE (Excluded NSFW from this, separate book with NSFW coming later)

ℹ️ IMGUR ALBUM - Fun experiment for everyone to try. Prompting w lyrics! Feed SD just with lyrics. Here is an example album - images generated with few lines from Iced Earth, metal band.

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

ℹ️ What went into making of all the models

QGO-10b / F222 / BasilticAbyssDream / Wonder / Refined / Clarity / HASDX / RPG / HARDBlend / Hunter 69 / Dreamshaper / Cyber Realistic / Deliberate / Elegance / Elysium / Galaxy 3 / Generic Female / Project AIO / LOFI / Realism Engine / rMad / Woop Photo / A to Zovya / WLOP / NED / Perfect Sketchbook / Dark Sushi / Epic Illustration / iCoMix / Comi Noir / Counterfeit / Meina / ReV Animated / Tri Phaze / EZ (Seems to be gone)

There were a lot of merges in the process and its kind of hard to tell the exact recipe at this point. Big thanks to all the authors of the merges above for their amazing work.

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

ℹ️ Which model should you chose? Quick read on these

Pruned models & their differences

Fp32 / Fp16 & their differences

ema-only weights use less VRAM - suitable for inference

ema / no-ema weights use more VRAM - suitable for fine-tuning

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

My workflow

I run ✔️ hires fix with R-ESRGAN 4+ for photoreal / realistic outputs.

I run ✔️ hires fix withR-ESRGAN 4+ Anime6B for anime / manga outputs.

I run ✔️ hires fix with 4x-UltraSharp at times for various tasks.

CFG scale / steps vary per model / desired output, so does the sampler. Experiment :)

You can find many more upscalers here

I mostly avoid using ❌ restore face.

I rarely do in-painting. Matter of fact, none of the images I posted went through any in-painting.

After up-scaling in SD, I usually upscale my output images further here - it's free, fast and does a decent job - check it out.

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

Useful SD related links

PROMPTING

• Lexica / Instant Art / Prompt Hero / Danbooru / Endless Characters

STYLES & SIMILAR

• AMAZING MAGIC JOURNEY DISCORD SERVER - Preview on reddit

• MJ Styles

• Wildcards

OTHER

• SD Tools

• SD Manual

• SD Reddit

• Image Upscaling

• Multi Frame Rendering for SD

———— ▬▬▬▬ ———— ▬▬▬▬ ———— ▬▬▬▬ ————

Pay attention to the About this version section of model page for specific version information. ➡️➡️➡️➡️➡️

Model Overview:

• rev or revision: The concept of how the model generates images is likely to change as I see fit.

• Animated: The model has the ability to create 2.5D like image generations. This model is a checkpoint merge, meaning it is a product of other models to create a product that derives from the originals.

• Kind of generations:

  • Fantasy

  • Anime

  • semi-realistic

  • decent Landscape

• LoRA friendly

• It works best on these resolution dimensions:

  • 512x512

  • 512x768

  • 768x512

VAE:

• orangemix.vae.pt

• kl-f8-anime2.ckpt

• Blessed2.vae.pt

  
  

Prompting:

• Order matters - words near the front of your prompt are weighted more heavily than the things in the back of your prompt.

• Prompt order - content type > description > style > composition

• This model likes: ((best quality)), ((masterpiece)), (detailed) in beginning of prompt if you want anime-2.5D type

• This model does great on PORTRAITS

Negative Prompt Embeddings:

• EasyNegative

• Deep Negative

• bad_prompt_version2

• bad-artist

• bad-artist-anime

• Make use of weights in negative prompts (i.e (worst quality, low quality:1.4))

Video Features

Olivio Sarikas - Why Is EVERYONE Using This Model?! - Rev Animated for Stable Diffusion / A1111

Olivio Sarikas - ULTRA SHARP Upscale! - Don't miss this Method!!! / A1111 - NEW Model

Disclaimer:

Do not sell this model on any website without permissions from creator (me)

Credit me if you use my model in your own merges

I do not authorize this model to be used on generative services

I have given you plentiful amount information and sources within this section, I will not answer redundant questions if it already exists here in the info section.

if you would like to support me:
https://ko-fi.com/s6yx0

This is a model trained on detailed photos of my own paintings, then mixed back with SD a little bit. Note, that it uses details, so add "highly detailed" or something to get better results (look at my prompts). This model trained on landscapes and several portraits, but makes a good job in other objectives as well. Use "A PAINTING BY MSE" to get a more painterly look

Anime版本 不做对比了 就是更平面和动漫风格 看看例图就好了

-2.0的一些版本提升在于 强化了人物皮肤衣服的透亮感，强化了暗部的细节，并稍微加强了一些手部关节 质感提升更多

1.5的一些对比 质感光影提升 色彩对比很棒

请注意 我的负面词汇并不是单纯的词汇 而是一个个embeddings 你们可以在这个网站下到这些词汇并在负面提示词中用上 可以帮助你们达到我想要的效果

PLEASE READ: this model doesn't do anything but improve anatomy and light. That's why I don't give prompts. You should compare YOUR 1.4 and 1.5 prompts with using this model, so you see if it gives you better results with any prompt. Also don't expect a miracle, but an improvement. This model doesn't do a particular esthetic or style.

I made this model merge for my own use and my patrons. The purpose of this is to improve the anatomy of portraits so I get:

• Better hands

• Less long neck problems

• Better posing

This merge is for general purpose renderings, so it doesn't force a style. Use it for everything with your favorite prompts, SFW or NSFW. In the examples, I am using my prompts and my styles (created by myself at my patreon page).

If you want to support me and encourage me to share future updates, or just like my art, join my patreon page: (and get virtual female models, virtual painters, tutorials, prompts, and more)

https://www.patreon.com/intuitivearts/about

I would also love if you share your results using it here. I love to see what other artists do.
Thank you! :)