After many attempts and taking out many models, I have reached this version with a lot of quality. It still has some flaws such as deformations in the tail but that can be fixed with inpainting. This is the first model that I have uploaded to the page so do not forget to tell me if I have done something wrong or what I should improve. Thanks you.

V2.0

Same as V1.0 just made the last part 50/50 to fix breasts

V1.0

Mixed AOM3a1 with Babes (0.7-0.30) - AB

mixed AB with baka-diffusion (0.5-0.5) - ABB

mixed baka-diffusion with kawaii2d (0.65-0.35) - BK

mixed kotosmix with BK (0.75-0.25) - KBK

mixed ABB with KBK (0.75-0.25) - ABBKBK

ABBKBK is now called silvermix because yes

Digital Fantasy

THIS IS NOT the 1.5 DREAMBOOTH one!

It has NOTHING to do with previous merges, despite it sharing a weirdly similar style

Come hang out with us here: https://discord.gg/Da7s8d3KJ7 - We're running weekly AI art contests, and support other creators in their journey!

Backups of Beta:https://huggingface.co/Duskfallcrew/Digital-Fantasy/resolve/main/DigitalFantasy.safetensors
Backup Beta CKPT: https://huggingface.co/Duskfallcrew/Digital-Fantasy/resolve/main/DigitalFantasy.ckpt
V1 Backup: https://huggingface.co/Duskfallcrew/Digital-Fantasy/resolve/main/DigitalFantasyNewv1.safetensors
V1 CKPT File: https://huggingface.co/Duskfallcrew/Digital-Fantasy/resolve/main/DigitalFantasyNew.ckpt
Diffusers space: https://huggingface.co/spaces/Duskfallcrew/Digital-Fantasy-Anime

(This runs on free tier API, if there's a QUEUE please don't queue more, if it breaks, please either join the discord, duplicate the space on your private profile, OR pull a message on the repo and ask us to reset!)

BUCKET OF JOY:

Meina V9
Something 2.2
Epic Mix Illustration
Duel Comix
Color Boys 2D
Blueboys2d
Deepboys 2d
Duskfall Test (OG)
Duskfall AI
Ironcatfatetoon
Mind Player Floral Boys
Color Box Model
Osenayan Mix

...oh screw it we'll also share the merge screens we capped, because this was a VERY COMPLICATED MERGE (Not lying, we remixed similar ins and outs but didn't use block weight just weighted merge) - It took us MOST of today's twitch stream.

Not all of the merge mixes we made ended up back on our drive because of an error in

the process. We've tried to make sure we screencapped MOST of the process, but at this point we must've been distracted and only did half the screencaps.

"Clam" is named from using modes' name.

It has better performance in txt2img when using positive prompts describing what you want without any quality tags, and you can use EasyNegative as negative prompt.

hf:https://huggingface.co/Johnhex/Clam

How to Use

here is the parameter recommended to use in stable diffusion

VAE: animevae.pt [f458b5c6]

Sampler: DPM++ 2M Karras or DPM++ 2S a Karras

Step: 15~30

Hires Upscale: 2,Latent Denoising:0.5~0.8

CFG Scale: 7~15

Use Models

Sorry that I forgot to record using models' weight.

• Color Box Model [a80fee84]

• Ligne Claire Anime Diffusion V1 [ee5b91da]

• Anything-V3.0-pruned [2700c435]

• tmndMix-tmndMixPlus [65b8d0b2]

• momoko-e [a2a802b2]

基于AnythingV2.1和V5训练 | Training based on AnythingV2.1 and V5

https://civitai.com/models/9409/or-anything-v5v3

更新 | Update log

本模型采用了特殊的训练方法，如果想要调节模型的内容可以使用手术刀来精确控制：Trush︱模型手术刀 。目前我自己正在使用这个模型

This model uses a special training method. If you want to adjust the content of the model, you can use a scalpel to precisely control it: Trush︱模型手术刀 . I am currently using this model myself

[2023/4/17]

• MoonSky7 V1.0

  本次上传是阉割版，项目名称为[Anything 2.1H]，保留有tag更为听话是[Anything 2.1N],由于[2.1N]实际图片效果并不如[2.1H]（出图较差但是适合用于训练），如有需要也可以上传（依旧因为是上传花费的时间太长，大概以我目前的网速需要4个小时才能上传一个模型）

  This upload is castration version, the project name is [Anything 2.1H], keep the tag is more obedient [Anything 2.1N], because [2.1N] the actual picture effect is not as [2.1H] (the picture is poor but suitable for training), you can upload if necessary

例图 | Example diagram

• 例图采用随机批量生成，纯靠抽卡的来的例图我认为是不合理的

  Example diagrams are generated randomly in batches. I think it is unreasonable to rely solely on high-quality example diagrams obtained from extraction

• 本模型项目名称曾被叫做Anything2.1[H],所以生成图所显示模型名称可能会显示有所不同，直接用图片信息生成即可

  The project name of this model has been called Anything2.1[H], so the generated graph model may be displayed differently, and can be generated directly with the image information

• 前段时间我把这个模型当作Anything2.1-RE上传到隔壁去了（补充说明

  I uploaded this model next door as Anything2.1-RE some time ago (note added

说明 | specification

• 模型无需外挂VAE，外挂VAE可能出现颜色过于灰暗的效果。但是模型直接输出图如有出现其他问题，请外挂VAE进行测试（仅屎山tag测试时出现过这种情况，但是日常使用并没有出现）

  VAE does not need to be VAE externally applied to the model, as it may appear to be too dull. But if there are other problems in the direct output drawing of the model, please VAE vae for testing externally (this situation only occurred in the test of shit tags, but not in daily use)

• 部分情况下可能出现过度拟合，如有此类情况请和我进行反馈

  Overfitting may occur in some cases. Please give me feedback if there are such cases

• 本模型同样有着众多的衍生版本，是否分享出来还是看大家的意愿，大家想要我就发出来，不想要就不发了（主要是模型上传时间太长）

  This model also has a large number of derivative versions, whether to share it or not depends on your wishes, you want me to upload, do not want to upload (mainly because the model upload time is too long)

• 不建议继续进行训练，因为模型本来的底子还是Anything2.1/5，想要达到效果预计难度会非常大

  Do not recommend for Fine-tuning.

• 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

• CFG可以推荐设置在3~30之间

  CFG can be set between 3 and 30, AI painting has infinite possibilities

• 个人制作能力有限，模型并不会非常完美，并且可能与人们的期望不同

  My crafting skills are limited, the model will not be perfect, and it may not be what people expect

局限性 | Limitations

• 可能不适合搭配使用例如badhand或者EasyNegative等

  May not be suitable for use together such as badhand or EasyNegative etc

• 部分未经训练的部分可能依旧会出现无法准确反映tag的情况

  Some parts not involved in training still cannot accurately reflect the tag information

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.

• 禁止炼丹阁转载本模型！

  禁止炼丹阁转载本模型！

  禁止炼丹阁转载本模型！

• 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：

  License - a Hugging Face Space by CompVis

  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

• 秋叶的甜品店（频道）：在这里让我有了这个想法，并且帮忙测试模型

祝大家玩得开心＜3

This is my collection of stuff that doesn't usually make it to my Civit page. Whether it's due to quality concerns, or other issues, I never ended up posting these. These blends are experimental, so do not expect the highest quality. My personal favorite out of the three is Periclase.

Included in this collection is:
Checkpoint Blends:
- Orthoclase Blend
- Periclase Blend
- Gabbro Blend

All images made with Orangemix VAE and Hires Fix on.

Additionally, I have a discord. Feel free to join and discuss. I give out advice for making blends and LoRA's.

mcmirella

该模型基于我之前发布的做了一些调整，将第一版CuteYuki与我的其他个人模型进行了一定比例的融合，得到了一些不错的新特性，所以它更像一个分支版本，而非CuteYuli的版本更新。
以下为推荐参数：
Steps: 30, Sampler: DPM++ SDE Karras, CFG scale: 7, Size: 512x768, Denoising strength: 0.5, Clip skip: 2, Hires upscale: 2, Hires steps: 25, Hires upscaler: R-ESRGAN 4x+ Anime6B.

This model is based on some adjustments I made to my previously released work, in which I merged the first version of CuteYuki with some of my other personal models at a certain ratio. This resulted in some nice new features, so it's more like a branch version rather than an update to CuteYuli. 
The recommended parameters are as follows:
Steps: 30, Sampler: DPM++ SDE Karras, CFG scale: 7, Size: 512x768, Denoising strength: 0.5, Clip skip: 2, Hires upscale: 2, Hires steps: 25, Hires upscaler: R-ESRGAN 4x+ Anime6B.

Try it out here! https://mobians.ai/

Join the discord for updates, share generated-images, just want to chat or if you want to contribute to helping perfect the model! https://discord.gg/vDRax8GM

USE:

The model uses Danbooru tags similar to anime models. I would not recommend just generating a character with no follow-up prompts as the results won't look very good. A good negative prompt also helps greatly.

Limitations:

Some characters generate less consistently. This problem will hopefully be resolved in the full release of V3 as I gather more training data.

The purpose of this model is to help you to create realistic images with photographic quality.

You just need to write to write "BTYLER23 person" in order to call the model, try to use camera prompts for better results.

This model is free, please share and give some credit.

这是一个拥有细致线条与强效光感的模型，混合了chosenmix，wintermoon，cetusmix等模型，并分层融合了多个画风lora，最终形成本模型的独特风格

本模型的人体细长，皮肤偏白，容易形成绚烂的画面风格。

注意使用以下负面tag来避免人体结构错误并稳定画面：badhandv4, EasyNegative, verybadimagenegative_v1.3, 3d, sepia, painting, cartoons, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, normal quality, ((monochrome)), ((grayscale))

This is a model with precise lines and strong lighting effects. It combines models such as chosenmix, wintermoon, cetusmix, and layers multiple art styles, such as lora, to create a unique style for this model.

The human body in this model is slender with fair skin, which creates vivid and gorgeous image styles.

To avoid structural errors and stabilize the image, please use the following negative tags: badhandv4, EasyNegative, verybadimagenegative_v1.3, 3d, sepia, painting, cartoons, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, normal quality, ((monochrome)), ((grayscale)).

Legacy Model Merge AnythingWorld

This model is made up of several contributors, preserving for the community.

Prompting Info

Includes an embedded Vae (A3 / NAI I believe)

I tried pruning it - completely changed the results.

• This model really only can do Anime, anything approaching photorealistic or even just detailed art leads to 'melty faces'

• Highly recommended to use '3D' in your negative prompt.

Recommended Negative Prompt:

3d, realistic, bad anatomy, bad proportions, (worst quality, low quality:1.4)

Credits

Model Merger sianworld

• The actual content is a wide array of models, Anything and others.. not sure on the total mix recipe - all information around this mix has been nuked off the internet.

Model presentation

This is a model related to the "Challenge of the WeekEnd" contest on Stable Diffusion discord.

This version is trained on dedicated tokens per user, and doesn't work like you may be used to.

I try to make a model out of all the submissions, for people to continue enjoy the theme after the event, and see a little of their designs in other people's creations. The token stays "SDArt" and I balance the learning on the low side, so that it doesn't just replicate creations.

The pictures were tagged using the token "SDArt", and an arbitrary token given to the user that submitted it.

The dataset is available below and is composed of 23 pictures.

This week's theme

The storm was a maelstrom of darkness and red, a tempest of writhing fury that threatened to swallow everything in its path. Against this chaos stood the ordered landscape, an unwavering bastion of precision and clarity, holding back the furious assault. The battle raged on, each side locked in an unforgiving struggle for dominance.

Amidst Chaos & Order, a figure emerged. A young girl with eyes that glowed like the sun stood tall and unafraid as the battle unfolded. With a calm determination, she held out her hands and began to guide the energies around her. The forces of Chaos & Order clashed and collided, lightning bolts and sharp lines slashing through the air with reckless abandon. But the girl remained unflinching, her movements fluid and precise as she weaved her way through the tumultuous battlefield.

The opposing forces began to merge as swirling clouds of Chaos began to dissipate, softening into curves and patterns that blended seamlessly with the structured landscape of Order. The sky bloomed into a tapestry of colors, a reflection emerging forth from two opposites.

In the end, it was neither Chaos or Order that prevailed, but the harmony that emerged from their interplay. And the young girl, with her eyes shining like stars, knew that this was the true power of Balance.

💢 Two forces, one outcome. Chaos and Order clash in a visually stunning battle for dominance! 🔳

Encapsulate the tension between chaos and order in a dynamic, visually striking composition. The human experience of the push-pull struggle between two powerful forces.

Challenges:

• Use contrasting colors and shapes to create a sense of conflict between two or more opposing forces.

• At least one aspect that represents Chaos and one aspect that represents Order.

• Create a sense of movement and energy that captures the dynamic interplay between Chaos & Order.

Trained tokens

• SDArt

• rcn

• in

• aten

• opi

• omd

• kuro

• nlwx

• asot

• psst

• buka

• muc

• kts

• utm

• avel

• mss

• guin

• pgs

• crit

• mlas

• phol

• dds

• pte

• rean

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

is made for fun (The trigger word can be anything, the main thing about minecraft!!!)

A simple model trained to make handsome anime guys.

I will give the right to use the images generated by my model sales to my friend "聖聖聖也" -> his PIXIV page

Please report to me any other unauthorized use of the images generated by the sale of my model. This is not unconditional, in the new version afterwards, the images generated by my friend "聖聖聖也" will be used as preview images for the new version.

"I would like to express my gratitude to Philip_ovo for helping me publish the model on Furry Diffusion Discord. If possible, please also support the Lora model MJFurryLora And NJFurryLora created by Philip_ovo."

This English version of the description has been modified based on the translation by marbrdf on ChatGPT.

This model is designed to generate kemono-style furry using booru tags and can generate human-like figures without decreasing the quality of the output. The model is designed to produce images without blurring the background.

See the model description for more details.

• Models are named alphabetically based on the order in which they were uploaded, and the later models are not necessarily better. Choose Model A, B, C, D, E, F or G based on your needs. This is similar to orange and 7th, and you can intuitively compare them with reference images to see the differences.

• If you only want to use a short prompt without adjusting the weights and without using the kemono-style Hypernetwork or Lora to generate kemono, choose Model B.

• If you have a basic prompt foundation and can adjust the weights but do not want to use the kemono-style Hypernetwork or Lora to generate kemono, choose Model D.

• If you want a very different style and can accept using the kemono-style Hypernetwork or Lora to generate kemono, choose Model C.

• If you want to generate attractive human-like figures and kemono-style furry that are close to the current mainstream anime style, and can accept using the kemono-style Hypernetwork or Lora to generate kemono, choose Model A.

• If you are looking for a more realistic texture similar to AOM2 or AOM3, and can accept not using the kemono-style Hypernetwork or Lora to generate kemono with a more realistic painting style, then choose Model F.

• If you want clear lines and celluloid-style coloring, and are willing to accept that the characters generated by Model E may appear very young and petite, then choose Model E.

• If you are looking for a highly impasto coloring style and incredibly detailed and accurate scenes with a rich and colorful palette, and are willing to accept that the generated kemono poses in Model G may have a more fleshy feel, such as a more visible navel, then choose Model G. Please note that this model requires the use of kemono-style Hypernetwork or Lora to generate kemono, as otherwise, unnecessary human ears and incorrect animal noses may be generated.

This series of models combines five Lora styles I trained myself with five kemono-style furry Lorastyles in a hierarchical U-net model that I repeatedly fine-tuned through layer merging.

Like Counterfeit and Pastel, this model uses a hierarchical U-net model combined with Lora. However, because it includes two completely different species with different body structures, humans and kemono furry, in the merging process, its stability is not as good as that of a general model.

The models used in the process include but are not limited to the Anything series, Orange series, Pastel+ (Yohan series), Counterfeit series, and 7th series, totaling more than 50 models combined, plus my own personal Lora model, so I cannot share the recipe.

Although this series of models can generate NSFW content, it is recommended to use any Lora related to NSFW for overlaying. The example images do not use Lora related to NSFW, but overlaying any of them will significantly improve the quality.

This series of models was generated with the default HI-RES setting, and I am not sure how much the image quality would be affected without it. However, based on the images generated by others in the discussion area without HI-RES, the results seem to be good.

I will not upload more example images because the style is very diverse, combining more than 50 models, including some realistic style images created by others in the discussion area.

2023/2/24 CrosskemonoC

2023/2/24 Deleted most of the example images, including NSFW, and replaced two images. Following Civitai's rules, example images with minors and NSFW are not included.

2023/2/24 CrosskemonoD

2023/3/1 CrosskemonoF

2023/3/7 CrosskemonoE

2023/3/8 CrosskemonoE_2

2023/3/13 CrosskemonoG

2023/3/14 CrosskemonoG_2

2023/3/23 Crosskemono2.0TEST

2023/3/24 Crosskemono2.0TEST_2

2023/4/1 Crosskemono2.0

2023/4/17 Crosskemono2.5

這個模型是以活用booru tag簡單生成kemono 風格的獸人，同時又能在模型中生成人類卻不降低生成質量，並且保持背景不會模糊的前提下製作的。

詳見型號說明。

• 按照上傳順序對模型以英文字母順序命名型號，並不是越後面上傳的模型就越好，請根據你的需要選擇A型、B型、C型、D型、E型、F型、G型，這有點類似orange以及7th，你可以從參考圖像直觀獲得對比，每個型號都有較大的差異。

• 如果你只想用簡短的prompt，不調整權重，也不使用kemono風格的Hypernetwork或者Lora來簡單生成kemono，那麼請選擇B型號。

• 如果你有一定的prompt基礎，可以接受調整權重，但不使用kemono風格的Hypernetwork或者Lora來生成kemono，那麼可以選擇D型號。

• 如果你想要一個很不一樣的風格，同時又能接受使用kemono風格的Hypernetwork或者Lora來生成kemono，那麼請選擇C型號。

• 如果你想要和目前主流的anime模型接近的畫風和構圖，想生成好看的人類又想生成kemono風格的獸人,同時又能接受使用kemono風格的Hypernetwork或者Lora來生成kemono，那麼可以選擇A型號。

• 如果你想要類似AOM2orAOM3的那樣更加立體的質感，又能接受不使用kemono風格的Hypernetwork或者Lora來生成kemono時畫風較為寫實，那麼請選擇F型號。

• 如果你想要清晰的線條的賽璐璐上色，又能接受使用模型E時，該模型生成的人物會顯得非常年輕和嬌小，那麼請選擇E型號。

• 如果你想要非常厚塗的上色風格，以及非常細緻且正確的場景，非常豐富多姿的色彩塗抹，又能接受在使用模型G時生成kemono的時候生成的姿態會非常具備肉感，也就是肚臍可能會比較明顯，那麼請選擇G型號。(這個型號需要使用kemono風格的Hypernetwork或者Lora來生成kemono，否則可能會生成不必要的人類耳朵和不正確的動物鼻子)。

這個系列模型是將我自己訓練的5個風格lora+5個kemono風格獸lora合併到我使用階層U-net合併的模型內，然後反復階層合併微調的模型。

和 counterfeit 以及 pastel 一樣 使用 階層U-net合併+lora合併的模型，不過因為在合併過程中加入了完全兩極的人類和獸人兩個身體構造不同的物種，穩定性較一般模型要差，望周知。

中間使用過的包含civitai和hugging這2個網站上有名的模型合併

包括不限於

anything series

orange series

pastel +（yohan series）

counterfeit series

7th series

等合計50多個模型合併，加上Lora模型為個人私享，所以不能分享食譜。

這個系列模型雖然可以生成nsfw,但推薦搭配任何跟nsfw有關的lora疊加使用，示例圖像沒有疊加nsfw相關的lora使用，實際上疊加任何一個都有很大的改善。

這個系列模型是在默認開啟 HI-RES 的情況下來生成的模型，我不清楚不使用HI-RES會對生成的圖像質量有多大的影響，但是看討論區其他人在不使用 HI-RES 的情況下生成的圖，感覺還不錯。

我不再上傳更多的示例圖像，因為基於我自己的模型並合併了50多種模型的緣故，風格非常的多樣化，評論區上也有人用來創作寫實風格的圖像，所以請參考評論區其他人創作的圖像。

2023/2/24 CrosskemonoC

2023/2/24 刪除大部分示例圖像，包括NSFW，並替換了兩張圖片，遵循civitai的規定在示例圖像上不放入包含有 minors NSFW 的內容，所有示例圖像的prompt 均不包含 loli or shota 的內容。

2023/2/24 CrosskemonoD

2023/3/1 CrosskemonoF

2023/3/7 CrosskemonoE

2023/3/8 CrosskemonoE_2

2023/3/13 CrosskemonoG

2023/3/14 CrosskemonoG_2

2023/3/23 Crosskemono2.0TEST

2023/3/24 Crosskemono2.0TEST_2

2023/4/1 Crosskemono2.0

2023/4/17 Crosskemono2.5

Yes, yet another nsfw anime blend model.

It is recommended to use an appropriate VAE to avoid color gradients, and the pastel-waifu-diffusion.vae model works especially well.

Anime: v12 / Realistic: v4.1

Hello everyone! These two are merge models of a number of other furry/non furry models, they also have mixed in a lot of my own LoRAs. And they're all about male anthro dragons.

These models are heavily bara oriented, the training dataset is mostly male NSFW, so it's easier for them to generate male nudity, no clothing and NSFW content.

They mainly use e621 tags, but also support danbooru tags, and already baked in vae so you don't have to use any.

Although they are 'doragoon' mix, they can generate other male furries.

They are not specially trained for anthro females. I didn’t try a lot, but I think these models can do females, and they are okay for females.

There are two types of them that are mixed with different models to generate different outputs.

The anime one:

It's used to generate anime style dragons or kemono dragons. It's also good at generating other furry male anthros. Recommended sampler: DPM++ 2M Karras and Euler a.

The realistic one:

It's used to generate realistic or photorealistic style anthro dragon males and other furry males. Recommended sampler: DPM++ SDE Karras and also Euler a.

Example images are slightly curated and they are direct outputs without inpainting.

Known issues:

For both:

• Struggling with wings and tails, because they are anatomically complex. May needs a little inpainting.

  Inpainting is always good.

• Without Hires Fix, both of them tend to generate blurry faces and eyes, so I recommend using Hires Fix for the whole time.

  Latent, ESRGAN_4x or Anime6B.

• Can not do aquatic anthros.

• Getting no better results when using male dragon related LoRAs.

• No eastern dragons support yet.

For the anime one:

• it has noise offset issue, it generates very bright characters in dark scenes, so use ContrastFix to fix it.

  https://civitai.com/models/8765?modelVersionId=10638

  Select SD1.5 version.

• And I don't think it's 'anime' enough for now. Its style is somewhat fixed, and it's a little overfitting.

• Can be unstable.

• Also anime one is worse at hands.

• Can not do ferals.

  (Can do a weird type of ferals: anthro feral xd)

For the realistic one:

• generated characters' eyes will be blurrier in full body portrait.

• Can do just a little bit of ferals.

2023.4.16 GhostMix-V1.1

UPDATE DETAIL(更新说明)

Improve the stability while using GhostMix on charactor, and make the face more like real man. Improve the quality of the model.If you like my model, feel free to upload the generated pictures to share with us. It will help me for the next version of GhostMix, thanks again.

提升了GhostMix在生成人物时的稳定性，使得人脸更逼真。模型质量进一步提升。如果觉得我的模型不错，欢迎上传生成的图片分享，会队我调下一个版本GhostMix有很大的帮助。关于图片复现问题可以看3.11的下文，谢谢。

2023.3.11 GhostMix-V1.0

Introduction(中文简介在下面)

First of all , I want to thank all the people who use this Checkpoint. And this is my first Checkpoint.All the sample image can be reproduced. This Checkpoint works well on both SFW and NSFW.THE NSFW PART IS VERY GOOD!!!
I uploaded this Checkpoint yesterday and from yesterday to today , I am still trying all the possibility of this Checkpoint. So if you try the model and find some good promts , I hope you can upload it and share with me, it will help me for the next version of GhostMix, Thank you agian!

Recommend Some Promts:

1. Fractal Art（highly recommend，awsome）

(masterpiece, top quality, best quality, official art, beautiful and aesthetic:1.2), (1girl:1.3), (fractal art:1.3),

1. Realistic Art-Girl & Flower field

((masterpiece, best quality)), 1girl, flower, solo, dress, holding, sky, cloud, hat, outdoors, bangs, bouquet, rose, expressionless, blush, pink hair, flower field, red flower, pink eyes, white dress, looking at viewer, midium hair, holding flower, small breasts, red rose, holding bouquet, sun hat, white headwear, depth of field,

1. Color Art

(flat color:1.3),(colorful:1.3),(masterpiece:1.2), best quality, masterpiece, original, extremely detailed wallpaper, looking at viewer,1girl,solo,floating colorful water

1. Realistic Art-Nude Girl(NSFW-recommend)

(masterpiece, top quality, best quality, official art, beautiful and aesthetic:1.3),  illustration, beautiful detailed eyes, (1girl), (nude),detailed nipples,huge breasts,puffy nipples, oniomania, erotic,beautiful detailed pussy,soft breasts,upper body

VAE&Textual Inversion:

VAE: kl-f8-anime2 or vae-ft-mse-840000-ema-pruned(anime suggest: kl-f8-anime2)

Textual Inversion: ng_deepnegative_v1_75t, easynegative

About Image Reproduction:

Some user said that they can not reproduce my result.Maybe the setting goes wrong.I just reproduce my cover image. If you want to reproduce my result, Checkpoint Model，Postive Promt，Negative Promt，Textual Inversion，Sampler Steps，Sampler，CFG，Resolution，Seed , ALL the things should be the SAME! Then be careful if you open controlnet or lora ,don't make them influnence your result.

中文简介

首先感谢每一个使用这个Checkpoint的人，这是我融的第一个Checkpoint。所有样图自测都可以复现。SFW和NSFW的图都挺漂亮的，NSFW的图非常棒。这个模型昨天上传到今天，我一直在尝试这个模型的可能性。希望使用这个Checkpoint的朋友们，如果你试了，觉得有不错Promts，欢迎上传图分享，让我也了解一下这个Checkpoint的可能性，也可以帮助我调下一个版本GhostMix，再次感谢。

推荐Promts:

1. 分型艺术（超级推荐，必出好图）

(masterpiece, top quality, best quality, official art, beautiful and aesthetic:1.2), (1girl:1.3), (fractal art:1.3),

1. 真实风格-少女与花海

((masterpiece, best quality)), 1girl, flower, solo, dress, holding, sky, cloud, hat, outdoors, bangs, bouquet, rose, expressionless, blush, pink hair, flower field, red flower, pink eyes, white dress, looking at viewer, midium hair, holding flower, small breasts, red rose, holding bouquet, sun hat, white headwear, depth of field,

1. 色彩艺术

(flat color:1.3),(colorful:1.3),(masterpiece:1.2), best quality, masterpiece, original, extremely detailed wallpaper, looking at viewer,1girl,solo,floating colorful water

1. 裸女(色图-推荐)

(masterpiece, top quality, best quality, official art, beautiful and aesthetic:1.3),  illustration, beautiful detailed eyes, (1girl), (nude),detailed nipples,huge breasts,puffy nipples, oniomania, erotic,beautiful detailed pussy,soft breasts,upper body

VAE&Textual Inversion:

VAE:kl-f8-anime2或者vae-ft-mse-840000-ema-pruned(动画风建议kl-f8-anime2)

Textual Inversion:ng_deepnegative_v1_75t,easynegative

关于图片复现:

有同学好像没法复现我的图，可能是设置有点问题，刚刚我才把头图给复现了。注意几个点：Checkpoint 模型，正向Promt，反向Promt，Textual Inversion，迭代步数，迭代方法，CFG，分辨率，随机种子都必须一模一样！然后复现的话，要把controlnet和lora关掉，怕影响复现。

No baked VAE.
The recommended VAE is "vae-ft-mse-840000-ema-pruned.ckpt".
If you try it and make a good one, I would be happy to have it uploaded here!
The merging source is listed in each version detail.

VAEは焼いてません。
おすすめのVAEは"vae-ft-mse-840000-ema-pruned.ckpt"です。
使ってみて、いいのができたら、ぜひここにアップしてもらえると嬉しいです！

マージ元は各バージョン詳細に記載しています。

AC_H-3是和Alice-Yozakura共同制作,有非常好的背景和光影

Hugging Face: https://huggingface.co/SakuraFoxKira/AC_H-3

参数和tag可以在Hugging Face里获取

As the name i want to point it towards midjourney-like quality without being the same in style, it's still my style. It won't have images and won't be merged with OpenJourney either, I want it to be as diverse as possible and applicable to any simple to complex prompt. That's what I want to aim for in this DucHaitenJourney model.

Please support me by becoming a patron:

patreon.com/duchaitenreal

DPM++ 2S a Karras cfg 11

will be better in large resolution 768x768, 512x512 will be poor quality

negative prompt:

illustration, painting, cartoons, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, bad hands, ((monochrome)), ((grayscale)), collapsed eyeshadow, multiple eyeblows, vaginas in breasts, (cropped), oversaturated, extra limb, missing limbs, deformed hands, long neck, long body, imperfect, (bad hands), signature, watermark, username, artist name, conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body, error

使用了Niji图片，初筛了大约1.3w张用于训练

除了默认的V5风格的画风外还建议通过watercolor和realistic等tag调整画风

推荐采样器Unipic，CGF6以下

this is a model that puts heavy emphasis on artistic and surreal elements with extremely high image detail, extremely flexible, can create beautiful images with simple prompts, negative prompts are not very important. Works best with step count greater than 30 preferably 50 steps, and native resolution up to 768x768, for example 2:3 frame resolution is 704x1024. CFG to around 11 to 15, highres fix will look better but not necessarily if you feel it takes too long

For those of you who don't have a pc or a weak computer, you can consider using my model via sinkin and mage website using the link below:

https://sinkin.ai/m/EYWOblK

mage.space

Please support me by becoming a patron:

patreon.com/duchaitenreal

negative prompt:

(worst quality:1.5), (low quality:1.5), (normal quality:1.5), lowres, bad anatomy, bad hands, vaginas in breasts, ((monochrome)), ((grayscale)), collapsed eyeshadow, multiple eyeblows, (cropped), oversaturated, extra limb, missing limbs, deformed hands, long neck, long body, imperfect, (bad hands), signature, watermark, username, artist name, conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, painting by bad-artist

dsg

It's a merged model that focuses on the person rather than the background


I tried to simplify the line as much as possible


I simplified it and maintained a clear outline


If you don't add 1girl, you can only see the background, so add 1girl plz

(I'm also making models for nsfw, so please wait a little bit :)

YabaL_Mix V2

• 이 버전이 전작보다 손가락의 타율이 좋습니다.

• This model draws fingers slightly better than the previous version.

• 경량화 버전을 추가했습니다 fp16

• I have added a lightweight version with fp16

• 개인 병합 모델 입니다. 

• This is a personal merge model.

• 병합식은 모릅니다!

• I don't know the merging formula!

• 기존 버전에 조금 조금씩 넣었다 뺐다 하면서 혼자 즐기려다가 이쁘게 잘 나와서 공유 합니다. 문제가 될 경우, 말씀 해주시면 삭제 하겠습니다.

• I was adding and subtracting a little bit to the existing version while enjoying it alone, and I'm sharing it because it looks pretty good!

  If there's a problem, please let me know and I'll delete it.

• 가로 이미지는 DDtailer 사용 했습니다.
  세로 이미지는 DDtailer 사용 하지 않았습니다.

• Horizontal images were processed using DDtailer.

  Vertical images were not processed using DDtailer.

• 추천 셋팅 (Recommended settings)

• DPM++ 2M Karras

• Sampling steps : 30

• Upscaler : R-SERGAN 4x+ Anime6B

• Hires steps : 15 to 20

• Denoise : 0.4 to 0.6

• CFG : 06 to 15

• Realistic : 가능합니다. 

• Realistic: Possible.

• 권장 값 : 1.0 to 2.0 

• Recommended value: 1.0 to 2.0

• NSFW: 가능합니다.

• NSFW: Possible.

High-Res-Fix and Dynamic Thresholding

I would appreciate it if you left some feed back, and images.

This model's goal is to generate photo like images.

Inpainting version on HuggingFace

All images where generated using High-Res-Fix with 4x-UltraSharp upscaler.

vae-ft-mse-840000-ema-pruned

Color Boys 2D model is a simple and cute style, suitable for drawing male and female characters together.

It might be a little hard to paint only men.

Ddetailer(Inpaint face) was used in this sample image.

Duskfall's Digital Fantasy

WARNING: This model PRODUCES A LOT OF ARTIFACTS and does not respond well to prompting. It is up to you if you merge or use, let me know if you get it to stop having rabies. This is not going to have any prompt samples OR resource attachments because it's downstream use it really only in merges for colors and style. Links in this and most of the model card are taken from the huggingface space.

https://huggingface.co/Duskfallcrew/duskfall-s-digital-fantasy

This model is a Dreambooth model trained by Duskfallcrew with Hugging Face Dreambooth Training Space with the v1-5 base model

You run your new concept via diffusers Colab Notebook for Inference. Don't forget to use the concept prompts!

All samples and info are here:

https://civitai.com/user/duskfallcrew

If you want to donate towards costs and don't want to subscribe:

https://ko-fi.com/DUSKFALLcrew

If you want to monthly support the EARTH & DUSK media projects and not just AI:

https://www.patreon.com/earthndusk

digidsk1 (use that on your prompt)

这个CKPT主要强调一种明快清爽的画风，比起油画和厚涂类模型，表现阴影的色块面积较为控制，画面整洁干净，非常适合例图中时尚服装的展示。

注意：

特别感谢zakp无私分享的的LoRA模型 fashion girl 使我得到了这样的画面效果，由于这个CKPT已经融合了fashion girl，请不要再在使用过程中再次添加，否则会造成过拟合

PS：这个CKPT是我尝试了不少次的融合配方，得名于乐事薯片，V1.0是黄瓜味（取名障碍的我一直是看见啥取啥的

This CKPT mainly emphasizes a bright and refreshing painting style. Compared to oil painting and thick paint models, it controls the area of the color blocks that represent shadows, making the picture clean and neat. It is very suitable for showcasing fashion clothing in example images.

NOTE:

Special thanks to zakp for sharing the LoRA model "fashion girl" selflessly, which has enabled me to achieve this kind of visual effect. Since this CKPT has already incorporated the LoRA, please do not add it again during use, otherwise it may cause overfitting.

P.S: This CKPT is the result of multiple attempts at fusion, and it's named after the Lay's potato chips flavor, v1.0 "cucumber" v1.1 “lime” (I have a naming disorder and always name things based on what I see : ) ).

Negative Prompt: rMadaNegative4 (SD 1.5) | Stable Diffusion TextualInversion | Civitai

VAE: stabilityai/sd-vae-ft-mse-original at main (huggingface.co)

Optional: rMadArt3_NoiseOffset (SD 1.5 Lora) | Stable Diffusion LORA | Civitai

将训练的动漫模型与几个风格更亮的模型合并，对细节的把握，和色彩的控制更加全面了。实现了还不错的效果。画人物的话推荐使用c站的embedding—easynegative，能实现更好的局部细节。

后续考虑将一些lora合并进去，让动漫光效更加风格化。期待的可以点一点喜欢，也欢迎大家返图和提出建议。共勉！

Combining the trained animation model with several models with brighter styles, the grasp of details and the control of color are more comprehensive. Achieved a pretty good effect. For drawing characters, it is recommended to use the embedding-easynegative of station c, which can achieve better local details.

In the future, I will consider incorporating some lora into it to make the animation light effect more stylized. If you are looking forward to it, you can like it a little bit, and everyone is welcome to return pictures and make suggestions. mutual encouragement!

Suggestion/建议:

Step 1/第1步: txt2img/文生图:

Hires. fix/高分辨率修复: On/开

Upscaler/放大算法: R-ESRGAN 4x+

Hires steps/高分迭代步数: 20-30

Denoising strength/重绘幅度: 0.25-0.35

Step 2/第2步: Send to img2img/>> 图生图:

Resize mode/缩放模式: Just resize/仅调整大小

Denoising strength/重绘幅度: 0.1-0.2

Script/脚本: SD upscale/使用SD放大(Upscaler/放大算法:R-ESRGAN 4x+)

自己网络查找的

Creates a 2D animation style.

We combined our existing ray tracing training version 4 [unpublished], various pictograms from Laura [13], and numerous other unpublished and undistributed personal models.

I used my personal merged model, Royale Anime Hard 1, which became the final version.

It was merged 8 times to the point where it was hard to tell the originals apart, and fused with study material [I don't know the number].

The last individual merged model seems to have helped me a lot (my finger and toe batting average skyrocketed).

Suggested triggers:

( masterpiece:1.2, best quality ), full body, cowboy shot,

and ( masterpiece:1.2, best quality:1.2 ) full body, cowboy shot,

Triggers to use when creating the same image. Add the following items to the Use Styles feature to import it

Ray Tracing, Global Illumination, An Extremely Delicate And Beautiful, Best Shadows, Shallow Depth of Field, Brilliant Colorful Paintings, Dramatic Shadow, Dramatic Lighting, sharp focus, (8k), (4k), extremely detailed, intricate, hyper detailed,

My model works well for raytracing and CG light illustrations.

Steps 28-80 / Recommended Steps 34-45

CFG 10-17 / Recommended 12-16

Denoising strength between 0.35 ~ 0.5.

(For better quality, try increasing the number of stabs to 80, and the CFG to 20).

Purple Eyes -Settings

Detection Detailer - bbox\mmdet_anime-face_yolov3.pth

Padding value (in pixels) when inpainting at full resolution 3

CFG 15

简单来说就是把常用的模型和lora进行了融合（因为调用起来实在太麻烦）于是就现在这个模型，玩的开心，底模是冬月L

Machine translation：

Simply put, it's a fusion of commonly used models and Lora (because it's too cumbersome to call), so let's have a good time with this model now. The bottom model is Winter Moon L

这个模型风格炸裂，远距离脸部需要inpaint以达成最好效果
This model is strongly stylized in creativity, but long-range facial detail require inpainting to achieve the best results.
请参考我的prompt来使用
recommended positive prompts: official art, unity 8k wallpaper, ultra detailed, beautiful and aesthetic, beautiful, masterpiece, best quality, (zentangle, mandala, tangle, entangle)

use ng_deepnegative_v1_75t and badhandv4 in negative

inpaint the face-->only masked-->set to 512x512-->Denoising strength:0.2~0.5

formula: (noosphere_v10Bakedx0.7+dalcefoPainting_2ndx0.3) x 0.6 + 饭特稀V08 x 0.4
我的TG频道可以看到更多例图 https://t.me/majic_NSFW

INM

发布一下最近训练的 SDicon-Dark 1.0 版本，适合生成暗色调B端运营图~欢迎大家拿去玩玩看看效果怎样~

——————

使用声明：

本模型仅供学习研究用途使用。

Fleeting-Radiance Mix

在刹那间绽放的光彩，如同人生中一瞬的璀璨，飞逝而过，却又在心灵深处留下永不磨灭的印记

The fleeting radiance, like a momentary brilliance in the fleetingness of life, passes by in a flash, yet leaves an indelible impression deep in the soul

Introduction：

Chinese:

一个泛用的，比较清爽的二次元模型（自认为）

在融合过程中，我追求细节，结构与稳定

不过，请注意，FR在融合过程中，从未、以后也不会使用到2.5D和真人模型，

这是一个纯粹的二次元模型

我推荐使用最新版本的FRMiX，因为在迭代过程中，我尽力让新版本表现得比先前更好

不过，如果你喜欢先前某一版本的画风，也可以下载对应版本

在设置上，我推荐高分辨率，竖图（如600*960），高清修复（算法我常用4x-UltraSharp和R-ESRGAN 4x+ Anime6B，重绘幅度0.3-0.4）

采样器我推荐 DPM++ SDE Karras，步数30-60（我常用60，细节更到位）

提示词相关性5-7

vae我使用的是clearvae，不过我并没有将它整合入模型，你可以选用你喜欢的vae

如果你需要fp16版本，请告知我（否则我仅会在最终版发布时同时发布一个最终版的fp16版本）

希望大家能够反馈自己生成的图片，这可以帮助我改进FR，谢谢大家！

我的bilibili（欢迎私信）：

https://space.bilibili.com/351739534

English:

My English level is very poor, so I had GPT translate this for me:

A general, relatively refreshing 2D model (at least I think so).

During the fusion process, I pursue details, structure, and stability.

However, please note that FR has never and will never use 2.5D or human models in the fusion process.

This is a purely 2D model.

I recommend using the latest version of FRMiX, because I have tried to make the new version perform better than the previous one during the iteration process.

However, if you like the style of a previous version, you can download the corresponding version.

For settings, I recommend high resolution, portrait orientation (such as 600*960), and high-definition repair (I often use the algorithms 4x-UltraSharp and R-ESRGAN 4x+ Anime6B, with a Denoising of 0.3-0.4).

I recommend using the DPM++ SDE Karras sampler, with 30-60 steps (I often use 60 for more accurate details).

The CFG Scale should be 5-7.

I use clearvae for VAE, but I haven't integrated it into the model. You can use your favorite VAE.

If you need an fp16 version, please let me know (otherwise, I will only release an fp16 version with the final version).

I hope everyone can provide feedback on the images they generate, as this can help me improve FR. Thank you!

FRMix 1.0 version(Attached with seven rendering images)：

ShiratoriShion/Fleeting-radianceMix · Hugging Face

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

Type-2 Celeste is a general-purpose stable diffusion illustrations model.

The example images have been generated using CelesteV2 with a HIRES fix immediately applied to them during generation.

Make sure to use at least a value of 50 or more steps + a HIRES fix for the best results!

Goes without question, you can use this model and any generated images without crediting me or the model though I would greatly appreciate if you could show me your generations <3

The second model that tries to merge

Based on kiwi4 made by kiwi, merge with various models, some components overlap with CursepoisonMix to

The purpose is to make cute girls with big breasts, so it is more NSFW