🧠🧬Agartha 🧪🍄

Welcome to Agartha .......aka ChadStyle!!!

This mix is inspired from the depths of nostalgia and creativity in all of us. This is a true labor of love. I have mixed this so many times I lost count but have never been happy with it until now. This is by far my favorite style that I have mixed. Ever since the first day I stumbled across Jak’s Creepy Critter’s I was hooked on trying to master a merge that would send each creator on a trip that they will would never forget. I am so happy to finally bring this model to the public. This is by far the cream of the crop for me atm. With this model you can create you wildest childhood dreams and nightmares. Goosebumps was a true passion of mine growing up, so recreating the tales in horrifying HD was a must. I also love creating weird drippy things, so of course FOOD_CRIT was added. Model was inspired by my love for Moist Mix V2 and all things eyegasmic I have found on Civit.You can make some truly bizarre and mind blowing creations with this model so have FUN! Truly, Thank you for stopping by, I really look forward to seeing what you create with this magical merge that I put my heart into!!.. ONE LAST THING….You are amazing!

Fore even more value, If you are interested follow me for the latest updates coming up. https://twitter.com/ultraplayerone

1. Lyriel 1.6

2. Babes V2

3. AiNow

4. 3D Magic Space

5. Project CryEngine 

6. Chimeramix v2 

7. Aw Portrait

8. Neurogen 1.1

9. The Great Fault 1.1 

10. Goosebumps Cover Art by Tim Jacobus

11. A-Zoyva Ultra photoreal

12. RoseRenderedMix

13. Animatrix V2

14. 526Mix-Anime+Animated

15. 526 Cusp Of Serenity

16. ColorfulSurrealismAI

17. Artificial Journey

18. Sentinel Diffusion

19. Slimy’s Supermodel V2

20. Lexartdiffuzex

21. Rev Animated 1.2.2

22. Xeno Art style V4.5 (2x)

23. Jak’sCreepyCritters 768px (Food_crit)

24. Color Fusion 

25. Fruit Fusion 

26. Fever Dream 

27. Holographic model (2x)

28. Rainbowpatch 1.2 

29. Eris 

30. JimEidomode

31. DucHaiten-StyleLikeMe 

32. Orusium

33. Darkful

34. NeverEndingDream

35. Dreamshaper V5

36. MoistMix V2

37. StyleJourney

38. DucHaitenDarkside V4

39. Halcyon

40. Artius

41. WyvernMixV8

42. Colorful V3

43. Juggernaut V1.7

44. Odyssey

45. RmadaV8ColdVae

46. Mechamix

*If you want to create more appealing faces for humans I suggest using "(comedone, skin cancer, psoriasis, eczema, herpes)” it in your negative prompt.

  If you are creating creatures or odd things, my suggestion is to leaving off   “(comedone, skin cancer, psoriasis, eczema, herpes)”. (NOTE:This is not my negative, i found it on someone else’s post.) The ai likes to NOT give you goosebump or creepy art if you use that in the negative prompt while focusing on that type of image.

Another trick I have learned is trying things like “(LESS_THAN_GREG_RUTKOWSKI)” in the negative prompt, it makes a DRAMATIC change in your image… like mind blowing sometimes. (NOTE: THIS part is my idea, one day i just said fuck it, im putting alot of this postive stuff in the negative prompt but wording it “LESS_THAN” format and BOOM!!! ) I look forward to seeing what you try in your negative prompt. If you get a pokemon that is deformed for example , try (DEFORMED_POKEMON:1.3) in your negative prompt that worked for me to remove extra pikachu ears and arms. This is experimental stuff so please don’t hold me to it, just remember to have fun.

This model is based on my Hodgepodge model. If you enjoy prompting on Hodgepodge, you will love the result of this model as well.

Do not try to get photorealistic images on this model, it is not for that. This model is for those who want to get high-quality graphics and images in anime style.

If you like the results of my work, feel free to like and stars. The model is free for anyone who wants to use it for personal purposes. I do not and will not take money for my work. Your reviews, stars and likes are the maximum reward I can count on.

分辨率:624x936

Clip skip: 1~3

VAE: None

半身的效果会更好，但我喜欢全身..

========================================================

Resolution: 624x936

Clip skip: 1~3

VAE: None

The half-body effect would be better, but I prefer full-body.

try something new

Negative prompt: (worst quality, low quality:1.4), (bad_prompt_version2:0.8), EasyNegative, badhandv4, text, name, letters, watermark

Steps: 20,

Sampler: DPM++ 2S a Karras,

CFG scale: 7,

Denoising strength: 0.5,

Hires upscale: 2,

Hires steps: 20,

Hires upscaler: R-ESRGAN 4x+ Anime6B

Clip skip: 2,

VAE:kl-f8-anime2.vae.pt

Posts of images welcome.

Please feel free to leave a review about the model.

If you would like to see more consider supporting me you can do it here:

https://www.buymeacoffee.com/goldenwraft

https://www.patreon.com/GoldenWRaft

https://www.deviantart.com/goldenwraft

Used as base models:

Clarity:V2: https://civitai.com/models/5062/clarity

ConsistentFactor:V4.0: https://civitai.com/models/9114/consistent-factor

Colorful:V2.6: https://civitai.com/models/7279/colorful

Do you want to use artists?

https://stablediffusion.fr/artists

• Adding [by Artist, etc] at the end of the prompt can give you the desired result

Nicknamed: C3

Credit to justmaier

Version info can be found under the review section.

I hope you enjoy.

Trained for generating Selena

BBMIX-JOY

It was a merged model for me to enjoy painting alone, but I have many friends asking what model it is, so I'm revealing it. The face is chic, but the body is a model whose weight is adjusted so that it is output in a healthy way.

Her round appearance, big waist and thick thighs are attractive.

referral prompt

• from above, from below

• shiny skin, nice leg line: 1.3, thick thighs, thin waist,

negative

• {1$$EasyNegative|verybadimagenegative_v1.3}

VAE

• clearvae_main , clearvae_main2

• vae-ft-mse-840000-ema-pruned

We will continue to add sample images after finding good Loras so that we can copy and use EXIF in the sample post.

Precautions (Be sure to observe this)

1. Models may not be traded or sold. Because this is free. .

2. Models are not available for online paid services. (Free service available)

3. In using the model, all responsibility lies with the user.

How to high quality

1. You can output an image like the sample by using an extension to increase the quality. (All sample images were used with the extensions below)

• 1. Hires.fix -> SwinIR_4x or R-ESRGAN 4x+ Anime6B

• 2. Dynamic Thresholding (CFG Scale Fix)

• 3. Detection Detailer (must use)

• 4. SD upscaler or https://oo.pe/https://www.upscayl.org/

1.Hires.fix ( built-in ) or

2. Dynamic Thresholding

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

3.Detection Detailer

https://github.com/Bing-su/dddetailer.git

Questions that do not use the above four cannot be answered. Please note that images printed without using the four extensions may not look good.

TIP

In order to keep token 75, do not insert resolution, detail, or quality tokens. Since extension add-ons such as DD or DT handle it anyway, there is no need to insert tokens like the ones below.

bad prompt (do not use)

official art, unity 8k wallpaper, ultra detailed, ultra high res, 8k uhd, film grain, delicate, RAW, light particles, detailed skin texture, detailed armor texture, detailed face, intricate details, ultra detailed

Reality Emphasis Prompt Excludes Detail Emphasis Prompt.

This model is the result of a merging of Counterfeit and Hassaku. The model has a bright and beautiful anime style with high detail.

Using model:

Use Counterfeit 2.5 VAE : https://huggingface.co/gsdf/Counterfeit-V2.5/resolve/main/Counterfeit-V2.5.vae.pt

Textual Inversion: EasyNegative, bad-hads-5, Deep Negative, bad-artist-anime.

For high detalization use Latent upscale with 0.6 denoithing.

BBMIX-RUIS+

It was a merged model for me to enjoy painting alone, but I have many friends asking what model it is, so I'm revealing it. The face is chic, but the body is a model whose weight is adjusted so that it is output in a healthy way.

Compared to the existing RUIS, the face is slimmer and the lower part of the face is reduced. So the small face is attractive.

ex) RUIS+LUCI = RUIS+

referral prompt

• from above, from below

• shiny skin, nice leg line: 1.3, thick thighs, thin waist,

negative

• {1$$EasyNegative|verybadimagenegative_v1.3}

VAE

• clearvae_main , clearvae_main2

• vae-ft-mse-840000-ema-pruned

We will continue to add sample images after finding good Loras so that we can copy and use EXIF in the sample post.

Precautions (Be sure to observe this)

1. Models may not be traded or sold. Because this is free. .

2. Models are not available for online paid services. (Free service available)

3. In using the model, all responsibility lies with the user.

How to high quality

1. You can output an image like the sample by using an extension to increase the quality. (All sample images were used with the extensions below)

• 1. Hires.fix -> SwinIR_4x or R-ESRGAN 4x+ Anime6B

• 2. Dynamic Thresholding (CFG Scale Fix)

• 3. Detection Detailer (must use)

• 4. SD upscaler or https://oo.pe/https://www.upscayl.org/

1.Hires.fix ( built-in ) or

2. Dynamic Thresholding

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

3.Detection Detailer

https://github.com/Bing-su/dddetailer.git

Questions that do not use the above four cannot be answered. Please note that images printed without using the four extensions may not look good.

TIP

In order to keep token 75, do not insert resolution, detail, or quality tokens. Since extension add-ons such as DD or DT handle it anyway, there is no need to insert tokens like the ones below.

bad prompt (do not use)

official art, unity 8k wallpaper, ultra detailed, ultra high res, 8k uhd, film grain, delicate, RAW, light particles, detailed skin texture, detailed armor texture, detailed face, intricate details, ultra detailed

Reality Emphasis Prompt Excludes Detail Emphasis Prompt.

A more anime/digital style model. Currently under testing and will only be available in FP16

Declaration: I suggest illustrators and arts cannot be replaced by AI, although these models can accelerate design/drawing, the details, sprite-inside, visual-logics cannot be Datafication in Neural Networks.

-----------------------------——·--··c🫰c--··—--------------------------------

⚡️⚡️New Modill v2.0 released⚡️⚡️

The train dataset has 500 illustration, 3000 steps.

Train dataset:816*816

CFG scale:7-8

Sampling Steps:20-30

Suggested sampler: Euler, DPM2 a, DPM++ 2S a Karras

Trigger words: editorial illustration

Because the illustrated characters aren’t normal body structure, some multiples arms, legs often occur, pls add Hires.fix to adjust.

Next version will focus on 2.5d and 3d illustration.

All train data collection are human-manual, and training cloud environment costs me much. hope your donation to support my new model. https://ko-fi.com/cecily34

If you have any questions related to the model or its samples, how to get started or just want to share your creations, feel free to join my discord: https://discord.gg/GgZFF3qPEB

That is my pleasure.

🌸 Loves U.

🔥Modill🔥 (Modern-Illustration) is a trained checkpoint to make attractive and creative illustrations/painting. It can generate some flat illustrations for both business and creative-design. Trained by 289 brilliant illustrations.

⭐️Advantages

• Modill is trained by 289 brilliant illustrations from different designers/illustrators. It can draw exaggerated-body characters and raster texture.

• No strictly restrictions on style. The train data includes different styles, no-overfitting can generate more special outputs.

⭐️Recommendations of parameters

• Sampler: DPM2 Karras, 20~40 steps.

• CFG Scale: 7-9.

• Resolutions: 512*512

• Negatives: poorly lit, duplicated leg, no text, one shoe, multiple head, strange face , error head, missing hand, blur, stereopsis, sex, waterpoint

*** Add some Style Lora models might generate great arts. ***

⚡️Appendix for X/Y/Z script

Easy image quality that has a good context awareness!

This model was made with the intent of having great image quality, and good context awareness. It will respond to most types prompting styles, wether it be descriptions, lyrics, or simply tags. Masterpiece is not required, none of the images posted have 'masterpiece, best quality', or anything along that line. The image quality is good by default!

V1.1 Update summary

Much better context awareness, small loss of hand quality.

The continuation of a weird model!

Having used alfahentai for lot random art, it was only being held back by it's own confusion at times. This model is built upon the idea to get rid of that confusion and make a confident model. All the images i posted are only using Nerfgun3's Negative Hand embedding. The rest is pure txt2img prompt. No inpainting or other manipulation was done.

Big shoutout to Cryonicus, I want to believe his model was the missing part my model needed. Make sure to give his model a try too!

What this model is and isn't

Unfortunately, getting the art to improve to an incredible standard, it cost the model it's ability to make proper genitalia. Breasts still work though. Again, the model has been merged as much so that it retains it's general knowledge. It does NOT respond well booru specific tags/prompts. It might have recollections of some things you might expect in a lot of booru tagged models. For example Wariza doesn't function. Masterpiece has essentially been removed as tag and is generally not needed at all. For all the images I create, unless I stole someone's prompt, it will not have masterpiece/best quality ever. Honestly, i've tried comparisons and generally doesn't seem to add anything.

This model shines in it's ability to grasp concepts, context and general cohesion. I've tried various prompting styles on this model and they all seem to work. It works best with natural language and descriptions.

Model Usage

The model works best on a low CFG, i stick at 5. At 6 you already start getting the first symptoms of burn, at 7 it's really bad, and 8 and above simply seem unusable. The model has a lot of detail it can resolve, so it likes high stepcount, like it's predecessor. With UniPC and cfg 3 or 4 you test prompts with steps as low as 7. Take note, that sometimes your prompt will be interpreted differently if you change the aspect ratio. The unpruned image has a bit better image quality in my experience.

Sampler: UniPC

CFG: 5

Steps: 20-50

Clip skip: 1

Upscale: 4x Ultrasharp, denoise 0.5

Negative prompt used

negative_hand, (low quality:1.3), (worst quality:1.3), (lowres:1.3), (error:1.3), (fault:1.3), (undefined:1.3), (crude:1.3), jagged, spiky, ugly hands, extra arm, extra hand, split arm, missing finger, extra finger, three fingers, four fingers, six fingers, (disfigured:1.3), (unclear:1.3), (indistinct:1.3), merged fingers, (bad anatomy,:1.3) misplaced hand, misplaced foot,, cleavage, sex, nsfw, ecchi, slutty, leotard, (text:1.3), (signature:1.3), (title:1.3)

A VAE for generating higher dynamic and more natural image in comparison with the common VAE like 840000, Anything, kl-f8-anime2, etc.

日本語の説明は後半にあります。

This is a 2.5D merged model to realize 'fusion' in various aspects. Because of the complex merging process, I believe you will enjoy the characters and details that only this model can produce.

Specifically, the model has the following 'fusion' features.

• This model is a 'fusion' merge model of both photorealistic and anime-like models, to create a 2.5D model.

• Characters will be produced as the 'fusion' of Asian and Western styles.

• The merging process is complex, and the merged models have unique 'fusion' details that are different from those of other merged models.

The merging process is generally as follows.

First, many photorealistic models are merged. The criteria for selecting models is if they can produce a photorealistic image when outputting a background without people. More than 15 models that meet this criterion are merged, including minor ones. Finally, only chilled remix and epiCRealism have a composition ratio of more than 10%. (A)

Next, a separate animated model is multi-block merged to create a clear, detailed, and descriptive animated model. This is basically the same as the one used in CityEdge StyleMix, and the main models used are: anything v5, Counterfeit v3, hassaku.(B)

Then, the model (B) is multi-block merged with the model (A) to create a 2.5D model. Finally, a thin layer of FaceBombMix, etc. is applied to ensure stability as a 2.5D model.

Both SFW and NSFW outputs are capable. High quality and unique images can be output in a short prompt. VAE is not needed, but you can try some.

List of my models

• CityEdgeMix : Simple photo-realistic model based on Chilled family

• CityEdge_ToonMix : Simple 2.5D anime-like model based on CityEdgeMix

• CityEdge_2dToonMix : Pure 2D anime-like model merging only 2D models

• CityEdge_StyleMix : Anime-like model with detail depiction by making use of block-merge

• CityEdge_FusionMix : 2.5D model focusing on unique depiction through merging many models and block-merge.

さまざまな要素を融合（Fusion）させて作成した、2.5Dマージモデル。非常に複雑なマージプロセスを経ているため、このモデルでしか出せないキャラクターやディテールが楽しめると思います。

具体的には、以下のような特徴を持たせています。

• 全体の描写は、フォトリアル系をベースにアニメ系を「融合」し2.5D化。

• 人物描写は、アジア系をベースに欧米系を「融合」しエキゾチックに。

• 複雑なモデルマージにより他のマージモデルとは異なる独自のディテールを表出。

マージプロセスは概ね以下のようになっています。

最初に、多数のフォトリアル系モデルをマージしています。モデルの選択基準は、人物なしの背景を出力した際にフォトリアルな画像を出せるかどうか。この基準に合致する15以上のモデルを、マイナーなものも含めマージしています。最終的に構成比が10%を超えているものはchilled remixとepiCRealismのみ。(A)

次に、別途アニメ系のモデルを階層マージし、クリアで詳細な描写のあるアニメモデルを作成しています。これはCityEdge StyleMixで利用したものと基本的に同じもので、使用している主なモデルは、Anything v5, Counterfeit v3, hassaku。(B)

そのうえで、(A)のモデルに(B)のモデルを階層マージし2.5D化。最後にFaceBombMix等を薄く重ねて2.5Dモデルとしての安定性を確保しています。

エロと非エロ、どちらにも対応。短いプロンプトで高品位かつ独自性のあるユニークな画像が出力できます。VAEは焼き込み済みですが別のものを使用してもいいと思います。

モデル一覧

• CityEdgeMix : Chilled系をベースにしたシンプルなフォトリアル系マージモデル

• CityEdge_ToonMix : CityEdgeMixを2.5D化したシンプルな2.5Dアニメモデル

• CityEdge_2dToonMix : 2Dモデルだけをマージしたピュア2Dアニメモデル

• CityEdge_StyleMix : 階層マージを駆使してディテール描写にこだわったアニメモデル

• CityEdge_FusionMix : 多数のモデルのマージと階層マージによる独自の描写にこだわった2.5Dモデル

This is a merge model for making cute cartoon faces.

I aimed to avoid the so-called "master-piece face" even with a simple prompt like "1girl, solo."

Embedding used in negative prompt

https://huggingface.co/gemasai/verybadimagenegative_v1.3

https://huggingface.co/yesyeahvh/bad-hands-5

I publish my work on Twitter. Follow me!

https://twitter.com/digitalcatkuro

・可愛いアニメ顔を作るためのマージモデルです。

　「1girl, solo」の単純なpromptでも、

　いわゆる「マスピ顔」が出ないことを目指しました。

Negative promptで使っているEmbedding

https://huggingface.co/gemasai/verybadimagenegative_v1.3

https://huggingface.co/yesyeahvh/bad-hands-5

Twitterで作品を公開しています

https://twitter.com/digitalcatkuro

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

This is a global repository of the Dreambooth Checkpoints.

Some people asked me to upload the source Dreambooths from which I make the LyCORIS.

To not spam the site/profile with multiple models I decided to do it the same way I intended to handle the Game characters -> a single page where each model is just a separate version.

Upcoming: Billie Eilish, Eva Green, Olivia Wilde, Caity Lotz, Alissa White-Gluz

(if there are no more suggestions from the community, I will be picking myself who to upload next :P)

NOTE1: I will be uploading up to one Dreambooth per day (the visibility of other models suffer and I want all to have an equal chance :P, also this gives me more time for LyCORIS)

NOTE2: Feel free to suggest who should be added (you can look at my existing LORA/LyCORIS library to see which Dreambooth I have; but of course a completely new person can be suggested as well - however that would be best to do via my ko-fi page unless you can wait very very long :P)

NOTE3: This is for those who prefer to get the maximum quality! If you are fine with the quality of the LyCORIS, you may want to skip it (however, I imagine I might upload some models here before I convert them into LyCORIS for time reasons).

However, If you prefer LyCORIS - you need not download anything from here because I eventually will upload all those models in the LyCORIS version :)

I'm Bored so I made this mix lmao

so it's a merge of 3 model

MeinaHentaiv3 - Civitai | Share your models

anyhentai - AnyHentai | Stable Diffusion Checkpoint | Civitai

HassakuBlend - Hassaku (hentai model) - V1.2 | Stable Diffusion Checkpoint | Civitai

Primary Model - MeinaHentai

2ndary Model - HassakuBlend

3rdary Model - AnyHentai

Interpolation Method - Weighted sum

Multiplier - 0.3

Baked VAE - grapefruit

NO VAE IS NEEDED, IT'S ALREADY BAKED

if you use VAE please remind to use Clip Skip 2 so it doesn't overdo the VAE effect.

compatible with ANY LORA and ANY EMBEDDINGS EVEN THE HARDCORE ONE (I tried it, turns out really great)

this model also can do multiple stylized arts (not nsfw only)

yea that's all it, send me your creations down bellow and ask anything tbh, I'll keep an update for this model hopefully

large breast beautiful milf

The finetuned HanakoMix Model gives you high quality anime images with a touch of photorealism to generate awesome character pictures

This model excels at outputting portraits of people at a level that could be mistaken for photographs.You can draw realistic figures without using Hires. fix.It is also suitable for NSFW image output.

It is not compatible with most LORAs, as they tend to lose their realism when used.

You can view the work I have created using this and other models on Instagram.

https://www.instagram.com/ken_aiart/

NSFW works are also submitted to patreon. If you like this model, support is always welcome!

https://www.patreon.com/ken_ai_artist

With Stable Diffusion, we can now create AI art generation images using our own trained images. in this model, we can generate images of women wearing dress clothes called in Arabic (dishdasha) or (دشداشة ) it is clothes used in an interior home, wearing clothes for Arabic women ones as popular images.

A 2GB model is on the way

Trained on images of Nintendo Miis.

Checkpoint can be combined with likeness LORAs to make Miis or Miis of oneself.

Keywords: modern_mii, classic_mii, highres, eye_whites, fighting

a couple anime/hentai models mixed with a touch of Dream like diffusion.

A colourful hentai mix also great at sfw content.

2023.5.19

更新了Utopia版本。这个版本基于Fantasy2.0模型微调而来。在减少模型体积同时，图像真实感进一步提升。

Updated Utopia. This version is based on the Fantasy2.0 version, fine-tuned. While reducing the size of the model, the image realism is further improved.

2023.5.17

更新了reality 3.0。这个版本在减少模型体积同时，图像真实感与面容美感进一步提升。

Updated reality 3.0. In this version, while reducing the size of the model, the realism of the image and the beauty of the face are further improved.

2023.4.21

我将胶片风Lora小比例融合进了Fantasy模型中，优化了此模型下皮肤的质感。但请注意，要获得完美的胶片质感，仍需要搭配胶片风Lora使用，只是使用比例可以降低0.1~0.15左右。

Fantasy 2.0模型更偏幻想风格一些，Reality 2.0模型更偏写实风格一些。两个模型质量都不错，大家可以根据需求下载使用。

2023.4.20

更新了reality2.0版本，这个版本相比之前两个版本，较大幅度提升了皮肤和样貌的真实感。建议在生成人像时，搭配胶片感Lora（1girlmix）使用，真实感更佳。

同时最重要的，新模型搭配兔狲Lora使用，生成兔狲逼真图像的成功率大幅提升！（没错，我搞这个模型，其实主要是为了训练猫猫）

chilloutmix模型与本模型的reality版本、fantasy版本的对比：

Comparison between the chilloutmix model and the reality and fantasy versions of this model:

我比较推荐用reality版本的模型，在色调、颜值和真实性上比较平衡。建议将clip skip设为1，这样生成的图会更美观。

I recommend using the reality version of the model, which is more balanced in tone, appearance and authenticity.

本模型由多个checkpoint模型合并而来，具体有哪些我也记不太清了。以下是MoonMix与chilloutMix模型在生成人像时的一些对比。总的来说，MoonMix相比chilloutMix，真实感和锐度有所下降，但美感与丰富度有所提升。

This model is the result of merging several checkpoint models. I can't remember exactly which models there are. Here are some comparisons between MoonMix and chilloutMix models in generating portraits. In general, MoonMix is less realistic and sharper than chilloutMix, but the overall beauty and richness is improved.

演示样例中所使用Lora模型“1girlmix”可以在此处获取：https://civitai.com/models/33208/lora

The Lora model "1girlmix" used in the demo example is available here: https://civitai.com/models/33208/lora

the model have been update , with some loras that have been merged

model information :
this model are merge from

Meinamix, Chikmix and Darksushi

already baked VAE

recommended sampling method:
Sampler DPM++ SDE Karas : 20 steps

CFG Scale: 7

Resolution: 512x768, 768x512, 512x512, and 768x768

Hires.fix : R-ESRGAN 4x+Anime6b, with 7 steps at 0.1 up to 0.5 denoising

Clip Skip: 2

to find negative embeddings :

https://huggingface.co/embed/EasyNegative/tree/main

badhandv4 - AnimeIllustDiffusion - badhandv4 | Stable Diffusion Textual Inversion | Civitai

veryBadImageNegative - veryBadImageNegative_v1.3 | Stable Diffusion Textual Inversion | Civitai

VAE:ClearVAE - Main | Stable Diffusion Checkpoint | Civitai

some loras that already merge with this model :

https://civitai.com/models/25758?modelVersionId=31269

https://civitai.com/models/25769?modelVersionId=37110

https://civitai.com/models/58390/detail-tweaker-lora-lora

I hope everyone is happy with this model, can't wait to see your result!

The current state of representation of people of all cultures in Stable Diffusion is very, very bad. Look at Filipina women for instance. The skin tones, the body shapes, pretty wrong in most models I could find. Using amateur exhibitionist pictures as training data, I set out to start changing that with a filipina model. Come to this one for all your Filipina gonewild needs.

It was based on URPM as a start ckpt. 512x512s in the training data, fit to the frame not cropped. It doesn't handle faces very well because so many are blurred in the training data, but if you know the usual tricks to get coherent faces, it's fine. Captions were done with BLIP. I quite like the results.

Which country/culture to do next?

Still Under Development

Descendant to Something_Amazing

I will be working with this model more.

The progress I've been making this is pretty nice, if there any suggestion guys of what other model's should I merge it with?

rev

A version of a realistic model that I can afford to share. There are plans to refine and reduce its sensitivity. Test, report what you noticed wrong, I will change it in the next versions.

NOTICE:LET ME KNOW before you put this model on commercial usage.

Hard to tell how many models used to merge

Highres-fix(upscaler) is strongly recommended(using the SwinIR_4x,R-ESRGAN 4x+anime6B by myself) in order to not make blurry images.

Looking forward to your reviews!

Show More

已融合vae

可以使用(8k, RAW photo, best quality, masterpiece:1.2), (realistic, photo-realistic:1.37),omertosa,1girl,(Kpop idol), (aegyo sal:1)best quality, ultra high res, (photorealistic:1.4), (1woman:1.5),(best quality)),((illustration))，((beautiful detailed eyes))，((extremely detailed CG unity 8k wallpaper)),((art of light novel)),from side below,fish-eye shot,In the sea of flowers，[dusk:0.6|stars:0.4]，sunset，(1girl:1.5)，blond long hair，gary pupils，wear a wreath，white veil，red roses adorn the skirt，white gauze short cloak，long dress，white silk stockings，{young girl}，rose headdress，sitting，{{masterpiece}}，hang head，Hold a bunch of white rose，lowered eyebrows，small,(1girl:1.5),

(bad_prompt_version2:0.8),(lowres:1.5), (blurry:1.5),(bad hands:1.5), (missing fingers:1.5), (extra digit:1.5),Multiple breasts,(mutated hands and fingers:1.5 ),(long body:1.3),(mutation,poorly drawn:1.2),black-white,bad anatomy,liquid body,liquid tongue,disfigured,malformed, mutated,anatomical nonsense,text font ui,error,malformed hands,long neck,blurred,lowers, lowres,bad anatomy,bad proportions,bad shadow,uncoordinated body,unnatural body,fused breasts,bad breasts,huge breasts,poorly drawn breasts,extra breasts,liquid breasts,heavy breasts,missing breasts,huge haunch,huge thighs,huge calf,bad hands,fused hand,missing hand,disappearing arms,disappearing thigh,disappearing calf,disappearing legs,fused ears, bad ears,poorly drawn ears,extra ears,liquid ears,heavy ears,missing ears,fused animal ears, bad animal ears,poorly drawn animal ears,extra animal ears,liquid animal ears,heavy animal ears,missing animal ears,text,error,missing fingers,missing limb,fused fingers,one hand with more than 5 fingers,one hand with less than 5 fingers,one hand with more than 5 digit,one hand with less than 5 digit,extra digit,fewer digits,fused digit,missing digit,bad digit,liquid digit,black tongue,cropped,watermark,username,blurry,artifacts,signature,3D,3D game,3D game scene,3D character,malformed feet,extra feet,bad feet,poorly drawn feet,fused feet, missing feet,extra shoes,bad shoes,fused shoes,more than two shoes,poorly drawn shoes,bad gloves,poorly drawn gloves,fused gloves,bad cum,poorly drawn cum,fused cum,bad hairs,poorly drawn hairs,fused hairs,big muscles,ugly,bad face,fused face,poorly drawn face, cloned face,big face, long face, bad eyes, fused eyes poorly drawn eyes, extra eyes, malformed limbs, more than 2 nipples, missing nipples, different nipples, fused nipples, bad nipples, poorly drawn nipples, black nipples,colorful nipples, gross proportions. short arm, (missing arms), missing thighs, missing calf, missing legs,mutation, duplicate, morbid, mutilated, poorly drawn hands, more than 1 left hand, more than 1 right hand, deformed, (blurry), disfigured, missing legs, extra arms, extra thighs, more than 2 thighs, extra calf,fused calf, extra legs, bad knee, extra knee, more than 2 legs, bad tails, bad mouth, fused mouth, poorly drawn mouth, bad tongue, tongue within mouth, too long tongue, black tongue, big mouth, cracked mouth, bad mouth, dirty face, dirty teeth, dirty pantie, fused pantie, poorly drawn pantie, fused cloth,poorly drawn cloth, bad pantie, yellow teeth, thick lips, bad cameltoe, colorful cameltoe, bad asshole,poorly drawn asshole, fused asshole, missing asshole, bad anus, bad pussy,bad crotch,bad crotch seam,fused anus, fused pussy, fused anus,fused crotch, poorly drawn crotch, fused seam,poorly drawn anus,poorly drawn pussy,poorly drawn crotch, poorly drawn crotch seam,bad thigh gap,missing thigh gap,fused thigh gap,liquid thigh gap,poorly drawn thigh gap,poorly drawn anus,bad collarbone,fused collarbone,missing collarbone,liquid collarbone,strong girl,obesity,worst quality,low quality,norma quality, liquid tentacles,bad tentacles,poorly drawn tentacles,split tentacles,fused tentacles,missing clit,bad clit,fused clit,colorful clit,black clit,liquid clit,QR code,bar code,censored,safety panties,safety knickers,beard,furry,pony,pubic hair,mosaic,excrement,faeces, shit,futa,testis,EasyNegative, paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), skin spots, acnes, skin blemishes, age spot, glans,extra fingers,fewer fingers,strange fingers,bad hand, (bad_prompt_version2:0.8),illustration, 3d, sepia, painting, cartoons, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, bad hands, normal quality, ((monochrome)), ((grayscale:1.2)), futanari, full-package_futanari, penis_from_girl, newhalf, glans penis, collapsed eyeshadow, multiple eyebrows, vaginas in breasts, pink hair, holes on breasts, fleckles, stretched nipples, gigantic penis, nipples on buttocks, analog, analog photo, anal sex, signature, logocloth, poorly drawn cloth, poorly drawn anus, bad collarbone, fused collarbone, missing collarbone, liquid collarbone, strong girl, obesity,bad body, bad shose,{bad leg}, {{more legs}},poorly drawn asymmetric eyes, short legs, long skirt, indoors, 2girls, only_upper body, only_lower body,{{{duplicate}}},lowers,error,cropped,signature,watermark,username,out of frame,blurred,bad shadow, {from above}, [petite], lowres, bad anatomy, bad hands, text,error, missing fngers,extra digt ,fewer digits,cropped, wort quality ,low quality,normal quality, jpeg artifacts,signature,watermark, username, blurry, bad feets,extra limbs,large breasts,artist name,bad anatomy,bad hands,bad body,bad proportions，worst quality,low quality，optical illusion

Not satisfied with current state of feet models on Stable Diffusion, I set out to train one based on content available that has feet in the frame(s), but not as the primary focus on the photo. This model helps with more photorealistic details on feet and often emphasizes particular positions that are probably well known by foot fetish lovers. Captioned with BLIP.

On its own, it sometimes will impose a stylistic look (gold, vintage pinup type of feel), but merges excellently with other models. Mixing with chillout, Realistic Vision have given me lots of fun results. Still loads of training and input data work I could do so this is just the start.

This model can be used to make a lot of fun images (see examples)

CLIP skip 1
VAE - vae-ft-mse-8400000
https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.ckpt

Short bit about V3 here:
I was noticing that V2.31 and 2.4 were doing some weird stuff outside of my will.
So I decided to remerge the model from the start using rebasin.
It has used the same models minus Cine and it seems to be working more like intended now.
Big thanks to P_I_N for using his hardware to speed up the process.

2023-5-17日常产出

最近打算用这个模型了，目前使用比较舒适

话说训练好lora融进模型进行调整的模型算是训练模型还是融合模型，融lora是否也属于一种训练呢，实在是不想dreambooth了，比较麻烦。ღ( ´･ᴗ･` )比心

Recently I planned to use this model and it is comfortable to use at present

In terms of training lora well, the model blended into the model for adjustment is a training model ora fusion model. Is blending lora also a training model? It is really not dreambooth, which is troublesome

As always, the example images I've posted have NO postwork done to them beyond the facial fix (in some cases) and upscaling. The point is to show how well it works out of the box - if you take the time and create really good prompts and use good negative prompts (an example is at the end of this post) you can get even better results.

This is likely (assuming that there are no glaring issues or problems with this version) the final version of this model. If there are some discoveries I make which provide huge improvements to things, I might reconsider, but as of this writing, I'm very happy and satisfied, so it's time to move on and start training some new things and getting to work on my game engine.

(Updated for V4.5 FINAL VERSION: May 18,2023)
(Updated "Triggers" Sections: May, 15, 2023)

Intro

I wanted a model or workflow process that allowed me to quickly and easily churn out art for games and/or photo stories. I didn't want long complicated prompts to create a certain art style, but I also didn't want to be forced to use LoRa's and TI's for every character. Unfortunately, most of the models where I loved the art styles didn't do so well of loading characters (celebrities for mixing, fictional characters, etc.) and the ones that knew tons of famous characters and people tended not take to the art styles I was going for (at least not without complex prompts).

Triggers:

DIGITAL ART

WARLOCKANDBOOBS
Thickens the outlines a bit while keeping the skin a bit smoother. NOTE: At high weights, this can increase breast size and tend to push subjects toward medieval outfits.

CARTOON

MASTERPIECE

<SIZE> BREASTS - <SIZE> ASS - <SIZE> BELLY
Small - Medium - Large - Gigantic (See below for "Flat Chest" - "Small" does NOT necessarily make breasts smaller.) Just about anything with the word "breasts" in it is going to make them bigger. The more weight you use (and for Large and Gigantic calls) the higher the tendency to make it NSFW by default. (Call your outfits and use SFW in the prompt. NSFW in the negative prompt helps, too.

HYPERBREASTS

HYPERASS

FLAT CHEST
This does a pretty good job of getting the breasts smaller. The weight of this is fairly low in the model so just saying "flat chest" will give very subtle looks. Play with the weights a bit. Depending upon the subject, something like (flat_chest:1.5) give or take 0.1 either way will give them a nearly flat chest. Numbers between 1 and that top point will move along the scale between normal and flat. Usually.

DETAILED EYE
This will give the eyes more detail, but it also tends to drive the camera up to closeup shots - and doesn't really do all that much on anything further back than a cowboy shot. Expect usage of this to change the nature of the shot and either call your camera angle in the prompt or use EasyPose/Controlnet to force the character where you want them.

ADD DETAIL
Based upon the Detail Tweaker LoRa - I baked in just a smidge of this (so there's no real reason or need to use it again with this model). Just put add_detail into the prompt with a weight (higher or lower as desired) to tweak the detail levels.

Fun Triggers

These are not really things I purposely put into the model, but that produce fun and/or interesting results. These triggers may very well work with other models you use, too.

Feminization: Something like "Chuck Norris, feminization," along with whatever else you want in a prompt will create something that would be maybe what Chuck Norris would look like as a woman.

Sanpaku: shrinks the iris of the eye a bit making more of the eye whites show and giving them a bit more detailed look. Increase the weight to make the irises even smaller (and more whites showing).

See "Version Details" to the right for minor revision update notes. This post will be updated with major notes, triggers, and other long-term items.

V4 Update Notes

Rebuilt from the ground up... got rid of some garbage (and pruned it down to half its original size) tweaked some weights that were bothering me, and generally got things cleaned up. I'm not quite happy with the style consistency, yet - but this is a good clean start on things.

NOTE: The trigger words listed are just the ones that were specifically added on top of the SD Base Model - and really only the ones I've found useful to make good tweaks and adjustments. This model is great at knowing art styles, people, and concepts that you might love (and/or hate) from the base model.

I also forgot one ingredient when adding hyperfusion - so it tends to go porny/nude when you start playing with breasts. I won't have time to update/fix until this weekend so... expect to have to do some work to add clothes if you're playing with breast sizes. Sorry about that.

V3.X Update Notes

Keywords "Masterpiece" and "add_detail" will enhance (or back down if negative levels are used).

This also responds to the Hyperfusion triggers (see version 3.2 notes) as well as all the custom art styles from Babes. (Plus many many more - but the Babes trained styles are ones likely not to be found in many other models).

A bonus (from Hyperfusion) is that it also lets you accomplish making breasts smaller - something very difficult to do on most models without bringing in a lot of extra help. With this "flat_chest:1.4" (or thereabouts) will do a nice job of shrinking things down in most cases. You can also use all the triggers from hyperfusion to make the breasts, butt, and belly do all sorts of things - and putting in weights there brings varied results too.

NOTE: At low levels, most of those calls don't go crazy by sexing everyone up. The higher your weights, the more likely that is to happen, so you may need to declare SFWs and even call a specific outfit in some cases. (I had to do that in the big bust sample image).

All previous triggers and keywords will still work, too.

Some notes on the model:

• It "mostly" defaults to SFW out of the box. i.e. If you just type in a person's name or character description with no clothes - it'll tend to put them in clothes - or at least a swimsuit. That said... with proper prompting, does nudes and porny stuff just fine, too.

• I didn't care so much about making sure resulting characters look "exactly" like the person you're calling out so much as wanting to make sure that if I called out that name/person again - they'd be consistent in the way they look from image to image. So, in other words, it's not so much about "accuracy" as it is about "consistency" - and that seems to come through. (This works well for my needs to create stories and whatnot)
  

Good Negative Prompt for this Model
canvas frame, 3d, ((disfigured)), ((bad art)), ((deformed)),((extra limbs)),((b&w)), blurry, (((duplicate))), ((morbid)), ((mutilated)), [out of frame], extra fingers, mutated hands, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), ((ugly)), blurry, ((bad anatomy)), (((bad proportions))), ((extra limbs)), cloned face, (((disfigured))), out of frame, ugly, extra limbs, (bad anatomy), gross proportions, (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), mutated hands, (fused fingers), (too many fingers), (((long neck))), ugly, tiling, poorly drawn hands, poorly drawn feet, poorly drawn face, out of frame, mutation, mutated, extra limbs, extra legs, extra arms, disfigured, deformed, cross-eye, body out of frame, blurry, bad art, bad anatomy, Scribbles, Low quality, Low rated, Mediocre, 3D rendering, Screenshot, Software, UI, watermark, signature, distorted pupils, distorted eyes, (distorted face), Unnatural anatomy, strange anatomy, things on face, ((tattoo)), ((tattoos)),

Obviously, some of those may want to be removed if you're going for something that has been excluded.

这一版的画风相比T-comic更平面，更日漫化，而且细腻的多！算是一次不错的尝试吧！和美漫特化的T-comic可以区分开了。

The style of this version is more flat, more diffuse, and much more detailed than the T-comic! It's a good try! It can be distinguished from the T-comic of the COMIC。

基于animefull-latest [e6e8e1fc]模型训练，为了取得更好的效果与之前炼的CyberMyself融了一点。

应该很听话。

Based on the animefull-latest [e6e8e1fc] model training, to get better results with the previously refined CyberMyself melted a little.

Aesthetics-focused anime finetunes. Waifu Diffusion 1.5-based, inherits that license.

Experimental variants can be found here.

Images generated with the Replicant2 version may be overly bright; if you find this to be the case, consider using this extension.

Rev2-Badprompt by gsdf

Welcome to my Mix! There is Real & Anime Version, be sure to check them both!

--

If you love my work you can buy me a coffee : ko-fi.com/tauron you dont have to, but if you do, it helps me stay awake until midnight (which is the only time im able to work on these stuff lol) and cope with life..

--

Updated to v2.1 Change Log

• Improved realism quality in general, it moves away from 2.5D feel from the previous version while maintaining fantasy capabilities.

• Improved vanilla NSFW prompts

• Improved SDE sampler quality - no more weirdly lit faces when using SDE sampler in ver. 1 especially in dark scene (observed faster baking time, or maybe its me getting more patient)

• Lower problems with Dynamic CFG threshold when using my lycoris in this version

• hands still sometimes messed up at higher CFG - observed with multiple test scene with controlled seed with 3 comparison model vs. my model. with : 1. no neg-Ti, 2. single neg-Ti, or 3. combined neg-Ti. (Ti in this case doesn't really do shit, it did its job reducing the problems - so its mostly problem with my model, i suspect maybe its because i merged with a SD21 model with hentai model, but who knows.)

• the VAE 84000 should be already baked in

• Do note that your results may vary than my testing. just pray to the god of Ai often.

--

How do I come Up with this Mix?

Well, I'm not an Ai engineer, so i dont really understand fully the inner workings of a base model, more than what Stable-Diffusion has put in their github page. I'm basically a designer that has high enthusiast towards Image generative Ai. So, this mix is done through trial and errors - basically I have a set of checklists that is mostly compiled after using SD since i joined CivitAi, the mix should be able to do all these checklists that i put for testing before i further improve or add another mix in.

What if the resulting mix fails one or more checklists? i rolled back to previous mix and tries to change mix values, or even change base model for mixing, thats why its going back and forth processes until I find one that is suitable for most use cases for me personally. There are 1 iterations somewhere in this model where i tried to combine the base model SD1.5 with a SD2.1 models, I dont know if this can cause problems in the future but so far, everything is fine to me.

Did you prune the model? - I'm not sure, definitely im not using .py script to prune the model, but there are pruned model used as a base mix. chilloutmix pruned is one of them (very-very early iterations).

Why do you put so many embeddings or loras to create the sample images? - these are a long processes since the beginning when i used SD, for embeddings, at first i just copied people's post negative prompt to learn, but as i understand more, i made my own, and that stabilizes over time, as you encounters a problem, you add negative prompts to counter that problem until it dissapears, and so on - finally it came to the point where you see in the sample image. these are not random as it may seem, and it is the result of controlled testing. (while maybe a very crude and a dumb-dumb way of brute force manual testing)

as for lora: all the loras used in the sample images are my lora. it is meant to show how the model interact with my lora, which is the point of the mix in the first place, only embeddings i use are other people's work, and credits due to all of them and you can find them in the resources section below.

Why make this Mix?

Short answer is to better integrate with my own lora.

I really love fantasticmix & henmix model, both are very great in creating the style i really like, and especially realism for henmix, be sure to check their models as well. I've been using their model alot and i love it, but i encountered these:

1. Since I'm starting doing my own lycoris, problems starts to arise like overbaking, distorted hands, multiple limbs, even with negative prompts and negative Ti already in place. especially with NSFW prompt, important research anatomies of human bodies gets really bad.

2. Low Sizes! another reason is that Im quickly filling up my disk drive, with so much models that i ended up not using or having to test, moreover 8GB models is a turn off for me, so I decided to just really make one that suits my taste & style and mix nicely with my own locon.

3. I like to have versatility in my model, i often switch between models because i want to go for conceptual art with realistic subject, other time i want to go for complete realism, this model tries to achieve that in a single model, while maybe its not perfect.

4. I like concise prompt, and not having to put too many enhancer prompts such as: photorealistic:1.4, ultra-hires, hi-res, hyperrealistic, masterpiece, 4k, 8k, rawphoto, cinematic lighting, absurdres , etc, etc, i just want to write realistic, and i want it to get there, so i can focus more on prompting the subject / environment / action instead of focusing too much on that part.

5. I use alot of 2:3 ratio instead of 3:2, so i need the model that is kind of optimized for this ratio.

6. If you ever watch kimi no nawa or suzune, it kinds of describe the anime style that i like. a very detailed environment, with a crisp outline. the anime version of this model achieved that with NSFW capabilities.

7. All in all basically i just want less headache and focus on creating beautiful stuff.

Prerequisite & Resources

These are the resources i use for my standard workflow :

1. VAE Prereq (not required for v2.1)

   vae-ft-mse-840000-ema-pruned.ckpt · stabilityai/sd-vae-ft-mse-original at main (huggingface.co)

2. Dynamic CFG Thresholding

   GitHub - mcmonkeyprojects/sd-dynamic-thresholding: Dynamic Thresholding (CFG Scale Fix) for SD Auto WebUI

3. DDetailer

   GitHub - dustysys/ddetailer

4. ADetailer

   GitHub - Bing-su/adetailer: Auto detecting, masking and inpainting with detection model.

5. Break & Prompt Matrix

   プロンプトの基礎と、BREAKってなに？｜BD｜pixivFANBOX

   Features · AUTOMATIC1111/stable-diffusion-webui Wiki · GitHub

6. Negative Embeddings

   Deep Negative V1.x - V1 75T | Stable Diffusion Textual Inversion | Civitai

   bad_prompt Negative Embedding - bad_prompt_version2 | Stable Diffusion Textual Inversion | Civitai

   badhandv4 - AnimeIllustDiffusion - badhandv4 | Stable Diffusion Textual Inversion | Civitai

   EasyNegative - EasyNegative | Stable Diffusion Textual Inversion | Civitai

   bad_pictures - v3.0 | Stable Diffusion Textual Inversion | Civitai

   yesyeahvh/bad-hands-5 · Hugging Face

7. Facial Enhancer Embeddings

   yesyeahvh/ulzzang-6500 · Hugging Face

   Pure Eros Face - PureErosFace_v1 | Stable Diffusion Textual Inversion | Civitai

Credits to all creators above that helps the model achieve its sample images quality above.

To prompt this mix

1. Style —> realistic , realistic details, detailed for realism / anime for anime / masterpiece , best illustration for conceptual art.

2. Subject —> describe your subject can be anything. 1girl, solo, cute, beautiful, slender, etc. if you use my character lora this should be where you put trigger words of the character

3. Attire —> describe the outfit of your subject. If you use my attire lora this is where you should put trigger words

4. Background / Context —> describe the context or the scene

5. Textual Inversion & LoRa brackets

6. You can follow your own prompting method, this guide only serves as a convention when you are having trouble producing similar results with my lora.

How to maximize this mix?

Sampler & Embeddings

Using 2M Karras is good enough most of the time to create good images with less baking time, you can further improve the detail with SDE Karras, in version 1, this messes up the face lighting quite often (such as lit face even though the context is dark), in version 2 this is fixed for most cases (at least in my test).

Using negative embeddings is a must to create sample image quality. I don't really care the inner workings of my model and i don't want to find out further, all i care is that the model can produce what i want at the quality i desire, and can tackle most scenarios that i need.

CFG Scale

• Lower CFG Scale means AI gets more creative and less overbaking problems. the down side is that the results tends to get a plasticky view.

• Higher CFG Scale means results get more realistic, and very accurate with training sets and more prone to overbaking problems.

• Dynamic CFG Threshold ( only use when necessary ) -- using this method allows you to bump up quality and achieve photorealistic easily, so you basically get both Lower CFG & Higher CFG Advantage and eliminate both problems, the downside is that sometimes it doesn't work with specific lycoris for some reasons. The settings that i most often use: Mimic CFG 5-10 , Clamp 99.5 Percentile, 2nd weight slider = 0 - Cosine Up, third weight slider = 15 Half Cosine Down (sorry i kinda forgot the name)

Repeats & Hi Res Fix

• Higher repeats means your Ai has more time to generate the image, resulting in better result. generally 50 is enough, with 1/2 value for high res fix repeat.

• I use 512 x 768, with ERSGAN_4x+ at 2x upscale. which then i will send to extras with upscaler to further 2x again with ERSGAN_4x or 4xUltraSharp (for anime). the settings i find superior are : GPFGAN @0.25, Codeformer @0.3 & weight 0.15

Utilizing ADetailer & DDetailer

• I use ADetailer to auto fix hands & enhance detail of the body. But sometimes this can go against you especially on anime version.

• I use DDetailer to fix face, but if you find it hard to install DDetailer, it may be better to use ADetailer second pass to fix face instead of body. it yields similar results as DDetailer.

The rule of thumb is that there are no constant equation, you need to find one your own that suits what you are aiming, and sometimes you change these value on the fly depending on your goal. its better to know what these parameters do to your results rather than copying values from people's post.

--

As an afterwords, i hope you enjoy using this model, a like or a review definitely helps, or if you find it distasteful, you don't have to use it. there are many models out there that produces similar or superior results.

Donate for civilians in Ukraine, Syria, Mali etc.: Red Cross | Doctors without Borders | DirectRelief | Feed The Children

🔶 Description

"What, another one of these?". Yes, I made it because everything I've tried had a minor flaw (nothing crazy, sometimes artefacting, sometimes face didn't look right, sometimes didn't abide by prompt). So I gave it a shot too. This is a doll aesthetic model with my finetuning for face and proper nudity. Pleasant semirealistic/2.5D style and very accurate to the prompt. Does not have clothing artefacts when prompted for nudity, upscaling artefacts or bias towards childish features.

🔶 Rule

Don't generate child porn/loli (this model doesn't do photorealism and does not churn out children by default unlike some other models but I don't want to see illustrated child porn either). Find free, anonymous help with pedophilic tendendies before a child is harmed.

✨ Guide for best images

(Update: Alternative try just High Res Fix with these settings, this was used for the example images of v1.1: Denoising strength: 0.55, Version: v1.2.1, Hires upscale: 1.8, Hires upscaler: Latent (nearest-exact))

Step 1 - txt2img

Step 2 - img2img with UltimateSDUpscale extension

Credits for parts of model

V08_V08, breastinclassBetter_v13, fashionGirl_v36Fix, koreandolllikenessV20

Credits for my LoRa training:

I did it with this colab for training and this one for dataset cleaning. Source images are diverse so this model does not mimick a single individual.

Pruning Info

The model is pruned with the following settings using this repo from 4GB to 2GB at virtually no difference in the generations.

python prune.py --fp16 --ema --no-depth delicate_v10.ckpt delicate-v1-0-full-ema-pruned.ckpt

Simply download the script prune.py, create a new conda environment (conda create -n prune), then activate it with conda activate prune. Run pip install numpy safetensors torch!=1.13.0. Then run the command above and replace the first checkpoint name with your non-pruned one and the second checkpoint name with whatever you want your pruned model to be called. Don't use safetensors as the extension (use ckpt instead), otherwise it may fail during inference. You can convert pickle ckpt to safetensor in the Checkpoint Merging tab of Automatic1111.

The model took 15 attempts or so probably and the LoRa in total 20 or so. I think it paid of that I didn't rush it though and hope you enjoy it too :)

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

(Use of VAE files is required

I'm using Counterfeit-V2.5.vae, but it's okay to use the VAE you used before)

Recommend vae : Anything-V3.0.vae

use word "mecha" to summon mecha

Negative prompt: (worst quality, low quality:1.4), (bad_prompt_version2:0.8), EasyNegative, badhandv4, text, name, letters, watermark
Steps: 20,

Sampler: DPM++ 2S a Karras,

CFG scale: 7,

Denoising strength: 0.5,

Hires upscale: 2,

Hires steps: 20,

Hires upscaler: R-ESRGAN 4x+ Anime6B

Clip skip: 2

kl-f8-anime2.vae.pt

This is an animated model that mixes a variety of things together. It is easy to use when using LoRA for animated models because it can output without any peculiarities.

I basically use this model for the LoRAs I create.

I would appreciate it if you could follow me on Twitter.

SeekYou - Mikan is up!

Yet Another fluffy Anime Model

Good at a variety of illustrations

It has some pretty deep potential 🙂

⬇Read the info below to get the same quality images🙏

////////////////////////////////////////////////////////////////////////////////////

Recommended Settings - VAE is included

• Clip skip : 2

• Hires. fix : R-ESRGAN 4x+ Anime6B / Upscale by 1.5+ / Hires steps 14 / Denoising strength: 0.4

• Adetailer : face_yolov8n

• Sampler : DPM++ 2M Karras / DPM++ SDE Karras / UniPC

• CFG : 8 Steps : 25

• Prompts

  • (best quality, masterpiece)

  • Neg : (worst quality, low quality)

////////////////////////////////////////////////////////////////////////////////////

Why is my image different from yours?

• Use prompts that match the danbooru tag

  • Tags like Realistic, HDR, etc. have the side effect of forcing the model's characteristics to change

  • Tag auto complete extension

• Use Neg only when necessary

• Refrain from excessive negative embedding

• Use Adetailer for eyes

• Use similar settings

////////////////////////////////////////////////////////////////////////////////////

Do you like my work? check out my profile and see what else!

And A cup of coffee would be nice! 😉

////////////////////////////////////////////////////////////////////////////////////

For anything other than general personal use, please be sure to contact me

You are solely responsible for any legal liability resulting from unethical use of this model

Overview

TLDR: We are releasing Vodka_V2 by FollowFox.AI, a general-purpose model fine-tuned on an updated dataset - now from Midjourney V5.1. And as usual, in this post, we will share all the details how we got there. What you should expect from the mode:

• We used an objectively better dataset - 2.5x larger, which was cleaned better.

• The resulting model is quite similar to V1 but marginally better. It’s a step up but not a breakthrough-type improvement.

• In the current state, we can generate some cool images with some effort

• The model is still far from effortlessly and consistently generating MidJourney or even top SD models level output

You can read all the details about the model training process on followfox.ai (link to the post), as we can embrace the open-source nature of this community. You can recreate the process, see exactly how we got here, and provide feedback and suggestions on individual aspects of the protocol.

Parameters and Workflow that Works Well for Vodka V2

There is a lot more to test here, but we will share a few observations:

• Compared to V1, you can try a wider range of CFG values; anything from 3 to 7.5 can generate good output

• Booru tag-only prompts do not work well since we didn’t tag the dataset

• Human sentence-type description followed by adjectives and “magic words” works quite well

• Almost all samplers seem to generate interesting results.

• SD upscale workflow (outlined below) with tile ControlNet enhances the image quality of this model

• Using EasyNegative TI (link) is recommended. “blurry” in negative prompts also helps.

Upscale Workflow to Try in Automatic1111

After generating the initial image you like in the txt2img tab (we recommend doing a grid of different samplers and CFG values for each prompt to find the promising ones), send it to img2img.

• Use the same prompt and sampler as in the original generation

• Set sample steps high; in our case, we used 150 for most of the images

• Set width and height to 2x the original. So 512x512 should have 1024x1024

• Set the denoising strength to something low; we used 0.2 to 0.25.

• For the CFG value, we used the (original - 0.5) formula. So if the original image was generated at 7.0, we would set it to 6.5.

• ControlNet settings: enable it; for preprocessor select “tile_resample,” and for model ‘control_v11f1e_sd15_tile’. You can also switch to the “ControlNet is more important” option. No need to adjust any other settings.

  

• Make sure to have the “Ultimate SD upscale” extension installed. Select it from the Script dropdown, select the 4x-UltraSharp upscaler, and set tile width and height to 640x640.

  

• Press generate, wait a bit, and you should have a decent output. You can repeat the process to go even higher resolution.

Conclusions and Next Steps

We believe the model development is going in the right direction, and we will continue releasing the new versions. And, of course, we will document and release every step of that journey.

For the V3 release, we already have a working hypothesis of where the blurriness and lack of details in some of the generations might be coming from, and we will try to deal with that.

Patreonも始めました。 よろしければマージモデルリリースの活動を支援して下さい！
I've started a Patreon. If you'd like, please support my efforts for the release of the merge model!

patreon.com/user?u=93245253

アニメ風を意識したマージモデルになります。

AgelessV2をリリースしました。

今回はV1で当初から懸念点であったLatent系使用時に出る線の粗さ、アニメ調を意識しすぎて少なかった描き込み量を強化・改善をしています。

おかげでLatent系使用時でも耐えれる良いモデルとなったと思います。

今まで通りSwinIRも当然ご使用いただけます。

※作例は全てt2iのみでCN等は不使用です。

We have released AgelessV2. This time, we have enhanced and improved the roughness of the lines that appear when using the Latent series, which was a concern from the beginning with V1, and the amount of drawing that was too conscious of the anime style and was lacking. Thanks to these improvements, I believe we have created a good model that can withstand even when using the Latent series. As always, you can of course use SwinIR. *All examples are made with t2i only, and CN etc. are not used.

setting:

• AgelessV2 :
  - 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: SwinIR_4x or Latent (nearest)

  - VAE: kl-f8-anime2 or clearvae_main or mse840000_klf8anime_klf8anime2
  - NP: EasyNegative

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.

融合了一些自己喜欢的模型

Mixed models that I liked

Model capable of creating low tech biomechanical characters, armors and golems from medieval fantasy scenarios

Trained and merge , use clip 2, > 768 *768 is fine

lose some art style, next version will fix this.

Caption by LLM model , can recognize different tags.

Use simple negative prompt. You can try different settings.

====================

Feedback is important to me ,Thanks for your likes and follow.

You can try it on ai-creator.net and ai-creator.net/arts with tags,based on different model or styles.

中文

我个人比较喜欢照片风格的图片，就擅自用@4044大佬的majicMix_V3和@LEOSAM大佬的FilmGirl_V0.2和一些乱七八糟已经下架找不到了的lora融进了这个大模型里面。

如果原作者介意的话可以评论留言，我看到会立刻下架。

建议设置:

• VAE: VAE -ft-mse- 840000ma -pruned

• 采样器:DPM++ 2M alt Karras(建议质量最好，您可以尝试其他采样器)

• 采样步数:30到50

• Clip skip:1或2

• Upscaler : 不建议修复，因为高清后的降噪已经失去了照片的感觉

• CFG Scale : 7~15

• 横纵比:640*960或512*768

免责声明

SFW和NSFW图像的创建由用户决定，用户对是否生成NSFW内容有完全的控制权。

English

Personally, I prefer photo style pictures, so I took the use majicMix_V3 of @4044 and FilmGirl_V0.2 of @LEOSAM and some messy lora that has been removed from the shelves and can not be found into this large model.

If the original author mind you can comment on the message, I see it will be removed immediately.

Recommended settings:

• VAE: vae-ft-mse-840000-ema-pruned

• Sampler: DPM++ 2M alt Karras (Recommended for best quality, you may try other samplers)

• Steps: 30 to 50

• Clip skip: 1 or 2

• Upscaler: hires.fix is not recommended, because the noise reduction after HD has lost its photo feel

• CFG Scale : 7~15

• Width*height: 640*960 or 512*768

Disclaimer

Creation of SFW and NSFW images is user's decision, user has complete control over to generate NSFW content whether or not.

Suggested negative prompts for photorealistic: (semi-realistic anime illustration 3d rendering drawing)

Beauty in the ordinary.

There are tons of checkpoints to achieve a "pretty" or "model-like" face.

It's time for a checkpoint that makes the ordinary and common, like a snapshot taken in passing, feel more real.

That's why ANINDE-mix was born.

It's not just a checkpoint to draw pretty people,

but about creating naturalistic realism with fewer prompts.

AND also very friendly for NSFW content.

With v.GLO, you get realistic Western people,

v.ASI to create realistic (not so ugly) Asians.

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

欢迎使用DepthStyle模型 - 这是一个深度图风格的模型

=================

Welcome to the DepthStyle model - this is a depth map style model

ChillRealHard by BADMONK

hardblend + chilloutmix + secretsauce

Increasingly complex mix I made for use in my own works. Version 1.3 is a bit more stylized, and may not be for everyone, but I like it, and it can still give the high level of detail I appreciate from the model so far. The recipe details are published below, and the result is a model that is still stylized, but intricately detailed and able to represent cultures from around the world. I've noticed in doing characters that have darker skin, I don't have to fight stable diffusion anywhere near as much as I used to. It's very versatile and finally feels like its more then the sum of its parts. I use it for concept art, and I feel it fills that role perfectly. Also yes, the model can do NSFW.

Source Models:
MeinaMixV9
Aurora
NeverEnding Dream
GeminiX_Mix
Utopia

Recipe info:
1.0: 55% MeinaMixV9 : 45% Aurora

1.1: MothMix v1.0 + 0.3 NeverEnding Dream

1.2 (unpublished): MothMix v1.1 + 0.15 GeminiX_Mix

1.21: MothMix v1.2 + 0.15 MeinaMixV9

1.3: MothMix v1.21 + 0.15 Utopia

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.). If you find you're still having issues with characters with dark skin, look into DuskFallCrew's Dark Fantasy Lora, its made for creating dark skinned characters in anime models. 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. After testing, high CFG + DT didn't show much improvement over low CFG (~7), so YMMV. 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 its composite models, 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 or 4x_fatal_Anime_500000_G, with 15 steps at ~0.3 denoising.
Clip Skip: 1 or 2.
Negatives: ' (worst quality:2, low quality:2), (zombie, sketch, interlocked fingers, comic), '
--------------------------------------------------------------------------------

License

This model 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.

Checkpoint Type: Horror | Model based on: YamerModel Gen2

Complete download: Click Here (8,49Gb) - Pruned download: Click Here(3,59Gb)

Recommended use: Potraits of a singular individual in a horror environment.

This model specializes in generating captivating horror images across a range of genres. While its optimal output size is 576x896, it also possesses the ability to produce non-terrifying visuals, such as "an adorable bunny in a magical forest". However, it's important to note that the primary purpose of this model lies in generating terrifyingly impressive images, serving as a testament to the remarkable results achieved solely through the use of the base model without any additional modifications.

Here are some modifications that were tested with this model:

• LoRA: Magonia, Biomechanicals, Pixel Soarting and TentacleHorrorAI
  

• Hypernetwork: Venom Hypernetwork, Pain and Anime Bondage

• Sampler: Euler Ancestral, DPM2 Ancestral

• GFPGAN: GFPGANv1.4

• ESRGAN: RealESRGAN_x4plus

• Stable Diffusion version used: Easy Diffusion v2.5.36

Creator Note: "Kindly remember to rate the model and show your support by leaving a like if you had a positive experience using it. Your feedback and appreciation are greatly valued." <3

Tips for this model:

1. Recommended prompt: Cult of Cthulhu, a dangerous looking girl green, blood on top of a table with a demon head, in the style of giger in the hole, with a lot of blood around, hyperrealism, hyper realistic color photo, realistic photography, dark atmosphere, realistic blood, HDR, ultra detailed, sharp theets, high quality render, ultra realistic shadows, ultra defined

2. Recommended negative prompt: out of frame, lowres, text, error, cropped, worst quality, low quality, jpeg artifacts, duplicate, wrong proportions, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, username, watermark, signature, easynegative, bad-hands-5, [negativebs:0.2], (low quality, worst quality:1.3), high contrast, bad-picture-chill-75v, watermark, letters, words, censored, multiple fingers, multiple legs, bad hands
   

   Don't click this.

Generate At 768x1152

Set Clip Skip To 2

Prompting:

Use these keywords DEPTH OF FIELD, FILM, PORTRAIT, REALISTIC

for multiple girls use https://civitai.com/models/51136/multiple-girls-group?modelVersionId=55644

Lunar Diffusion

mooncryptowow's Lunar Diffusion model is a high quality blend based off 1.5 and was lovingly crafted to create the most detailed images possible while retaining a large degree of creativity in the renders.

This model excels at making beautiful people and complex machinery, but I strive to create a completely general model capable of creating anything.

For best results render your images at sizes higher than 512x512. Model contains "vae-ft-mse-840000-ema-recommended" baked in.

Check out my art pages, or support me on Patreon to get access to all of my prompts and latest versions off all of my tools: https://linktr.ee/mooncryptowow

This model falls under the RunwayML Stable Diffusion 1.5 license which you can review here: https://huggingface.co/runwayml/stable-diffusion-v1-5

a merged checkpoint of Protogen Infinity and Icomix, with baked in vae.

-rpg is merged with RPGv4 for added realism? (have heard it looks more cartoon-like, however)

i recommend generating with highrez-fix set to R-ESRGAN 4x+ anime6b with strength set to 0.5.

To upscale, first send to extras and upscale with your fav upscaler, then send to inpaint to fix eyes/hands/etc.

multicontrolnet also works well for webcomic or posed images.

for positive prompts, you can try appending the following:

(8k), dslr, fantasy, photoreal, sunbeam, high quality shadows, ambient occlusion, sharp, high resolution, detail, (200mm)

i recommend easynegative for negative prompts, but you may also find the following helpful*:

(worst quality:1.4), (low quality:1.4), malformed hands, bad hands, mutated fingers, extra limbs, blurry, bad anatomy, disfigured, jpeg artifacts, child, depth of field, bad eyes, child, cat ears, lipstick, lazy eye, bad eye, scary, poorly drawn face, disfigured, lowres, ugly, aged up, missing fingers, missing limb, fused fingers, one hand with more than 5 fingers, one hand with less than5 fingers, one hand with more than 5 digit, one hand with less than 5 digit, extra digit, fewer digits, fused digit, missing digit, bad digit, liquid digit, colorful tongue, blacktongue, cropped, blurry, JPEG artifacts, worst quality, low quality, normal quality, bad face, fused face, poorly drawn face, cloned face, big face, long face, badeyes, fused eyes, poorly drawn eyes, extra eyes, gross proportions, illustrated, anime

*beginning of prompt taken from Icomix.

V6.0 NEW MODEL OUT : This version fix color contrast. Like always, better fur ! You can remove bad-artist and easynegatives from negatives prompt.

This version doesn't use VAE and doesn't need one.

Please consider supporting me so I can continue to make more models --> https://www.patreon.com/BB95FurryMix

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 v5/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,

Negative:

I recommend to use boring_e621, you can add bad-hands-v5 if you want

Settings :

Steps: 30-150

Sampler: DDIM or UniPC

CFG scale: 7-14

Size: from 512x512 to 750x750 (only v4/v5)

Denoising strength: 0.6

Clip skip: 1

Hires upscale: 2

Hires steps: 30-150

Hires upscaler: Latent (nearest)

Prohibition is a high quality model trained on Public Domain, CC0 licensed Life Magazine and Country Gentleman magazine cover illustrations between 1914 and 1937, but with a primary focus on illustrations produced between 1920 and 1933 during the Prohibition era.

Example prompt (keyword triggers are in bold):

sd-prohib a female bartender in a white shirt, vest and tie making a cocktail during prohibition, medium brushstrokes, painterly, dynamic, family life, idealism, pictorial, country gentleman, 1920, cover art, illustration

Recommended Negative prompt keywords: logo, wordmark, signature, title, text, words, magazine cover

Subtle variations can be produced by reusing the same seed and changing the year:[1914|1920|1924|1925|1928|1930|1933|1934|1937]

switching between [broad brushstrokes|medium brushstrokes|fine brushstrokes]

switching between [country gentleman|life magazine]

and switching between [family life|adventure|romance]

I post about Stable Diffusion, model making and AI at diffusionary.substack.com

+OMNI+

Hello thanks for stopping by. This is my first model to be uploaded after months of nonstop learning.

I hope you enjoy. Here is the Mix contents:

1. PornVision Final

2. Ultra Azoyva +FruitFusion (added my own trigger)

3. Ultra Azoyva +Colorfusion

4. Ultra Azoyva +FeverDream

5. Ultra Azoyva +DRIPPP

6. Crystal Clear

7. Unreal Engine 5 B2

8. Unity Engine B2

____________________________________________________________________________

First off, I want to thank all of the original creators of the models used in this merge. You are amazing and helping the community so much by pioneering the space. We are all in this together, and this model wouldn't have been possible without your hard work.

This model was aimed to deliver a focus on what I love to personally create: surreality, in the areas of photorealism, colorful abstract, sci-fi, 3D, food, and stylized art.

This model does include NSFW images, but I didn't focus any of the examples on that material as it's not my style. I'm happy to say that hyperfocusing on my style has landed me a dream position, using stable diffusion. I feel very fortunate and want to give back as much as possible. This is my first model, and I hope you enjoy it as much as I do.

At the moment, I only know how to upload the full version as I am working with the Draw Things app via MacBook Pro. When I find out how to upload the pruned version, I will do so. I really appreciate you taking the time and checking out my mix. I have put so much love, time, and effort into studying SD. I've learned so much since stable it has launched, and I'm so happy to start sharing what I have learned with all of you. Thank you for taking the time and creating something with this model. I really look forward to seeing your amazing art.

++EASYMODE TIP++

Copy and paste entire prompt below and Just change what is in “< >“ and the style of your choice aftwards)  Feel free to play with the weights of the three triggers words. ***Fever dream does have 2 more trigger words but I didn’t list them because I don’t personally use them. Feel free to look them up and try them out. I really just wanted this to be as simple as possible for anyone wanting to create with this model. I know trigger words and weights were difficult for me to personally grasp at first. I hope this all helps. I want to thank JimEidomode as he helped lay out his prompt in a way i could understand weights.

Mac Users TIP: In Draw things app I actually will copy and paste the following prompt below the trigger when you upload it, and change what is in <>. Saves insane amounts of time and my image quality is more consistent. This is just a tip and not required per say.

***I am NOT using any LORA’s on any images. No post processing has been done.

My favorite prompt :

Jaw dropping image of a (sci-fi mothership) gushing with (dripping cake) by (greg rutkowski) style, medium format, fuji superia X-Tra 400, iso 4,000, perfectly octane rendered, pastel color palette, perfect composition, smooth, hard focus, ultra detailed, cinematic, highly detailed, ray traced shadows, 8 k, 1 6 k ,3 2 k, (FRUITFUSION:0.31):(BRTDRPP:0.14):(LIQUID SPLASH:1.2)

Weights can be changed according to you preference.

Negative Prompt for all photos:

The first 5 are negative embeddings that can be found on Civit.

(neglowres_2400:1.444| (bad_picture_chill_75v:1.33) | (by_bad_artist_neg:1.444) | (easynegative:1.333) | (bad_prompt_version2:1.3333) | 3D | CARTOON | BENT FORKS |(NSFW:1.3) | (OVER_CONTRAST:1.333) 

Best Samples taken with :

Eular A

768x768

Steps:20

Guidance:7

Clip Skip:1

Base Vae was used i think.. or no Vae. I don’t know how to do change or use other ones so whatever draw things does with VAE is what is used. 

If you have any questions please reach out. I will do my best to answer any concerns or suggestions.

Here is a link to my twitter profile for contact

Twitter Link:

https://twitter.com/ultraplayerone

Focused mainly on realistic portraits and flexibility, with higher native resolution support. (V8: 896x1152, up to 1024x1280, 1280x1280)

Example images are somewhat cherry-picked, no hi-res upscale used, no face-fix.

CLIP Skip 2 might help with realism in some cases while sacrificing fantasy aspects, experiment with it.

V8/V7 comparisons:

1024x1280

768x1024

V1 Recipe (might not be exactly how I did it):

Exp1-Tier1: (using add difference for all)

• GreyModel v12.0.0 + Dreamlike Photo v2 (25%)

• + Woop Photo (25%)

• + Protogen Infinity (25%)

• + ArtEros (15%)

• + HSPU (15%)

• + URPM (15%)

• + F222 (15%)

• + AlmostDirty (15%)

• + BondageV10 (5%)

• + TWAM (5%)

• + SXD1.0 (5%)

• + 3DKX1.1 (1%)

• + Hassan1.5 (1%)

• + OpenJourneyV2 (1%)

• + StrangeDedication2 (1%)

Exp1-Tier2-Part1: (still add diff)

• GreyModel12.0.0 + AnalogPhoto (25%)

• + Dreamlike Photoreal v1 (25%)

• + Amphibia Diffusion (15%)

• + Artstation (15%)

• + CALARTS (15%)

• + ChromaV5 (15%)

• + Corporate Memphis (15%)

• + NitroDiffusion (10%)

• + Dreamlike Diffusion (10%)

• + FracTime (10%)

• + InkPunk (10%)

• + Midjourney-v4-diff (10%)

• + RPGv2 (10%)

• + SeekArt MEGA (10%)

• + Colorjizz768 (5%)

• + Double Exposure (5%)

• + MegaPals (5%)

Exp1-Tier2-Part2: (add diff)

• GreyModel12.0.0 + MosaicArt (5%)

• + RealisticCartoon (5%)

• + Strahd (5%)

• + Technicolor (5%)

• + DMT Entity (1%)

• + Elysium (1%)

• + BalloonArt (1%)

• + DnDClasses (1%)

• + DarkVictorian (1%)

• + ComicDiffusion V2 (1%)

• + DiscoDiffusion (1%)

• + Dvmjv4 (1%)

• + Halloween (1%)

• + Lichtspiel (1%)

• + Lnkdn (1%)

• + Midjrny-v4 (1%)

• + Mtg-style (1%)

• + NekoMix (1%)

• + Nekrofaerie (1%)

• + SDHeroCalm (1%)

• + SpiderVerse (1%)

• + Ssmoky (1%)

• + SynthwaveV2 (1%)

• + TDST (1%)

• + VanGogh (1%)

• + Vectorartz (1%)

• + Voxelish (1%)

• + Wa-vy (1%)

• + Woolitize768 (1%)

Exp1-Tier3 (add diff):

• GreyModel12.0.0 + Babetopia (20%)

• + Instababe (20%)

• + PovSkinDream (15%)

• + PowerfulMix20 (15%)

• + GTMult v1? (15%)

• + LatexRaw (10%)

• + ClassicNegative (10%)

• + Furrystaber (5%)

Exp1-Tier4 (add diff):

• GreyModel12.0.0 + RealismMix (20%)

• + Hassan1.4 (15%)

• + AttractiveMix (10%)

• + VoluptuousMix (10%)

• + SamDoesBimboes (1%)

The Tiers separately are UNUSABLE because of all the AddDiffs, but merging them together brings them back to usable territory.

Probably merged these with Weighted sum as

( (Tier1 + Tier2Part1)*0.5 + (Tier2-Part2 + Tier3)*0.5 ) + Tier4 * ?

Then merged it with a custom DB model (trained on a face, but it also learned a bit of analog photo style) to tone it down a bit

= Exp1Final

Exp2 (using weighted sum):

• ArtEros + RealVision1.2 (50%)

• Grey12.0.2 + ProjectPhoto BETA (50%)

• SexyRealism + TheGirlNextDoor (50%)

• Deliberate1.1 + AnythingFurry (25%)

• Visigen + Futagen (25%)

• ArtRealV + GreyProjectPB (50%)

• DelibAnyFur + VisiFuta (50%)

• DelibAnyVisiFuta + SexyRGirlND (35%)

• ArtRealGreyPro + DelibAnyVisiFutaSexyRGirlND (35%)

• + Emotional Puppeteer (1%)

= Exp2Final

Exp1Final + Exp2Final weighted sum 50%?

= Exp3

Moving it back a bit towards photorealism:

Exp3+ (not sure about the weights from here on, but used weighted sum)

• (Analog + SimpMaker3k1 + DoubleExp + Attractive + Licthspiel)

• + (SexyRealism + TheGirlNextDoor)

• + (RealVision1.2 + ProjectPhotoBeta)

• + (Dreamlike PhotoV2 + Deliberate1.1)

= WyvernMix V1

We encourage you to share your awesome AI generations.

If you want to support my work, you can buy me a coffee or tea: https://www.buymeacoffee.com/gswan

"Unstable Dissusion was created in response to Stability AI's decision to neuter the 2.0 model."

This model is based on the official unstable_diffusion . Credit to the original developers and donors.

The data was collected by volunteers in the Unstable Diffusion community, and the model was trained by the Unstable Diffusion development team. You can join the community here: https://discord.gg/unstablediffusion

Sharing this model with the community. As it closely resembles Lexica Art V2 or Lexica Aperture v2, this product is an excellent choice.

It's absolutely amazing how we're able to create incredibly realistic faces with only a few simple steps.

If you want to support my work, you can buy me a coffee or tea: https://www.buymeacoffee.com/gswan

Enhance your writing prompt by adding the estilovintedois prefix, as it guarantees a consistent writing style.

"While you're welcome to use this model, please keep in mind that any potential damages or mishaps resulting from its use are not our responsibility..."

"This model was trained on a large amount of high quality images with simple prompts to generate beautiful images without a lot of prompt engineering."

"It should also be very dreamboothable, being able to generate high fidelity faces with a little amount of steps."

This model is based on vintedois_diffusion. Credit to the original developers.
Vintedois (22h) Diffusion model trained by Predogl and piEsposito with open weights, configs and prompts (as it should be).

Example prompts

• Prompt: photo of an old man in a jungle, looking at the camera

• CFG Scale: 7.5

• Scheduler: diffusers.EulerAncestralDiscreteScheduler

• Steps: 30

• Seed: 44
  
  Prompt: kneeling cat knight, portrait, finely detailed armor, intricate design, silver, silk, cinematic lighting, 4k

• CFG Scale: 7.5

• Scheduler: diffusers.EulerAncestralDiscreteScheduler

• Steps: 50

• Seed: 44

• Prompt: a beautiful girl In front of the cabin, the country, by Artgerm Lau and Krenz Cushart，hyperdetailed, trending on artstation, trending on deviantart

• CFG Scale: 7.5

• Scheduler: diffusers.EulerAncestralDiscreteScheduler

• Steps: 50

• Seed: 44

• Prompt: destroyed city

• CFG Scale: 7.5

• Scheduler: diffusers.EulerAncestralDiscreteScheduler

• Steps: 50

• Seed: 44

• Prompt: victorian city landscape

• CFG Scale: 7.5

• Scheduler: diffusers.EulerAncestralDiscreteScheduler

• Steps: 50

• Seed: 44

• Prompt: prehistoric native living room

• CFG Scale: 7.5

• Scheduler: diffusers.EulerAncestralDiscreteScheduler

• Steps: 50

• Seed: 44

v2-1_768-ema-pruned-fp16