I'm not an expert in creating images but I've noticed that arms now look longer, there are perspective problems, men look more feminine, difficulty detecting certain physical and genital details.

This looks so good

接下来有更新FLUX版本的计划吗？

之前一直在用pony版本，然后刚刚开始用这个IL版本发现依旧是顶中顶。

作者，你好伟大🥹

私はv9以前に戻ります　V10以降手首とかの崩れ発生率が大幅に上がっていると感じます

ここまで早い速度で新しいVerを出さなくてもV8あたりで完成されてるmodelだと思っています

Greta Model, really helping me make some cute Baobhan Sith images.

支持作者，成熟风格是没错的，因为现在大部分模型都是偏loli的，再说了，喜欢哪个版本就用哪个呗，喷作者是没必要的

作者大大，你下一版本会更新啥啊，打算做之前的真实亚洲的pony模型的Illustrious版本吗

v11评价两极分化看起来相当严重……个人愚见而言，我自己常用的风格串在v11的确变得更精致更好看了，提升巨大，这是非常值得肯定的，也是我觉得v11风评分化最可惜的点。

但是当初选择老歪的这个系列的模型除了很多其他优点，就是冲着能容易生成比较纤细的身材来的（尤其是有大欧派的情况下，其他模型总是会容易变成坦克，而wai不会，这对我这个胸控而言根本无法拒绝），现在不少人吐槽v11的就是身材太容易成熟这点了。

我也尝试过用一些和纤细身材以及年龄相关的提示词进行改善，但要么直接变幼，要么就没什么效果……太幼的脸固然不太好，但是更容易生成成熟的身材也很难让和我一样喜欢纤细身材而选择wai系列的老用户完全接受v11。

老歪的微调模型质量说顶尖一点都不夸张，身材这东西本来也是萝卜青菜各有所爱，除非模型能够真正变得十分灵活，否则我想很难两边都照顾到。个人而言还是希望老歪能够稍微重视一下这个问题，实在没办法的话干脆把这个系列再分成两个子系列，面向不同人群，我觉得也算是个解决方法（）

Did i miss some drama? Why those comment that just thanks author or say they like model get heavily thumb down? Did I missing something? Model look fine on my end.

Love the model, just updated from V8 to V11, is there something I'm doing wrong? V11 was a objective downgrade for me in visual artifcating with the same style I used before.

这个模型是我用的这个系列的第三个，比之前的细节有明显提升，作者加油！！！

Thanks again for your great models.

Personally, if you could make a model that combines the sharp and clean contours of v11.0's character with the colors of v9.0, it would be the strongest model ever.

Is there any way to control dick size in that model? Dick slider from goofy_ai don't work at all, all dicks are way too long.

This statement might get me under fire, but oh well. v11, with the right prompt is amazing. The base prompts do create issues sometimes. So, I will share what prompts I have been using. So far, using my prompts, I have not seen artifacts or other issues that others have been saying. Do correct me if I'm wrong, as I have not tested it with LoRa's that much.

Here are my prompts and other details,

CFG Scale: 7

Steps: 30

Hires Upscale: 1.5

Hires Step: 20

Hires Upscaler: R-ESRGAN 4x Anime6B

Denoising Strength: 0.14

Sampler: Euler A

Scheduler: Normal

Positive Prompt: {prompt}, masterpiece,best quality,amazing quality, absurdres, very aesthetic, newest, very awa,

Negative prompt: bad quality,worst quality,worst detail,sketch,censor, simple background, foreshortening, bad proportions, uneven eyes, small pupils, clenched hand, small fingers, (6 fingers, 6 digits, extra digits:1.25), clothes writing, abs, ribs, stomach bulge, wet,

從8用到11, 大概只有11是我在批量產圖的時候出現最多破圖的...這版本感覺在Lora上適配有問題, 我自己用起來是有用角色Lora時最容易出問題

i think v9 is the best, v10 and v11 occur finger problems.

I think v9 is best for now.

Many of characters Lora I use look weird after v10.

Honestly, v11 feels like an extreme downgrade compared to v10 in terms of drawing style. I compared them side by side and v10 is much better, so returning to it. Hoping that the new model will return the style of v10.

i think this is pretty good of a model. keep up the good work

so, how can we know which characters includes this model?

Come one guys, its been great model so far, especially since this model is free and there is no early access (buzz) in each update. I think all project models / lora in each update may experience a decrease in quality than before, and I think that's normal.

you can use the version that you think is good while waiting for a better update in the future.

current version maybe feels downgrading or worse than before. but lets believe on author it will be better on upcoming version.

Does it have NoobAI-XL Epsilon in recipe?

Version 9 works great with Lora ponies and the style is also like a pony. Version 10 already with pony lora does not work or works poorly, the style is much different. Version 11 did not check the work with the lora, the style is also different. So for the transition from ponies, version 9 is great, the new versions clearly do a great job, so you can start with them

is the artist list for this the same as the pony version or is there a list?

Life is short and youth is fleeting. People prefer their younger selves.

So I prefer V9 to V11.

Thank you for the useful information about WAI V11.

I tried lowering the CFG to 3.5 and printing, and it seemed to be printed similarly to V9 and V10.

There doesn't seem to be much difference in the way the hands crumble.

This is just my personal opinion, but I think it will look good if it combines the colors and reproduction of V9 with the composition etc. added in V10 and later.

Hi i am using your checkpoints daily.

can make one for only anime characters ?

Hey, is there any way to find out all available characters? I tried the link it doesnt work for me

I have trained all my LoRA models with this model, and I must say that it is one of the best when it comes to anime, manga, doujinshi, and CG is concerned. This update seems better than the previous one in almost every aspect, although I notice a slight degradation in eye quality. In other areas, such as hands and background scenes, it seems improved, I don’t need to add specific prompts to generate good hands.

The LoRA models work well, and when using Highres-Fix with high image scaling values, the image doesn’t suffer from distortions. I think in the next version, you should focus on improving the quality of the characters' eyes, as that’s the only aspect that seems worse compared to previous updates.

Great job! I’m eagerly looking forward to your next update.

how can i make an image without any character?

我挺想知道wai系光辉模型和noob的lora模型能不能通用，有一部分我经过测试可以通用有一部分又不可以，我不知道是我的问题还是其他原因

so what would there be in v12

各位大佬，想问下如果要制作角色lora用哪个模型做底模比较好呢

What exactly changed in v11?

各位大佬有没有comfyUI的工作流分享一下

Can you make version 9 available for online generation again please?

v90 for base generation and v110 for upscale refining work together as a great pair

Im getting a pink shade or yellow shade on pictures when hd resing with the online genarator. Even with cfg at 4 and denoise at .40

amazing model, but I have been having problems with getting certain positions. It tends to focus on doggy positions mostly...

Is there a way to see the metadata of the images we download for this model?

我想咨询一下

Illustrous 和sd1.5和sdxl哪个更接近？lora是否具备通用性

in v11, It feels like the eyes are easily blurred when generated directly. (Not restored in Hi-res fix)

good

How to fix face and particularly eyes ?

Literally amazing and beautiful! 💖

🫡

就我用V11生成十几张图体验下来，V11生成的人会更成熟。和以前版本比的话，一个是十六七岁的妹妹，一个是大了三五岁的姐姐。

至于变胖，有些图确实变胖一点，看着怪怪的，但有些图又没什么变化。

我不好说哪个版本更好，就我个人喜好而言，我更喜欢之前的版本

Does anyone know how to specify the male protagonist of FateGrandOrder using Prompt? I'm having trouble with the female protagonist appearing when I specify Fujimaru Ritsuka.

This is good

how to use ADetailer in ComfyUI?

One little suggestion for future versions: some training on "prone bone" position so it stops putting an actual tibia in the bootyhole.

V9 was a 10/10
V10 was a 6/10
Finally V11 is a 8.5/10
Why? Read on...

PROS
After using V11 for a bit longer and trying out different settings. There is some important updates that make it worthwhile.
-Much better prompt adherence
-Better anatomy.
-New style looks good with most IL/NOOB LoRAs


CONS
-New style can be have too strong a bias or overpower the image based on your prompt. You might need to lower CFG

-Some people might prefer the V9 style which was very bright, colorful and highly anime-esque.
(V11 looks more painterly and 2.5D by default)

Nice one

面部总有阴影，想问一下有没有面部无影或者给面部打光的方法。

There are always shadows on the face. I would like to ask if there are any methods to make the face to light it up.

It would be my favorite model by the amount of details and high definition it gives, no blurriness. But unfortunately it creates all characters with smaller faces and tinier bodies when in comparison to model like PASanctuary, I am not sure if there's a way to remedy that without sacrificing quality...

找了很久，终于找到了

Works great, dunno why it was cooked on this site initially but running it locally works like a charm however I'm having issues with inpainting, can anyone recommend a checkpoint that works well with this one ?

Really hoping v12 reverts the style change because while the improvements in detail and prompt adherence are great the problem is not every character needs a tiny head, stretched out body and a fat ass.

How do people still use on-site generator on previous versions? I want to use v9 or v10 too >:)

Trying to run local on A1111. I can't use Openpose in ControlNet. Is there any specific download for it to work?

after reviewing the vast amount of comments good or bad or requests brings me to the conclusion that people love to demand//complain about something that is free to use or do not understand anything about how to configure their settings on their own should make their own checkpoint and distribute it freely for others and feel how your effort is being trashed. this is why we can't have nice things. the least you can do is figure shit out yourself than to bother someone who gives this out for free to use for your private degen use. internet always brings out who you truly are and you all should just vanish from existance. so many dumbasses making people who try to do good work look bad.

Honestly, I was really hating this model when it came out, but after some more testing, I really like it
I love the new, more mature style a lot more

It would be perfect if it didn't have so many issues on the body, especially when upscaling

I was very hesitant to use this model since I really enjoyed the flexibility other models have in style prompting. But once I started trying this model I can't stop, it's quality is incredible, the adherence to prompt is great, it delivers entirely, and while it is less flexible it's also alot more stable.
10/10 would recommend again.

Anyone who interested in GGUF Quantized version, i made it specially for comfyui-gguf usage

https://huggingface.co/kekusprod/WAI-NSFW-illustrious-SDXL-v110-GGUF
You're welcome!

Can anyone direct me to a file filled with all the current available character’s name

amazing

v11版本感觉对画师tag压制很严重，适合不带画师的裸出图吧，在画师tag的情况下更喜欢v9，风格更明显，至于v10感觉在两者之间，感觉v9的基础上只需要增加lora的适配性，和新角色支持，这样出来的v9加强版比目前的v11,v10要好得多。

歪佬这模型解剖学好的离谱了吧？

- 当然，也可能是我用NoobVpred用的

- noob风格超棒，但是细节和人体不大行，要是两个模型能整合一下就好了

♥

While v11 output objectively better overall images, I feel like it does a weird thing with their eyes.

V11出的圖風格感覺比前幾個版本都更加成熟一點，

對於喜好大姐姐和熟女的我說堪稱完美，

在生成御姐角色的時候面相和身材比例感覺也更加契合

除了感謝以外沒有其他形容詞好說了

This checkpoint won't recognize some "Ponys", like Maud Pie. Please no kink shaming :(. Or am i prompting wrong?

Does this model have an issue with prompt editing? You know, the [A:B:N] thing in A1111. Character expressions often go blank when I use prompts like [anything other than expression:0.3], exactly at that point of generation.

Regional Prompt extension is a possible culprit, though I've never seen this kind of issues with PonyXL and its variation models while using that extension.

Sooooo clean! (using Ver 9)

Awesome

非常强大的模型，t2i非常好用
- 人体很好，简单的姿势几乎没有人体错误，复杂的姿势也可以做到
- 对着提示词有着极高的服从性，基本上提示词里的内容都能准确展现出来，提示词之间相互污染的也极少
- 模型也可以通过提示词发现一部分提示词里衍生内容，抽卡也不错
- 兼容大部分pony lora
特色：模型自带的风格极其强烈，如果要100%复刻风格较难做到，角色lora效果不错，但是风格类lora会混入模型风格。

Hi, there are so many versions, which one is the best overall? for more built in characters and styles?

Very very awesome.

算是一个简单的问题反应吧

该模型下的所有版本貌似对long nipples这个关键词极度敏感？虽然我没法稳定复现，但是这个关键词只要加了权重，在最新版本forge的环境下大概率会出现炸图崩坏的严重问题，最严重时只是给这个关键词加个括号也会导致炸成鬼图或单色图，我最高只敢加权重到1.2，加到1.3就开始出现不稳定的情况，加到1.4及以上几乎必定炸图，用其他诸如hassaku跟noob以及其它IL系模型会轻一些，但这个模型问题格外严重，请问有什么好的解决或改善办法吗？

I see the list of characters, but is there a list of all artists? I've been blindly trying them, not being sure if the trigger words are correct or just missing.

Thanks a lot for all your hard work you put into this model and the continous improvement of the versions! :-) I have tested version 11 thoroughly and I am deeply impressed with the results that I get. Prompt adherence easily outclasses any other model I ever tested so far by at least an order of magnitude.

I have two detailed observations I would like to share:

1) I noticed that general LORA compatibility is good, unless it comes to LORAs that intentionally (or unintentionally) try to modify the art style. This holds true even for art style LORAs dedicated to Illustrious models and is especially apparent when looking at the eyes of human characters. I would deeply appreciate it, if you could look into this for potential future versions and allow changing the base style of the model with LORAs more easily and consistently.

Having said that, let me emphasize that the base style of your model and the degree of detail is outstanding and I really like that. I also recognize that the many built-in (danbooru) artist styles are easily available inside the checkpoint which will potentially be enough for many users, but some might still want to add specific art styles through LORAs which are not built-in.

2) The general "art style consistency" for the "vanilla art style" in your model is breathtaking and beats many other models I have tested. This is an excellent feature and should be maintained in future versions, IMHO.

Once again: Thanks for your amazing work and the efforts you put into this! Keep it up! :-D

Has anyone figured out a means of negating the style change in v11? I keep coming back to try it and have been really impressed by a lot of aspects of it but I always end up reverting to using v9 due to the realistic proportions just throwing everything off.

Very nice, thank you

有人知道如果畫面有兩個角色，提示詞怎麼下比較不會讓兩個角色的一些外觀提示詞互相汙染嗎

暫時最好的IL模型，沒有之一，感謝作者的付出！

我在測試使用vpred + zsnr訓練出的LORA與WAI V11的相容性,我發現使用你的NOOB AI VPRED 模型作底模以vpred + zsnr 方式訓練出的LORA，相同的素材和TAG訓練出來的LORA，在WAI V11 上運行竟然比以WAI V11為底模以EPS訓練的成果更好，不管在色彩 服裝還原度 與光影表現上⋯⋯就很神奇

另外我也不明白為什麼大家說肢體有問題，我以V11 畫了超過500頁的插畫，幾乎90%出品以上都是開箱即用，除了少量手指偽影外，完全不需要額外修正。 當然我是只用自己重新訓練的LORA 配合使用，不排除是因為別人套了PONY 系的LORA 導致

為什麼作者你這麼強，有空能不能也更新一下NOOB AI VPRED 的版本，讓其更貼近V11的藝術風格，哈哈

WAI V11我唯一遇到的問題就是原生出圖，眼的高光不太理想（普遍都很平坦），但我的工作流上有以相同提示詞相同底模對臉部修復，經過修復後眼睛又很有神，可能是因為解像度不夠？

Has anyone noticed a problem with the bellies? I keep getting strange results when using hires.fix, each time double navels are generated regardless of denoising strength. I would be grateful if someone could tell me whats the matter

honestly, I prefer version 10, the strokes are prettier in most illustrations, but above all, it better accepts the styles of artists who are already trained in the model. Anyway, both versions are available for use, so it's not a big deal, this is one of, if not my favorite illustrious model

we need more One Piece Men Loras working with this one! <3

Pretty solid. I like it.

Is it NoobAI mix or not? Why no information about recipe? We left SD1.5 shitmix era year ago 😂

What is this?

Where i can find the in-built artist styles?

Has anyone managed to get Controlnet to work with this model? If so what are you using?

V11不开ADetailer的话眼睛崩坏的几率有点高，特别是加lora时，经常看着面部修复把各种色块混合的眼睛修复正常

All the images I'm making look bloated and with some sort of color filter (yellow, green, pink) and I'm using Automatic1111, not the civitai generator. Any tips?

Edit: Installing the extension provided with the model seems to have fixed the issue for me.

yaayagood

ok

偶尔色彩会过于复杂，步数越高色彩越复杂

This model is powerful!

how to make an image with no characters?