How do I run it, I'm a beginner, what are the steps? I also have amd

Okay, so it seems to be a problem with Forge not setting the PNG info properly.

If you have a problem like this I recommend you every time you post an image with this model, come to the model page and scroll down and click on add post

oh, txk for ask this question :)

cool! thanks for your model, great work

i would even say that the dark-skinned characters turn out a little better than the rest.

I use either DPM++ 2M SDE + Karras (scheduler type) or Euler A + Simple (scheduler type) depending on the image.

you need to get rid of the "foocus enhance style" - then it worked for me

@berberry616 Thank you very much!

Thank you so much. I had the same issue <3

same here

@maximilianbuackmann934 If it helps, Pony does not seem to have this problem at all. At least the version I'm using. https://civitai.com/models/404154/wai-ani-nsfw-ponyxl

That's weird, it is my goto checkpoint NSFW with Fooocus.

@Gianni1900 I get nudity, but that’s about it.

Use only Checkpoints based on Pony Diffusion V6 for that sort of generations. Forget anything related to SDXL in Civitai's filters. Only: Pony.

@hansolocambo Thank you for the advice. :)

Decrease guidance. When you let model to fantasy a lot, it can fail. Find your balance for value of prompt strength per model

I had this when I was using the wrong kind of VAE

make sure kerras is set for the scheduler

If you're using Invoke, it'll sometimes keep the initial prompt in the "positive style prompt" and the "negative style prompt", and it will screw up any new prompt you type. You have to expand the prompt section until you see four prompt boxes, there's a little icont in the upper right of the main prompt box that does it. It's not quite a bug, but you frequently have to clear those out, especially if you're 'recalling metadata' from a previous image. This is on all sdxl/pony models, not just this one.

@Moxxle I've found that the issue is with Enhance style.

The issue I've found whenever I would enable the Enhance Style the generated images would just look deformed and unusable, but when I disable the style everything is normal still it's a shame I can't use the enhance style with this checkpoint or maybe there is a solution but I'm not trying to find one.

@mikas77 I'm glad you found the issue, at least.

I made 200$ from selling Juggernaut Images on Shutterstock (: . so i guess you can do it.

@vnuny Shutterstock doesn't accept AI image, how do you sell there?

@Tianci idk man they just accepted them, i guess if your images are realistic enough to them, they let you do it. (;

При доходе до 1 миллиона долларов можно ничего не платить правообладателю (таковы правила лицензии). Но Вы не имеете право регистрировать авторское право на изображение. Продавать можно.

All vanilla outputs from any AI text to image model have been ruled to be in public domain, so no one can prevent you from using them. Makes sense, as the models were trained on a wide variety of publically accessible but mostly copyright material. If you do extra manual work on it beyond just text to image, you could claim copyright.

There is an FP8 variant of this checkpoint.

Jugg seems to cater for almost anything NSFW or SFW, it mind blowing the amount of stuff it was trained on.
I've been able to cut out the use of alot of LORAS because Jugg does it by default.

@JohnnyWu22 are you using v9 or v11 (XI)

@QH96 I'm guessing it v11, because the checkpoint file was downloaded on 5th September 2024

No SD checkpoint can do hands 100% right 100% of the time, but Jugg is pretty damn good compared to others

This model is different than other AI models, it usually requires in depth descriptions of things to try and get what you want to picture, read the prompt guide by Adam. He goes through everything and gives good examples. One way I've learned to prompt is thinking of how I would describe it if I was writing a book.

DPM++ SDE Karras should work on DrawThings. Also, try to set the clip skip to -2. Not sure how or if it's possible in DT. Make sure your image size is 1024x1024 or 836x1216

Get rid of weighting in your prompt and if in Fooocus only select 2 styles. This model is trained with a new text encoder training process and blows up fast with too much prompt weighting. Dial it down. Start simple then expand out.

It's available in our platform right now. Three models. Juggernaut Lightning Flux, Juggernaut Base Flux, and Juggernaut Pro Flux. Follow our Twitter/X for updates.

No. Only SDXL base model recommended refiner when it was released, but everyone uses a finetune, and no one uses the refiner model.

use wild card like celeb-female

@bigdog11 I figured it out, it's because of a lora

@RunningSalmon also you can put CFG scale 1 or just lower

i think you can use control net union

did you figure out anything ?

@KandooAI they tariffed the models.. that sucks!!! who thought this would be a great idea!!!!

The most popular XL model out of the generator. WTF

@ffc4now crazy right?

What?? Why we have to pay for this thing to just get access to the model. Like, we can wait for days to use it again??

@AyeYouNeek Have faith fellow civitan, this is just a short phase of hardship. Soon billions of Buzz will flood into Civitai because of the tariffs.

Newer versions don't always mean 'better'. Look at the absolute series of disastrous releases that Lustify made after v2. They put that checkpoint into the toilet.
But yeah, i'm with you in wanting a new shiny version to play with.

@JohnnyWu22 you're totally right. i'm not really an expert but i have the impression tje Juggernaut model improved quite a bit since it came out though. I'm not on f.lux so i was wondering wheter we could have a glowing exit for SDXL xD

@Aethnos I'm definitely not an expert either - I use Fooocus ! :-)
I Asked Perplexity AI all about Jugg, and it was it was the best SDXL checkpoint for NSFW and SFW realistic photos.
I was asking the AI about Flux and SD3.5 too, it said that the people who made those have hard coded a block of NSFW images. Looking at the Flux models on here, the best they seem to have is some topless stuff, so maybe they'll find a way around it ?

The new system is dumb and only takes short-term trends into account.

Local gen is the only way to avoid enshittification.

Time for all to invest in expensive GPU. How's that gonna work for ya civitai ;p

simple solution to that problem, block the user and you wont see any of their posts anymore 😊

@mariedoll123 You can do that? I mean, i know you can block them but that makes me not see his work, thank you very much.

@LordWolf255 your welcome 😊

She was part of the training data.

@KandooAI let me rephrase "they resemble her" but not her. In fact XI makes them look so beautiful. much better than the real deal! lmao

@mhocker590 i dont think so.

@KandooAI i can't wait! but im in love with XI

I have the same problem, when I use the recommended sampler the image looks worse and grainy and messed up.