✨ Tired of typing prompts + want viral-ready art?
Pair Lazy Embeddings 🛋️⚡ (zero-typing magic for all models ) with TrendCraft 🪄🚀 (30-day trend-trained detailer ) for effortless masterpieces. Check them out: Lazy Embeddings + Trendcraft (Pony XL , Illustrious , Noob Vpred, or SDXL base trained version)
👉 Join me on our discord server (https://discord.gg/tWrKk3jcXx)! Share art, ask for help, or just chat. We have CivChan bot plus games there!
👉 Support future updates: Buy me a coffee or tip buzz! Every bit helps. Thank you for fueling the art! 🚀
🚨 JANKU 🌟
✨ Key Features:
🔥 LoRA-Free Workflow: Generate high-quality art without stacking detailers (unless you’re adding niche styles/concepts).
🔞
𝐍𝐒𝐅𝐖Mastery:Better anatomy (yes, even in complicated poses) + improved male framing.
𝐍𝐒𝐅𝐖: if you want more control, put lazy𝐍𝐒𝐅𝐖in negative!
Large character list support: Character list shared in our discord server.
No more stability issues —generate confidently at 1536x1536 without up-scaling. My personal resolution to use is 1024 x 1536.
🎨 Crisp Details: Enhanced backgrounds, outfits, and character features.
👀 Prompt Responsiveness: Gets your vision right the first time!
🔧 VAE Baked In: No extra steps—just generate.
Read version details for the version you are using for more information!
✨Improvements from RouWei
Easy and versatile prompting with great aesthetic, anatomy, and stability
Vibrant colors, smooth gradients, and full brightness range
Over 35k artist styles pure training (no merged weights), and customizable options. The list in .txt can be found on our discord server.
Note: you must use
by artistnameto trigger it!
Eliminates watermarks, tag/character bleed, and strange artifacts for better coherence
Improved prompt following, stable styles, and accurate artist representation
🛠️ Recommended Settings
Steps: 25-30 (I use 30 normally)
CFG: 3-5 (I use 5 normally)
Sampler: Euler / Euler A (my go-to)
Scheduler: Normal or Simple (my go-to) or SGM UNIFORM
Model Recommended Resolution：768X1344 832X1216 768X1280 704X1408 1024X1536 (my go-to, no need for up-scaling!)Standard Illustrious & SDXL resolutions also supported
Prompting: Use my Lazy Embeddings for positive and negative and even 𝐍𝐒𝐅𝐖embeddings to prevent or add 🙈. Use lazywet to remove wet and shiny skin. No more guessworks for prompting! If you don't want to use embeddings, use your normal NoobAI/Illustrious quality triggers. There is artist, danbooru, and rating tag support.
Positive promptlazypos {character tags} {detail tags related to character} {background tags}lazypos, 1girl, makima \(chainsaw man\), yandere stare, smile, dress shirt, grey jacket, red hair, sidelocks, black necktie, looking at viewer, jacket, long sleeves, upper body, suit, necktie, shirt, simple background, medium hair, collared shirt, ringed eyes, shaded face, braided ponytail, head tilt, braid, white background, formal, yellow eyes, holding own arm, wing collar, closed mouth, solo
Negative promptlazyneg
Pro Tips
Using my Lazy Embeddings for onsite gen doesn't cost you extra buzz, if anything it helps you save buzz since you're not generating bad images!
Check my images’ metadata for exact prompts/settings. Import them into ComfyUI to reverse-engineer my workflow!
🩺 Fixes & Feedback
Eyes still wonky? Use eye detailer, but most of the time you don't need unless the subject is further away.
Found a bug? Report it—I’ll fix it in next version. I only generate with this model, so I’ll keep improving it. I won't publish something I won't use myself.
❤️ Why I Made This
Tired of juggling 10 LoRAs for 𝐍𝐒𝐅𝐖 art? Me too. I want a 𝐍𝐒𝐅𝐖 checkpoint with more knowledge of characters and other common booru tags. This checkpoint is my personal playground —designed for creators who want bold, detailed, no-BS generations.
Future updates? Always. Let’s push boundaries together.
Please share your feedback to improve the checkpoint together! ❤️
JANKU Checkpoint License & Terms of Service
Comply with the NoobAI-XL 1.0 License
User Responsibility :
You are fully liable for all outputs. Use prompts to filter content.
Restrictions :
No redistribution of modified versions without adhering to open-source terms.
No use for unlawful or harmful purposes.
Failure to comply voids usage rights.
Description
This version is created using a different unreleased style lora, no change to knowledge of base model. It is done by adjustment of 3rd style step of JANKU v4.0, here are the main differences:
Shiny hair with particles and just more detail particles in general!
Shiny eyes when it is closer!
Sharper lines!
Different kind of details compared to v4.0 base.
Different kind of lighting compared to v4.0 base.
Let me know what you think of this style as I might make it the default style going forward depending on feedback.
You will prefer this version more likely for SFW, photo perfect NSFW, or close up creations. The knowledge of both model is the same.
To celebrate CivChan new feature of chatting with discord users in our server, you can join the discord to download this model for free https://discord.gg/tWrKk3jcXx. If you want to support me still, please buy early access as it is only 169 buzz 😉
FAQ
Comments (135)
v4.0 Style A released. Please see version detail notes for more information! Check out my cover images!
To celebrate CivChan new feature of chatting with discord users in our server, you can join the discord to download this model for free https://discord.gg/tWrKk3jcXx. If you want to support me still, please buy early access as it is only 169 buzz 😉
:O new update !!
I have installed three versions: 3, 4, and 4A. Version 3 has good compatibility, while versions 4 and 4A are on par in terms of advantages and disadvantages. Since I can't find a reason to uninstall any of them, I have to switch among the three occasionally.
@LYNX233 I agree! Use what you like since you can generate locally!
Hello, where can I see the update of Version 4A
@moon_smile At the top right check the version details
@janxd got it, thanks
Hi,
I really like your model and I was trying it because you publicize it as illustrious and as having the illustrious license. However, now, I saw that you claim that it has the NoobAi license (without commercial use). Owing to this, I will not use this model, because I need commercial use.
regards,
This model was always with NoobAI license. The title of the model always include NoobAI in it.
@janxd I did not explain myself correctly... the info that gives the characteristics of the model characterizes it as illustrious and characterizes its license as the illustrious license. Thus, if one does not read your description carefully, he will consider your model as having the illustrious, and not the noobai, license. I understand that your model being based on noobai must have the noobai license, but, in order for you not to inadvertidly mislead your users, you should change the license on the characterization of the model on the info displayed on the top right corner. You can, for instance, call it a junku license!
Your model is really good, Thus, it would be great if you did a pony or illustrious version of it, so that it would have a commercial license.
@MairFunta You just ignore my comment about the title having NoobAI since the beginning and even the tags have NoobAI and NAI tags. Since v1.0 the recipe has been shared and always mentioned NoobAI EPS 1.1. No one is being misled here. If you are using models for commercial use, you should do your job and verify what license the model is using and not make any assumptions. Lots of comments since the beginning asking if it can be used commercially and I always answer NO since v1.0. I bet you are using a lot of illustrious models out there and don't know they have NoobAI in them and breaking NoobAlI license. It is very easy to test if a model has NoobAI in it.
@janxd I was just making a constructive critique... A newbie does not know that noobai is a different thing than illustrious, as the model presents itself as illustrious, one accepts it as such...
I was not saying that your intention was misleading anyone; I was just stating that at the top of the page you claim it is illustrious but at the bottom you claim it has a noobai license... I was just giving you a notice that somthing was wrong!
If you have a way to test whether or not an illustrious model is made from noobai, then I would like to know how that can be done, because I do not want to use any noobai one! Thanks in advance!
@MairFunta If a model responds to the following tag: newest, absurdres, very awa. Also if the model knows a lot of characters, it is most likely have NoobAI in it. There is also a technical way to check is by looking at the blocks.
You can blame CivitAI for the labeling. CivitAI did not have the NoobAI category till recently, but they didn't allow for resources mixing of different types. All my resources are marked Illustrious since the beginning before this new category. You can see older models before this change doing the same. If we change it, our resources can't be found correctly. This is why creators have to put the Noob license in description.
@janxd Thanks! I understand! I just want to be on the safe side!
Do you have more info regarding how to look at the blocks ? I can program in python, so, if there is a test I can do to the checkpoint in a technical way I will do it!
@MairFunta I can't help with this through text, but since you know how to program it should be easy to figure out by yourself through research. Look at NoobAI and poplular NoobAI models, compare/check other model weights. Find the common weights that are not Illustrious.
@janxd Thanks! If you remember any way I can test or any source I can follow, I would be grateful. Best regards.
@janxd I know how to program but I am a newbie in what regards AI. :D Thus, I do not know where to search for...
So I really love this so far, but this seems to struggle with effeminate male characters/femboys. In general, any males that aren't huge are nearly always depicted as females, or having no bottom-half genitals SFW specifically. Like tight fitting clothing will almost always be portrayed as having nothing "down there". any tips?
Use the tag bulge and adjust the strength if needed
Thanks for the tip @DoctorAnnihlax !
@janxd for the most part this works. Surprised by how much I had to increase strengths still compared to v3.0 still my favorite model overall for what I want
Thanks for this, it looks great! Just a quick question - for general use with slightly more complicated prompts, should I use this checkpoint with the other LoRAs you have or would you recommend without?
You can use whatever loras you want as long as they are marked with illustrious or noobai! You do not have to use loras unless the model don't know it.
Great, thank you! So I'm guessing something like TrendCraft would be good as a general use LoRA since those are more subtle changes, but something like a "virgin killer sweater" LoRA isn't needed since the model would most likely know it.
Thank you for the support!
太棒了太棒了baby，希望你保持热情，我需要你😘😘😘
Want to know if there are any improvements in the details of the genital area. Most models seem to only generate genitalia that "looks similar," but upon closer inspection, it doesn't hold up to scrutiny.
Models including this one generally refines for default style, adding characters, and artist styles. The genital area is small, so it is better suited for loras.
An exceptionally great model !
I’m trying to train a LoRA for prostate milking pose, but none of the other popular Illustrious models could understand the position—only this model was able to comprehend it.
I loved your models and they are my favourite 💓
Thank you for creating these models. Their body pose and facial details are detailed. I have also used Pie Models (Blueberry Pie) and Illustrij. Although they can compete with your models, I ultimately loved your model the most.
If possible, create a model with a detailed style of clothes. If you have ever watched Chinese anime, their clothes design is incredible.
Some anime examples - Jade Dynasty, Perfect World, Throne of Seal, and more.
Throne of Seal, this anime has the best graphics and animation. If possible, try this once.
Also, create a model for these types of characters.
Thank you for your support! I can add it to the list of things to work on, but my paid commissions comes first.
Love the model, I change the WAI NSFW for this one and it feels like an improve :)
Is there a way to know the characters tags?
Yes, if you visit our discord, you can download the txt file of them. There is also a new list in the next version with more new ones.
bro this model is amazing, is very easy to get gorgeous looking pics, with some other models i struggle a lot. Thanks bro
Hey I just wanted to congratulate you on this amazing model, I have some advanced comfyui workflows that depend on controlnet pro max repaint feature and I've been testing tons of different models with many degrees of success and let me tell you this is by far the best one. Thanks!
My personal #1 atm!
Love the styles and colorpalettes this generates.
Sometimes the images look stunning even before you refine them even...
This is my go to model to go try and tinker with any new idea I have.
Primarily NSFW, tho.
I really need to learn how to use these properly. I can never get my desired result in Fooocus.
厉害
NoobAI is a derivative of and inherits from fair-ai-public-license-1.0-sd, which explicitly states that NO additional restrictions can be added and it CANNOT be replaced with a different license. Please stop spreading NoobAI's invalid license around and acting like it applies.
Btrw, How do I get rid off the target/bullseye in the background when using the model, this is more like a general question.
very good!!!!!
This model is phenomenal, but it keeps adding vaginas to my futa characters. How might I prevent that?
I have tried futa_without_pussy in pos and pussy in neg, but it still happens.
Because futa with vagina it real futa. You want dickgirl.
Pretty good, I usually use wai but now I really like this one, especially its aesthetic style. And it is the best model which strongly respond to artist prompts, no any other models can do this so far. Top 1 tier.
However this model's stability is still need to improve, in certain number of prompt words the image aesthetice will break down, or even fail to generate (whole pictures white or black).
Btw, this model have more obvious prompts pollution, such as blue eyes will easilly make blue hair tips. Though this may be caused by its stronger aesthetic style, and no much SD model could handle this problem very well, not a big deal. Using lazy emb can alleviate this problem, but personally evaluate the ability of the model itself so just mentioned.
After using it for while, I found that the pussy may easy to be blurry even the camera just a bit far under artist style. It seems not that easy to fix by just adjust your prompt words, you need to change the seed and reroll. Another similar issue is hair style may often generated with hairpin under artist prompts.
But overall, this is definitely the best anime model for now. The more I use it, the more I like it, and the artist prompt book is amazing.
Great style, but this checkpoint will have a problem generating male characters.
I do offline stuffs. So I need to download checkpoints. Does anyone know the difference between 4.0 and 4.0 Style A? Love the 4.0 but is it worth an update?
Not a bad background, but 6 fingers, really? And even in 1536 x 1024 resolution it tries to cram two models instead of one, breaking the anatomy.
Was really hyped when I saw the artist list, thought I'd have some fun experimenting.
Don't bother. This model has exact same artist recognition ability as other illustrious models. Not sure what was used for the MEGA showcase, but most of these have exactly 0 influence on a simple portrait of hatsune miku with minimal negatives and positive prompts. All this model does in my experience is lose in stability and details to Hassaku and Wai models.
the MEGA showcases are generated by Rouwei 0.7 not JANKU
Does anybody know how to fix the model producing too many upper eyelids, like the first two example images on this model?
Use eye detailer and increase area plus denoise. I did not use it in those Pictures. Or just use 4.0.
Hello, I am a beginner. Where should I go to look for a character list? Thank you for letting me know.
Really good list to get started
https://well-fresh-lawyer.glitch.me/?source=https://files.catbox.moe/6holoy.json
v4.0 Style A has artifacts on the eyes very often
I hope this can be fixed because I do like the style of the eyes better
So good at least for everyday
nice
Nice model
I eagerly await the announcement of the new JANKU created based on the RouWei v0.8.
Nice, keep up the good work
I really like this Checkpoint, but is anyone else having the problem of it MASSIVELY skewing towards female characters? I'm writing prompts that are 100% masculine. No feminine terms used. 1boy, male, muscular, etc.....and I am still getting skinny big breasted female characters.
That's many illustrious models in general, which are overtrained on waifu art
Because most people want to see women, most of the art is of women and most of the AI models are trained on the women that the art is of. Regardless of people's views gay men who want to see men are a niche minority just like in real life. Maybe 1 in 1000 men are gay.
@hikingferal That is unfortunate. I had to switch over to another checkpoint. Having much better results with Nova Furry ironically, even with human males. I'm bi myself. So it would be nice to find a checkpoint with a little more flexibility.
Good
Crisp Images
Ai is going crazy
Add Style A in tensor please bro
nice style :)
So good
Good!
This is soo good 😩😍
Which characters are supported? Is it the same as the noobai list?
You can try the characters from the Illustrious list and from the RouWei list. Most should work but the model has forgotten some of them.
Pretty good
Wow
It has been useful flr the learning process
Nice
very nice
what would be the best approach to character lora training, if we want to maximize consistency with this model ?
train with illustrious v0.1 base
@janxd I trained Loras with illustrious v0.1 base, they work fine with Illustrious XL but not with Janku. What am I missing ?
Very Nice.
Any workflow for this that has lazy embedding and trend already in it?
Bro, did u know, comfyui working better with text or n?
What about pinning the work of someone who hasn't used your model?
Very nice model, Thanks!
Thanks for this model
This is a really amazing model, Thanks for this.
very good
really really amazing model 😍
generates really beautiful image but very weak in nsfw prompt adherence
EDIT: also tends to overfit... a LOT. OP recommends cfg 5 but I can't even generate eye color consistently at that CFG. Either have to use style strength controller to reduce overfitting or just stick to lower cfg.
are you using the lazy embedded on your prompt? because, I use them and there is no problem at all...
great Model.
any artist recommandations?
great!!
Love the this model but theres one thing bothering me. It cant really do hands and feet perfectly, when trying to generate feet, theres either 6 toes or 2 toes merging into 1 when the character is wearing thigh highs or socks, also when trying to do a breast grab, the fingers look like they are attached to the breasts. Even with very detailed prompting i could only resolve these issues by luck and continuing to prompt it would get bad again
Try different scheduler/sampler combo. Or try consistency fix (https://civitai.com/models/567975?modelVersionId=1939571). Start small with that fix.
https://civitai.com/models/1519509?modelVersionId=1723816
Also option, author has all-in-one LoRA as well. In this case, it works more on CLIP part less on image part, so you need to have option to use LoRA weight both for model and clip separated. You can use even model weight on LoRA set to 0.0 and CLIP at something like 0.3 - 0.4.
ComfyUI related, in other software you are probably stuffed.
How do I fix? Ultra white foreheads? I have tried everything i can think of... help
Have you tried adding prompts for hair bangs? Like long bangs or something.
This is great, as a begginer i'm having a lot of fun.
Great generator.
No commercial use?
excellent model, nothing superfluous
Do I have to use "by" when using "artist"? I've used it before and after, but they're so similar it's hard to tell the difference. Some artists can't even find peace.
How can I remove the reflective effect from the character's skin?
try shiny skin in negative
There's a matte skin lora: https://civitai.com/models/1105924
clueless_engineer Thank you very much, it really works very well.
clueless_engineer Thank you very much, it really works very well.
I really enjoy this gen but it can be challenging to get darker skin tones especially with multiple people in the prompt
v5 released! Looking forward to what everybody creates 😍
I'm creating a reminder to come back in two weeks.
Seems like I can not join the discord with the links, can someone share the character list please
I can send you over discord by DM, just DM me first @JAN
Isn't it oversaturated?
share image link that has prompt and resources for help
janxd Well, I mean your showcase pictures for 4.0 and 5.0 are oversaturated too
The colors look fine for me and it is color corrected based on my liking. You find images from other users in the gallery that don't use the same color correction as me.
For some reason using lazyneg makes the color looks more "cooked", i cant explain it lol, but it looks like it has oil on the image and the color is more satured
are you using the onsite gen? use cfg 3
also try an older version of lazyneg. I will do some testing with onsite gen later today to see if I can figure out the issue
Seems to be a bug with the onsite generation, nothing can be changed in the lazyneg that will fix it. Lowering the CFG to 3 should help.
Yes, happens only using onsite generation, i will try CFG 3 thx for the response!
CooperL I have updated lazyneg, please use v3.1 for onsite gen. It should work at normal settings. I did a bunch of images at CFG 5 with no issues.
Thanks for bringing this to my attention. There are others but we couldn't find which embedding is causing it. Appreciate you a lot!
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.