UPDATE v6.5:
This update brings small improvements :
Better logical composition, a bit less body fusion and mutation, better prompt following, a bit more creative.
For the style, I redirect it to a more amine style, I hope you find it nice :).
Cons : the model if not specified generate teenage girl or boy, so you need to precise the age of your character.
MISCELLANEOUS:
Rules & Respect :
1. Do not spam similar images in the gallery.
2. No hardcore, disgusting or violent sex image.
3. No, low quality or effort image.
Image Gallery :
1. The best image will be pinned on top of the gallery.
2. Creative image are welcome.
Thanks:
Hope you like this model and that you show me your images.
All the buzz that I gain from Model-EX is spent in the gallery.
Thanks to everyone and feel free to leave a comment. :)
MODEL INFO:
Here are the settings needed for this Model:
Clip Skip: 2 / Nothing Else
Steps: 30 / Recommended
CFG: 2 / Recommended
Sampler: Euler / Recommended
Scheduler: Exponential / Recommended
Description
Comparing v1.5 to 1.0 :
55% Better Background
20% More Creative
40% Better Light Logic
30% More Capable
60% More Depth Comprehension
50% Better hands and Toes
Better at large angle or wide scene
40% Better human Anatomy
Abnormal consistency of insane image
Easy to Controle Without Lossing The Model Creativity
Insane For City, Nature Environments (Whatever Angle)
FAQ
Comments (41)
Version 6.0 Sounds extremely promising just by the information you shared. Im gonna try it right away. Probably could look repetitive but ill use the same images i did the first times, to Compare the changes in version :3. Ty for your work! :D
Share them here to, I want to see them :)
@Nico53425342 For sure :D
v6 is absolutely fantastic! Probably the best illustrious model so far.Small flexibility with artist styles is the only downside i found so far. Great anatomy is probably the biggest strengh of this model
Thanks dude, if you find anything else tell me and I will try to fix it for next version.
But the Inflexibility with artist style is quite normal, it's mostly train on danboru tags and natural language, so artist style tags are almost inexistent, try using lora or use strong negative to deviate the style. :)
@Nico53425342 oh that explains a lot :D. Sadly after some extensive usage i found another issue. Model seems to have problems with generating varied outputs with the same prompt. What i mean is when i generate let's say 10 outputs from a single prompt they all look almost the same while other models generate varied and different results. It also depends on the prompt but so far it happenes around 70% of time
@DevilLady that not a bad news that, a fantastic one. That means the model understand the prompt and output what it think even if you change the seed (his mind) that mean the model is focused more on the prompt than creativity, it can be bad in some edges cases.
Hot western men possible for creation? Gay sex, anal penetration etc?
Wow wow wow, hot men not a problem, gay sex big problem, anal very big problem.
Keep things sane dude, I hope my model cant do that, just do normal hetero stuff...
@Nico53425342 I think you must be joking. I like your model and in general it performs well. The issue I have with it is when I am trying to use more than one male character in an image. If the dudes end up close to each other in the image, a lot of times they are touching each other, or trying to make out, sometimes a lot more than that even. Like you must have done something to the weights to make this model behave that way. How do I make it stop?
@DevilDogAV man, you need to fix your prompt, because I don't get those kinds of problem. The model respect very well your prompt, so if the context points to something else than sex, it won't do sex.
I use Comfy UI, so I know exactly how my model will perform, but for UI stuff like forge I have no idea of what is happening in the background.
Try removing in your prompt everything that points to gay stuff and precise the none gay image that you want in more details.
Give me feed back if you come across other issue.
@Nico53425342 It's cool, I was just trolling you. I didn't like your response to the OP. It's disappointing to see public comments like that. Your model is actually a bit resistant to making male characters at all. But that is related to the general female bias of Stable Diffusion itself. Just relax. There's no need to degrade someone's sexuality because it's not exactly like yours.
@DevilDogAV I'm confused, I have a hard time understanding what you understand.
"I didn't like your response to the OP" I'm the OP, how can I respond to myself ?.
I'm not a fan of gay or girl with dicks or anal stuff, so that why I respond like that to @zeomeow.
It is hard to express myself though text, so I never mean harm to anyone, and if it's seen as it, it should be taken as a joke.
I think there is a misunderstanding between us, please clarify for me what you mean :).
@Nico53425342 I meant the original poster (OP) of the question, the question that created this thread, zeomeow.
You are right, it can be difficult to properly express ourselves through text because it lacks the information we exchange through tone of voice and facial expression. I accept that you meant it as a joke. Thank you for clarifying and taking the time to respond.
I will take a crack at answering the original question.
@zeomeow Yes, this model can draw those things but it will take considerable weighting adjustments to the prompt. You will likely need "male_focus" and "yaoi" or "bara" included in the positive prompt and turn their strength up to 1.2 - 1.5. You will also likely need "1girl" and/or "pussy" and/or "breasts" in the negative prompt and turned up to 1.2 - 1.5 strength.
If FP32 is too slow for you, you can try to load it at FP16 There is no difference in quality, and it will x2.2 speed up. With FP16, I'm getting (1 step/s).
I recommend the usage of COMFY-UI to generate. With an extra Node called Self-Attention-Guidance.
It will enhance your image tremendously.
I don't recommend any other app to generate because you never know how and who is cooking you food.
mind sharing that comfy workflow? ^^
@TheQuacktastic Look at the suggestion list i just created a new post, its my current workflow
@TheQuacktastic ah and if you decide to upgrade the workflow with different parts share it with me it will be a pleasure
Hello!, ive always been using stable diffusion reforge (i think), i have heard of COMFY-UI, how heavy is in requirement to be use?, it requires more than Stable diffusion and is too hard to set?. Only wondering cause my PC is a little old in terms of Memory and CPU so if COMFY requires more ill probably wont be able to run it haha.
@Razuryam i don't think you should have any issues if you are able to run it on forge, i'd recommend trying it out
@TheQuacktastic Oh Really? ;o. Thank you so much for letting me now, ill try to watch a video for the install and try :3
@Razuryam Comfy is lighter than any other generation UI
@Nico53425342 oh really?, thats even better then lol
the prompt adherence is nuts, good job dude
yeah its nuts, i cannot comprehend
Found this model on the dreamerland AI tool. Can I use this model on dreamerland to generate images for commercial purposes?
sure, do whatever you want with it
Thank you so much. 😀
Surprisingly, this is probably the most docile XL model I've come across, keep up the good work. I think it's a little lacking in variety in drawing men, but I guess I'm just used to other models usually drawing randomly, I don't even know if that's a minus or a plus :)
Yeah, men are hard to draw because everything is overtrained with girls so they overshadow men's by far, there is nothing I can do about that, you should just use dark tweaks on your prompt...
A very beautiful, unique model. Always enjoyable in a new way
One of the best models I've tried, surprised this isn't as popular as others yet
Thanks man appreciate it
im so glad you made a model for ILL and that i randomly stumbled upon it. ex model has always been my favorite then i swapped to ill and was a lil sad ahahah
Thanks man for the compliment, it makes my day
Life tough dude hahaha. Just download the image and put it in a image2prompt and you will get a better prompt than what was used to generate it
hello! i really liked the model you created. can you share an example positive and negative prompt for me? thanks!
i Only do Positive, never negative here is what im using at the moment : Artwork, illustration, Masterpiece, Motion Lines, Body Movements, Dynamic Motion, Motion Blur, Colorful, Best Quality, Good Quality, Newest, Very Aesthetic, Absurdres, Vibrant Colors, Mesmerizing, Amazing Quality,
@Nico53425342 thank you so much for the reply!