Thank you for the model!

cant replicate images in preview , tons of loras i cant find

By far the best realistic base I've been able to find so far!

Pony version?

LOL I can't generate male characters

@ponyboi can you re-upload the spitroast pics for this model from 2 days ago? Or share the prompt? 🙏

Hii do u have a workflow for comfyUI? do u recommend any loras for the most realism?

This is my favorite new checkpoint but I wish it had more poses and positions, it seems to tend to default to a couple poses very insistently under constraint. Any suggestions to counteract this? Otherwise I hope you'll add a bit of variety next version!

Still learning but I am enjoying using this model the most, so far.

I have a noob question, though. Is there a way to comment out stuff in your prompts (i.e. tag it in some way so that it isn't used in the prompt). I just want to add some reminders and stuff.

Thank you for the model and thanks for help!

This produces some really great results and I like it a lot. But I'm also struggling to get it to obey my prompts. The subject always wants to look at the viewer despite prompting otherwise. I haven't had this issue with other Illustrious checkpoints (like CyberIllustrious).

Wow the generations look insanely realistic. Well done!

Sometimes you need to give a chkpt a second chance. I'm glad I did here. Getting very good results using free Google colab and forge. I'm not even needing an img to img upscale or hires fix, just a standard portrait or landscape with adetailer and one realism or detail lora (pony, sdxl or Illustrious bring different and good results).

Getting some really good results in some situations. But I am struggling to get it to produce the following: cunnilingus, female masturbation, any kind of dildo self play. Any tips or is it just not trained for that?

Requirements?

What are the requirements for this model?

The same issues like many others. Very good Pictures, very real, but I have the biggest problems to get my model look away: I havn´t any pic with full body and from behind. Do you have tipps for prompts to get full body shots, shots from behind and to make the model look siedways? Because I really want to use your checkpoint as it is very beautiful

คมๆ

Ngl, great results, but this model doesn't give a fuck about your prompts AT ALL.

只能生产欧美脸型 亚洲脸型基本出不来

great work I've been using it with sdxl vae.

adhering to simple and natural prompts makes it easily one of my top 3 favorite models.

sometimes, simple = more.

Thank you

Fantasy was trained out of that model, too realistic for me...

My current favorite model for inpainting boobs. Most models don't pay much more attention then "boobs" but you can get crazy detailed with your description and it will try it's best and usually succeed. Much better than stand alone boob models I've used. Only minor gripe is it sometimes ignores the obvious size of the covered boobs and and will make your Double D's into A's or vice versa and when you try and crrect this on the next pass, it will interpret your "medium sized breasts" prompt as "(Mammoth colossal supersized breasts:1.5) but it's beautiful 86.7% of the time

Great work! Love this model and your loras!

Where is solid generations amazed by photo quality?

Whenever I generate with this model, I get these white pearl-like artefacts throughout the image.

Aaaaaand now we have to bid to get it back in the generator. Wonderful.

Ok, this is actually one of the better il realism models, usually they look too fake, or do the speckles all over the place. My only problem with it, it struggles with older and plain looking women.

best pictures i produced yet, i'm getting very good results at 1200x1600

excellent model, really like it alot! Works well!

great model, lloking very good after good prompt, excellent job

so lovely

Very good checkpoint for realism ,excellent! Keep it up! <3

This model can do night or dark lighting, but with not-booru tags - by_night in positive and daytime in negative.

Great model! Why was onsite generation disabled?

The best of the best. Excellent response to prompts. Works great with the lors I've trained. A beautiful model for my creative ideas. Thanks a lot.

a good fit with your realism slider, also works with real skin with a bit of tweaking. My current go to since it does a good job balancing between sensual and photo-realistic without that uncanny or exaggerated quality other photo-real illustrious models tend toward

A great model for beautiful creations! Stunning faces! One of my favorites! Works very well. Good job!

How does one reduce the size of the ENORMOUS BAZOOMS in this model. WOO WEE F CUP CITY

LOVE it! I really appreciate that, as a newbie to ai image generation, this is the first model I've found (after trying several) that delivers extremely high quality realistic images without requiring novel-length, contorted prompts and tons of after-processing. So far it also responds very accurately to my posing prompts

Is this model also good at generating Asian faces?Cuz I dont see any asian face in gallery

How about training Asian models for similarity?

Which scheduler do you recommend for local generation?

why can't this be downloaded?

thats a lot of freckles

Very forgiving model. It can take a lot of different sampler/scheduler combos, and a wide range of cfg/steps.
You have done a solid job on the stability of this model.

no workflow, how would i know the requirements..

I wanna say, this is a pretty nsfw model, I don't mind it though, it's just funny I'm getting sexual scenes even though I didn't prompt for it.

Why this model says base model Illustrious, yet no IL loras can be used with it and everyone uses Pony loras?

Great model for realism, after making some changes to my base prompt for regular illustrious models it worked perfect, however pro tip it's still an illustrious model which mean a strong attachment to danbooru tags and the model creator suggest deformed:1.3 in the negative prompt which... deformed doesn't mean the dictionary meaning of "deformed" in danbooru tags, rather the anime deformed, which would be addressed by the "anime" tag, instead I suggest something like: abomination, corrupted anatomy, or artistic error. or all 3, anatomical nonsense doesn't ever seem to work well for me, and leave it unweighted

My personal negative is "(abomination:1.3).

(worst quality, distorted details, medium quality, bad details, lowres:1.2).

(blurry, artistic error, stretched, disordered:1.1)

(corrupted anatomy, ugly, unattractive, broken, damaged, bad textures),

baby, toddler, big belly, tanlines, fat rolls, cartoon, 3d, anime, doll, artificial, painting_\(object\), multiple views, long legs, long arms, film grain, bad hands, bad feet, bad symmetry, big feet, "

The last section you might want to have a look at for personal taste, also stretched may be a case by case basis if you want something stretched like tight clothing, mostly danbooru tags or at least borrow from words from danbooru tags to get a more targeted effect.

Hello, thanks for your great effort. Can it be used with Controlnet? I couldn't get it to work. I tried many workflows. Either the pose is okay but the image is bad, or the image is good but the pose is not taken into account. If you know of a workflow or control that works, could you share it? And yes, I'm a noob :)

I'm sure I'm doing something wrong lol. I put in light a lesbian scene and I keep getting a woman having anal/vaginal intercourse with a man.

Very good model, and several problems:

1, Don't follow prompts, this might even be the problem of SDXL

2, Tend to generate nude and nsfw images, even with stong neg prompt trying to prevent.

Thank you, likely a stupid question but I am still dumb and new :D, the workflow I have requires a VAE, but this is baked in. Do i bypass vae or just use it as normal?

Have to just stop, and say, Thank you so much for making this checkpoint. Absolutely beautiful results. One can only hope this staves off the world from leaving their bedrooms/computer.
World Peace. Not a bad accomplishment.
Wonder how many we can attribute to this checkpoint as a life spared by making some dude too tired to leave his bed.

This is my new favorite realistic fine tune for Illustrious. Good facial variation, lighting and detail, with some banding on skin depending on the prompt. No anime bleed over which can be an issue, especially for strange eye shapes that happen even with CyberRealistic Illustrious (my second favorite). The model is well behaved too, following prompts and not adding unwanted details such as automatic high heels and garish nail polish when using with certain… other keywords. Feet and fingers might still get mangled, but overall it is quite naturalistic.

I get really bad anatomy. I use forge UI and used the described settings, but the bodies look either wierd or have bad anatomy.

OK definitely like and gives a good effect when mixed with anime style prompts