🌐 Online generation available at Shakker.ai & SeaArt
✨ VERSION 2.0 · REWORK
Version 2.0 is a complete rework. This version addresses some issues from previous versions, offering cleaner image quality, enhanced overall realism, and noticeable improvements in facial features.
However, the use of Adetailer or Face Fix is still highly recommended.
🏠 Recommended Parameters for Local Generation
Samplers:
Recommended:
DPM++ SDE
DPM++ 2M SDE
DPM++ 2S a
DPM2 a
DPM++ 3M *
Euler A
Settings:
Steps: 25-35
CFG: 4-5
Clip Skip: 1
Resolution: >1024
Additional Information:
Euler A provides a soft and cleaner appearance, while SDE samplers produce a more detailed and sharper look. Experiment with different samplers to achieve your desired look
Use Danbooru tags.
Prompt Style:
Positive:
masterpiece, best quality, amazing quality, realisticNegative:
3d, anime, 2d, bad quality, worst quality, worst detail, sketch, censor, simple background
🏋️♀️ Training LoRas
Read the following article for tips and my training preset
🌞 Final Remarks
This will be an ongoing development aimed at improving the realism and various aspects of the model, so feel free to share your feedback.
Thank you for downloading and using the model. I hope you enjoy it!
🪄 Workflow
⚡️ Buzz for the Best Images ⚡️
Description
FAQ
Comments (48)
Please upload it on tensor art, i would be very grateful
What's new in v1.1 and how are you finding training Illustrious vs Pony?
v1.1 has better quality, more realism, and nicer subtle details.
From what I’ve seen so far during training, it’s harder to overtrain or mess up the output. I’m really liking how it performs.
Holy shit, this just... WORKS! Beautiful and easy to use!
Very nice model.
I am concerned about a lot of mysterious wrinkles around the waist and the rib bones that tend to come out.
It would be nice if the face was a bit more stable, even though it is an Adetailer premise.
Yeah, I’m trying to fix the asian bias and improve the faces, but since it’s so heavily trained on anime, it’s tough. Hoping I can make some progress on that
@ZyloO How do you train your models? Real life data doesnt help? I mean real life girls and sex.
Very promising model, waiting for new versions!
Just don't use "smug", for your own sanity. Faces are... eh... yeah.
Beautiful
I'm just getting noise for ou tput in ComfyUI if I try to generate any of the images I've found for this model. If I set clip skip to 2 I get an image that's similar but worse quality. Don't suppose anyone has any suggestions?
Add this to civitai generator please!
I already have two models on the generator, and I don’t think I can add more. I’d love to, but it’s not up to me. Perhaps if the model becomes popular and more users request it, things might change
@ZyloO I look forward to it! Because when it comes to realistic models, I only use yours because they are quite reliable!
@ZyloO In fact, you can replace this with one of realism pony. If you like, you can put down the lower version of realism pony. If you contact the authorities, they should do so.
Although the details are richer and more realistic than Pony's, the composition and aesthetic appeal are much poorer.
This is one of the first birds. We will see, whats the situation in half year. First realistic Pony models sucked too. Today, there are very good ones.
I disagree. I have equal or better composition and better prompt adherance, compared to the realistic pony models. Aesthetics depend, maybe slightly better in pony. You might want to look into perturbed attention guidance, it helps a lot with composition variety in my experience, decrease CFG when using it and use a high s2 in freeu @yuinyan490363
The eyes don't look good in the drawing.
Possible to escape from Pony style ike face ? I tried to describe face features and like small eyes, droopy eyes, small mouth, small nose,etc. but nothing works.
What kind of LORAs can be used with this model?
I've used this checkpoint on Tensor.Art and SeaArt.Ai with Pony and SDXL LoRAs
Isn't this supposed to be for Illustrious Loras as well (and mainly)?
I'm kinda confused here.
Why is this only capable of making ONLY woman?
Simplest explanation: Men are horny.
@Polygon Yes, i know me too, siri show me naked men
can you petition the mods to allow your checkpoint to be accessible by the onsite generator? Thank you. I know the rule of only 10.
I did, but I can only have two models on the generator, and I currently have two versions of Pony Realism. I don’t want to cut access to either of them, as many people still use them. In the future, when the site increases the model cap, I will request it again
@ZyloO I thought the creators were allowed 10 models
Hello :)
Can we use Pony Loras with your model?
Some of them work, while others only work to an extent. You will likely need to test them
@ZyloO Oh, I understand. So, what recommendation would you give me for using this model with self-trained LoRAs? Should I train new LoRAs based on IL to work with this model, or use the standard XL? Or perhaps train with special parameters based on Pony?
Thanks for responding :)
Just wanted to say I love your model
First I skipped this one, but now I tested it again, but now for img2img.... excellent, now I love this model
Did you use controlnet model?
@fengyun321 I don't need one in my usecase, I'm doing img2img for anime style images, to turn them realistic. It's still not that realistic, but I enhance it further with other models. The more important for me is that pony tend to change image a lot for img2img, but this model stays closer to source.
@sutikkukun Thank you for your answer, I will give it a try.
I hope for new release based on NoobAI model, especially vpred. It already has some cosplay knowledge.
Good job!
illu realism? YOOOOO
Illustrious but realistic? Your techno-sorcery is cutting edge.
Great checkpoint, a little bit tricky to work with LORA's but final effect is worth the effort
Are you using Pony loras or SDXL loras? And what is the trick you use?
@ronaldmikhailp236 Mostly Illustrious/SDXL ones, rarely Pony ones work. With that in mind, I set LORA's values <0.5, high values tend to make images look odd.
@xtcONE Thanks for the info brother. May all your generations be immaculate
@ronaldmikhailp236 thank you G! Have a good one!
Really good prompt adherence.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.