Join The Tinkerer on Whop. Membership gets you early releases, private tools and a bunch of extra stuff.
👉 Join on Whop
💬 Join the community for support, free tools and early news on Discord
This is the model you reach for when you want super clean, photo-realistic results.
It’s sharp, powerful, and somehow just nails gorgeous images every single time.
⚙️ Recommended Settings
Sampling method: DPM++ 2M SDE Karras
VAE: is already Baked In
Sampling steps: 30+ Steps
Resolution: 832x1216 / 896x1152
CFG: 3-5
Clip Skip: 1
Upscale: 1.5 upscale (0.3 - 0.35 denoise)
Upscalers: 4x_NickelbackFS_72000_G / 4x_foolhardy_Remacri / 4x_NMKD-Siax_200k⚙️ Recommended Settings DMD2
Sampler: LCM
Schedule Type: Exponential
Sampling Steps: 8/12 (12 for extra detail)
CFG Scale: 1-1.5
⚡ These settings are not the same as the standard SDXL workflow — be sure to adjust your pipeline!💡 What’s the difference between FP16 and FP32 checkpoints?
FP16 (half precision) uses 16-bit floating-point weights. It’s smaller in size, loads faster, and uses less VRAM - perfect for normal image generation. You won’t notice any visual difference for inference, so it’s ideal for everyday use.
FP32 (full precision) uses 32-bit weights, which means it’s larger and heavier on resources. It doesn’t improve generation quality, but it’s better if you’re training, merging, or fine-tuning models, since it keeps the exact weight data without rounding loss.
🧠 Prompting
Example negative prompt:
cartoon, illustration, anime, painting, CGI, 3D render, low quality, watermark, logo, labelor use CyberRealistic Negative
✨ Optional Adetailer
Adetailer model: face_yolov9c.pt
If you only want the main face being refined set 'Mask only the top k largest' to 1.🔗 Links
Backup location: HuggingFace
☕ Fuel the Project
If this model gave you stellar results, consider buying me a coffee. Or two. Or, you know… a GPU.
⚠️ Heads-Up
This model can make mature or sensitive content. You’re fully responsible for what you create, so use it responsibly (or if you can’t resist, at least be artistically irresponsible).
💬 Want Better Prompts?
Use this custom ChatGPT to auto-generate polished prompts for CyberRealistic XL:
🔗 [Try it now on ChatGPT]
Description
V1.0 with VAE included. Without VAE will be uploaded to huggingface.
FAQ
Comments (20)
Dude, please, make a inpaint version!
be friendly my boddy !
what should be so much better in this inpaint model?
jugger and yamer already exist as inpaint ^^
I suggest you to make your own one (you can find tutorial on this site), or use Fooocus which allows you to use any no inpaint checkpoint in inpaint.
a hint
leave the negprompt blank ^^
try it ...
only use if somewhat you whana dont see
that's what you get when you're only focused on the technical aspect; then you're not so great with prompts. Nice tip!
Love it, this may be my new fav (forced)nsfw sfw model. Surprisingly it is absolutely amazing with NSFW, but it requires nsfw loras to get the most out of nsfw, luckily the model is extremely and surprisingly Lora friendly. I've downloaded at least 20 sfw models and they all refuse to cooperate with nsfw loras, and I mean absolutely refuses; mutations/deformity.
In fact, out of like 30 txt2img generations I don't think I've seen a single wild deformation(except for fingers, nobody gets that right, its just a problem that plagues stable diffusion entirely), and I do strictly full body shots. This model is perfectly fine tuned.
Thank you for your message. Yes, it did take some time to get it to where it is now. I had some personal criteria it needed to meet. Hands are still a bit of a challenge; I've been working on getting them right for a while now. Hopefully in the next version.
Got link to nsfw loras? I see a bunch stating that but which do you guys mean?
Hands down the best SDXL checkpoint, as I said in my review. Incredible for a version 1.0. All the images I've posted are raw outputs without any additional fixes or Lora.
Thank you for your compliment! For you, it's version 1.0, but I think I've already made about 35 versions over the past few months :)
@Cyberdelia Ah, that explains why it feels so refined and clean. I get very few errors or glitched poses.
@Beethy Awesome images you made btw!
Amazing Checkpoint.
I would love to see it getting expanded into more fantasy/cartoon/concept-art capabilities. (Although it already has them)
Very Style-LoRa friendly!!! – I can't wait for the next iteration! Congrats!
And, maybe just don't focus on NSFW but on great artistry instead. So much potential!
Don't fix something that ain't broken. Just use lora's for your "fantasy/cartoon/concept-art". It works very well, I have tried it on over 145 lora's. Each time perfect.
One again Im stunned by how LORA compliant/compatible this cyberrealistic checkpoint is. Almost every other sdxl lora fails at the accuracy of a lora except for this one. Ive tried like 8 checkpoints with the "ahegao lora" and only Cyberrealistic XL got it right without fail. Even Juggernaut didnt get it right. Truly hands down the best fined tuned realistic XL model to date. People are sleeping on this.
I wanted to test this out for a couple of days before I left a comment, but... man, this has to be the best photorealistic model on the site right now. It does everything I throw at it flawlessly, integrates LORAs with exceptional ease, and does a decent go at hands and basic anatomy with less prompting than any other model. I've toyed around with local models for over a year now, and this was the first model that made me leave a review, post pictures, and leave a comment. It's well deserved in my opinion. I hope this model gets a lot more attention, because it deserves some huge hype! I can't wait to see where you take it next.
hey,
you have any idea wh i cant make an inpaint model of this
error: Bad dimensions for merged layer
and
list' object has no attribute
it happend on new models since 2 weeks
jugger works fine ;)
I dont know if its just me, but the model seems to white wash lora models even if i prompt for dark skin.
This is the one
You made me try SDXL and I find the images to look much more natural than with the SD 1.5 model. The variety of the images is much higher also and it doesn't have that perfect look that makes the images all plastic. This model is also able to do cartoon characters with realistic shapes which I like. I have a great time with model, thanks a lot!
By the way, whoever said SDXL is bad at NSFW doesn't know ****. I feel my SDXL Lycoris models have super powers compared to my SD 1.5 models, they are much more flexible yet have the same training parameters.