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
This is version 2.4 (internal) with some additional training from version 3.0. There were some issues with version 3.0.
FAQ
Comments (26)
v3.1-SDXL
9,5/10
:)))
the score is going up, well done Cyberdelia.
Использую FOOCUS , и модифицированный мной dpmpp_2m_sde_gpu
It's strange how underrated is this checkpoint. My theory is that thematically NSFW models seems to have all the attention these days on the SDXL sphere (unlike CyberrealisticXL which is more like an all rounder).
Whoever interested in nsfw, you can merge this checkpoint with Lustify v.4 and you'll get the best one of the best NSFW checkpoint that exists so far for SDXL. You get the unmatched image quality of Cyberrealistic and the flexibility of Lustify.
Back to the subject of the improvements in the latest installment of Cyberrealistic: eyes and faces have improved dramatically. All in all, It's a really good and firm step in the right direction. It keeps evolving. So great work!
I've already said it, but I love the aesthetics. I love the light/ambiance (it's one of the few checkpoints that gets rid of the "white light" predominance so intrinsic to SDXL platform which i personally find disgusting and nausea inducing). And i love the artistic vision behind Cyberrealistic checkpoints which are truly a work of beauty.
Hats off to Cyberdelia, one more time.
Yeah, its underrated, because cant do porn. The cyberpony latest versions practically supersede the SDXL versions.
@valyruca128 3.1 can do NSFW/porn :)
@Cyberdelia Really? Anyway i moved on Pony. I wont go back. 3.1 can do effortless sex poses like your pony? Only some handful XXX checkpoint could do this in SDXL.
Im waiting for your next Pony versions:)
Edit: if its true, you should advertise as the most beautiful and lora compatible XXX checkpoint. You should put sex scenes to the main sample pics.
JFC - with the new NSFW in 3.0+ and this little light bulb... mind blown
@valyruca128 I don’t really want to post NSFW images. I understand that most users want to create these kinds of images, but I’m trying to provide as many SFW examples as possible.
@Cyberdelia totally understandable. Moreover, I'd do exactly the same and its better this way, honestly. With my previous comment i just was trying to pinpoint why the checkpoint was so underrated -considering the quality which is top tier- in my opinion. I wasn't trying to stir any controversy up. Looking forward to new installments of CyberrealisticXL. Best regards.
@ElectricDreams oh I know :) And also I don't have anything against NSFW images ;)
Absolutely. It's esthetics and lighting is competiting or even better than JibMixXL and RealVisXL but at the same time it can do NSFW and has a really good skin texture. Faces are a tiny bit behind these two but ADetailer will help there. On the way to be my goto checkpoint.
Hey, will CyberRealistic also be available as SD 3.5 (Large, Large Turbo or Medium)? LG Dominik Goyke
Really great skin detail and face variety in 3.1!
supr model, mb lighting/hyper version?
Since the community seems only interested in Flux cuz of extremely easy Lora making, and also on Pony.
Any chance we can see real Fine Tuning of SD 3.5 by Cyberrealistic?
Hi guys, when i try to downmload this to google colab it says username and password failed?? Here the code it returns in colab? Any ideas please?
bro .. are you SUPER GENIUS or a MAGICIEN ??? V3.1 is the supernatural SDXL discovered !! thank you BRO. works perfectly with ControlNet .. please make V3.2 or other .. and thank you very very much
Totally agree. It's amazing, also with ControlNet.
Day 354... still waiting the new installment of CyberrealisticXL...
Just kidding. But in all seriousness i hope Cyberdelia is still working on it or has plans in the future to keep this series going on. Still the best allrounder there is (for XL) and it consistenly gives quality outputs whilst respecting loras fidelity. Image Quality wise, Cyberrerealistic has always been top tier.
Landscapes, both photos and illustrations generate very interesting and not boring. Photorealism of people, especially with faceid is amazing. Ideal model CyberRealistic XL v3.1, we are waiting for the continuation, please do not abandon such a successful model.
Hope you also like the latest version
@Cyberdelia Many thanks! 4.0 met all expectations. I noticed that for landscapes when using euler + align your steps gives a spatial perspective, better than dpm++. In terms of detailing with different samplers, there is no ready-made recipe, each plot is unique and you can find one that suits better. Interesting model, new light, new details!
I gave it thumbs up because the Pony one excellent, but this version is so blurry with around 42+ steps. I'm using the instructed negatives, and no matter how complex or simple the prompt gets, it's completely blurry.
works like a charm
Update incoming! ;)
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.