Join The Tinkerer on Whop to get early releases, private pages, and the Tinkerer Discord role - all in one place. š Join on Whop
CyberRealistic is a flexible, easy-to-use photorealistic model built from nonstop testing, custom blends, and just the right amount of chaos. The backstory? Itās honestly a wild mashup of different checkpoints, but what really counts is what you get in the end: sharp, expressive, clean renders that just work.
Itās tuned for both textual inversion and LoRA, so itās great for anyone from total beginners to hardcore prompt wizards. If youāre making portraits, messing around with new styles, or just want a model that steps aside and does its job, CyberRealistic wonāt let you down.
āļø Suggested Settings
Sampling method: DPM++ SDE Karras / DPM++ 2M Karras
VAE: is already Baked In
Sampling steps: 30 Steps
Resolution: 512x768
CFG: 5
Upscale: 2x
Upscaler: 4x_NickelbackFS_72000_G
Denoising strength: 0.3š§ Prompting
Negative prompt examples
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurryš¾ Backup & Resources
backup location: huggingface
ā Support the Project
If this model helped you hit your vision faster, cleaner, or just plain better ā consider [buying me a coffee]. It keeps the updates coming and the experiments wild.
š” Need Better Prompts?
This custom ChatGPT was made to top-tier prompts just for this model:
š [Try it now on ChatGPT]
ā ļø Friendly Warning
This model can generate mature content. Use responsibly. Respect laws, platforms, and people.
Description
Model for inpainting and outpainting.
Make sure that the model name has the ending -inpainting
FAQ
Comments (33)
v3.3 was my favourite. One of the best and it still surpasses most recent chekpoints even if it was made months ago. It's old but gold. I hope 4.1 is better. We'll try.
The version 4.1 is really good. I've done some tests and the results i've got are really impressive.
One of my best model.
Is there a 4th version of model on hugging face? thx
Not yet, will try to update is soon
Why is there two files for the model?
Pruned FP16 and FP32 models
FP stands for floating point. It is a computerās way of storing decimal numbers. Here the decimal numbers are the model weights. FP16 takes 16 bits per number and is called half precision. FP32 takes 32 bits and is called full precision.
The training data for deep learning models (such as Stable Diffusion) is pretty noisy. You rarely need a full-precision model. The extra precision just stores noise!
(V41) - WAY better! This is fantastic! Thank you sir, massively appreciated!
This is great. While requiring a little more prompting than something like epicrealism, it listens to instructions well and gives a wide range of different faces. Bravo
As one of the people who originally complained about the previous release, I'd like to say that 4.1 is INSANELY BETTER. It's great work and truly feels like an evolution. Congrats on the hard work and thank you for listening to the users.
Thank you for the compliment, it is greatly appreciated.
For the 3.3 version, both the standard and inpainting models have 5GB versions, but for 4.1 the files are both smaller. Why is that? Are they less capable?
No, they are pruned. The results of pruned and non-pruned are the same.
Technical article: https://medium.com/@souvik.paul01/pruning-in-deep-learning-models-1067a19acd89
Can you make LCM model ?
Hi all!!! Who knows why images generated using Automatic 1111 are very different from those generated by civitai with the same settings... Thank you very much for your answer.
There could be plenty of reasons.
1. Maybe the civitai image is missing a lora or embedding that wasn't listed. Sometimes I try to copy an image and the lora listed on the civitai image prompt no longer exist on site.
2. People could have used a1111 to create the image, using all manners of extensions (Adetailer, ControlNet, LatentCouple, etc) that isn't written into the image png. You can try downloading that image and drag that png image into the a111 'PNG Info' tab to see what could be missing, though that isn't perfect. There could be a better way to find the setting used to that image creation, but that beyond my know how.
3. Some people generate using comfyui instead of a111, allowing much much more control over the image creation that can't be done on a111 or civitai.
All that being said, take everything I said with a pinch of salt as I am still learning the ropes on all this.
As commented, but also, many people like to edit their prompts if they made something special, to keep their magic to themselves.
I'm a newbie sorry for asking: does it need any hires/fix?
From my experience, it saves a lot of time to ignore hires/fix and upscale the image after it is generated. I use the Ultimate SD Upscaler extension.
@makkuraĀ thank you so much
@makkuraĀ or generate without hires and when You have nice photo You can take seed and generate again with hires. After many testing it has better quality than upscaling later in img2img
I love your model, just gave it some buzz! Could you make a version of your model that is more prone to fantasy elements, such as magic and armor, and less NSFW? The model is very capable at the moment, but it doesn't recognize well fantasy elements, and tends to make partial armor (with legs exposed and such).
Keep it up!
Thanks for the buzz - appreciate it. I understand your question, but it's difficult to create a model which has everything (that's why Loras are great). 2.5D type of models are better doning this kind of stuff.
@CyberdeliaĀ I've never found a Lora that does female armor that isn't skimpy or missing parts. If you know one or can point me towards the place I can find one, please let me know! I promise I'll make it worth your while!
@RheisSivviĀ Late to the game, but try prompting for exactly the armor you want and incresing the weight, and also, in negative, prompt things like skimpy, bellybutton and such. It usually works for me to avoid the chainmail bikini.
@tvange365 Thank you for your reply! I'm still struggling with it, but I'll try more modular and descriptive prompts like you suggested. I'll post my results soon!
GJ,thanks for sharing, I'm a newbie and this might be a stupid question: do I have to use the inpainting version when inpainting? What features does the inpainting version have?
No, personally I don't use the inpaint version :) You could have better results - so if it doesn't work with the base version try the inpainting version.
Really good Model, I have been using the unpruned Version to upscale SDXL gens with a controlnet guidance with excellent results.
The CyberRealistic SDXL version is going to be released soon!
ooo Very nice! Take your time.
Thank you Very Much for this stellar work! Really turns out well
You should like win an award. The cyberrealistic models are like the best.
Details
Files
cyberrealistic_v41-inpainting.safetensors
Mirrors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors
cyberrealistic_v41-inpainting.safetensors