CivArchive
    Preview 8048126
    Preview 8048128
    Preview 8048124
    Preview 8048121
    Preview 8048120
    Preview 8048135
    Preview 8048125
    Preview 8048122
    Preview 8048119
    Preview 8048127
    Preview 8048130
    Preview 8048133
    Preview 8048132
    Preview 8048601

    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, label

    or 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

    Some slight enhancements in hands etc. (though not flawless), along with various other minor and more significant improvements.

    FAQ

    Comments (30)

    qtpii328107Mar 16, 2024¡ 3 reactions
    CivitAI

    Excellent

    JimsmithwickMar 16, 2024¡ 1 reaction
    CivitAI

    Model download hash mismatch, CivitAI gave E9337AE032D6 but download gave E9337AE032D6A2D411702FC6ACAD76BFBDDDEB1A697A62BFDB83881E949532C8.

    Any ideas?

    Cyberdelia
    Author
    Mar 16, 2024¡ 2 reactions

    I've seen on Discord that there's a bug, and they're working on fixing it.
    "SHA256 hashes fix incoming, models uploaded this morning with the short/invalid SHA256 will be rescanned."

    JimsmithwickMar 16, 2024

    @Cyberdelia thanks for the reply!

    yofoton174609Mar 23, 2024¡ 1 reaction
    CivitAI

    well done moving to SDXL. SDXL is better than stable cascade and anything else in my opinion, since there are more LORAS to SDXL, and 1024p is enough for any kind of upscaling. Perhaps try a "SDXL Lightning" version next? (sampler:euler&scheduler:sgm_uniform, cfg:1, 4steps)

    Cyberdelia
    Author
    Mar 23, 2024¡ 1 reaction

    I put so much hours in Cyberrealistic SD15 that it is difficult to move away from it. So personally I use SD15 Cyberrealistic most of the time. Still try to support/update it.

    yofoton174609Mar 27, 2024

    @Cyberdelia yeah, i know what you mean. Sometimes its simply not worth switching up to a new base-model. What do you think about SD3?

    KodexMar 26, 2024¡ 4 reactions
    CivitAI

    this makes no sense... I keep switching back to this model... What makes it so good above the mainstream? cfg high values?

    Cyberdelia
    Author
    Mar 27, 2024¡ 4 reactions

    Made with love ;)

    ObsidianDreamsMar 28, 2024
    CivitAI

    Hello! I really love this model, it's one of the best I've tried for SDXL so far. Would it be possible for you to share an unpruned version so I could train my Loras using this? Much appreciated!

    SilmasMar 30, 2024¡ 1 reaction

    I would ever train LoRAS on the base SDXL model.

    Cyberdelia
    Author
    Mar 30, 2024¡ 1 reaction

    @Silmas Correct

    ObsidianDreamsMar 31, 2024

    @Silmas I like to experiment. I've had very good results training on models other than the base model. But thank you :)

    SilmasMar 31, 2024

    @cassiebug1988 sure you can have good results, but this LoRA is maybe not usable for other models then. That is why I always train the LoRAs for the base model.

    ObsidianDreamsApr 1, 2024¡ 1 reaction

    @Silmas totally understandable, but personally as I only really use a few models anyway, I like to train on the model I plan on using with the lora. I have made plenty with the base model too that I like!

    Cyberdelia
    Author
    Apr 1, 2024¡ 1 reaction

    @cassiebug1988 I have the unpruned version (hopefully), will try to upload it on huggingface. Don't know when - I need to search my backup drive which is not here.

    ObsidianDreamsApr 1, 2024

    @Cyberdelia I appreciate that so much! No worries if you can't, and no rush. thanks again!

    FunkMonkersonApr 4, 2024¡ 1 reaction
    CivitAI

    I can’t seem to use this in the on-site generator. Is that intentional or am I just dumb?

    Cyberdelia
    Author
    Apr 5, 2024

    I really don't know.

    FunkMonkersonApr 5, 2024

    It directs me to “use on of my favorite generators” like rundiffusion or seaart.ai. But I can use your older models to generate on Civitai. Maybe Civitai doesn’t allow sdxl on-site generation yet?

    FunkMonkersonApr 5, 2024

    Also, awesome work!!

    KringZachkvar66Apr 9, 2024¡ 2 reactions
    CivitAI

    major problems with limbs, fingers, legs

    amazingbeautyMay 27, 2024¡ 2 reactions

    what's left...

    OmikonzJun 18, 2024

    @amazingbeauty suitcase

    nomad42023Jun 18, 2024

    @amazingbeauty boobs

    fablegeniusJun 20, 2024¡ 2 reactions

    what's a better checkpoint in your opinion?

    Phenix60Apr 26, 2024
    CivitAI

    around a month ago i starting using comfyui, i'm still learning thingand now i wanted to try to do outpainting with sdxl on comfyui, i haven't got great result so far with the workflow i made or maybe the model/ setting i use, then i found your 1.5 model which seem to have a normal version and inpainting version, but for sdxl there only one version, so my question is can i use it ( this model the sdxl one) for outpainting or is outpainting in sdxl not working that great or need specific workflow or model to work ?

    andymuir77760Jul 7, 2024

    I gave up trying to find a good outpainting workflow in comfy. Everything i tried seems flawed in some way or other. In the end, started using fooocus as it does outpainting very well

    sofozMay 21, 2024¡ 1 reaction
    CivitAI

    This is my favorite checkpoint so far but it doesn't perform that well when it comes to hands and fingers.. Any suggestions?

    fablegeniusJun 20, 2024

    got a suggestion for a checkpoint that does perform better in those areas?