Join The Tinkerer on Whop to download this model and many more — all for one monthly payment. Plus, get early releases, private tools and a lot more. 👉 Join on Whop
CyberRealistic Z-Image Turbo is a build based on the original Z-Image Turbo. It follows the original Z-Image Turbo settings and philosophy, with minimal intervention, exploring how well this setup translates into the CyberRealistic workflow.
⚙️ Personal Settings (Forge Neo)
or
Required Additional Files
Make sure you also have the following:
16 GB+ VRAM: qwen3_4b.safetensors
8–12 GB VRAM: qwen34bfp8_scaled.safetensors
All VRAM sizes VAE: ae.safetensors
This model is shared as-is, for users who enjoy experimenting, benchmarking, and pushing models outside the comfort zone.
Feedback, findings, and edge cases are welcome - this release exists primarily to learn from real-world usage.
Description
Join The Tinkerer on Whop to download this model and many more — all for one monthly payment. Plus, get early releases, private tools and a lot more.
👉 Join on Whop
A lot of improvements in this version - better colors, improved NSFW and overall just a step closer to perfection 😉
I’ve also included +175 sample images showcasing a wide variety of styles and content, ranging from SFW to NSFW.
FAQ
Comments (16)
I'm shocked at the level of quality I can get with this model! It's trully amazing!
Also, it runs STABLE on Forge Neo with a mere RTX 3070 8GB VRAM.
Contrast and colors are much better on v3.0 than v2.0.
My only complaint is that it won't generate thick or voluptuous women, no matter what I put in the prompt.
I have no problem generating voluptuous women :)
https://civitai.com/images/124589558
Thanks about the compliments!
@Cyberdelia Thanks for the answer.
We probably have a different concept of "voluptuos women". Maybe it's a regional thing.
I'm trying to achive body types close to this:
https://civitai.com/images/119970816
https://civitai.com/images/119970807
And maybe some like that:
https://civitai.com/images/123486288
https://civitai.com/images/123486050
I'll probably need a Lora, but my RTX 3070 struggles with Z-Image + Loras
@snaker07 Like this: https://civitai.com/images/124786472
And this :)
https://civitai.com/posts/27379591
"voluptuous plus-size" is the secret
@Cyberdelia you actually got much better results than I did. I'll review my prompting! Thank you!
Can you please make gguf version or fp8 scaled?
I already made an FP8 version 🙂 If you mean a different FP8-scaled variant or a GGUF conversion, let me know exactly which one you’re after
@Cyberdelia Thanks for your reply! FP8 runs very slowly on GPUs with the Turing architecture (RTX 20 series). There isn’t enough VRAM to run FP16. However, Turing works very well with the FP8 scaled format — for example, z-image-turbo_fp8_scaled_e4m3fn_KJ performs very well on RTX 20 series.
So I just wanted to ask if you have the possibility to make an FP8 scaled version of the model. Or maybe gguf Q6_K.
Any plans of a full ZIB version?
Thank you for your work.
I have some full versions... but yeah, don't have great results with them.
I am trying to run this on Swarm UI but there is no shift settings and all the images are extremely dark. How to solve this?
Stupid question, but did you install the latest version? (SwarmUI v0.9.8 Beta). I see that Zit Image is supported, so it should actually work.
if that doesn't work I would highly suggest forge neo with pinokio. That's what I use. It's fool proof and I like how you can access it locally from a smart phone while generating. I have prompt preset styles saved that make better outputs than some ComfyUI users with 56 nodes I have seen. If you end up doing that and being interested I would be more than willing to share.
@Cyberdelia Yes sir. I have the latest version. I got the v3.0 model now and it's working as expected. I dont know how it is fixed. Thank you for all of your great work.
@Serencebey There is a shift setting. In the generation parameters section make sure all available options are displayed. then search for shift in the search bar.