2603 ZIT — By Stable Yogi
Fast photoreal Z-Image Turbo with a natural, real-life look — believable skin, faces, and lighting instead of the over-sharp "stock studio" feel. ~8-step renders, SFW → spicy.
Highlights
📷 Natural, real-photo realism — soft, believable lighting
🧬 Clean skin texture, crisp faces, reliable anatomy (holds up on complex poses)
🎨 Versatile — portraits, full-body, editorial, candid
🧩 Excellent LoRA base
What's new
Retuned for true-to-life realism — dialed back the over-processed studio look
Ships in fp8 (Forge + ComfyUI) and Q8_0 / Q4_0 GGUF (ComfyUI)
Recommended settings
Steps 8–9 · CFG 1.0 · Sampler Euler · Scheduler Simple
Resolution 896×1152 or ~1024 square
Which file — Q8_0 GGUF = best quality (ComfyUI) · fp8 = Forge or ComfyUI · Q4_0 = smallest / low-VRAM
Want the stronger, higher-detail Pro build? → Get the Pro version of my models here
Description
XMAS is a separate experimental build, focuses on balanced images and , and some secret ingredients under testing.
Get the Fp16, Q5 and Q4. Get the Pro version of my models here
FAQ
Comments (19)
BEST ZIT MODEL EVER!
after testing multiple models, this one by far has the most realistic/life-like outputs! the quality is amazing, the lighting is perfect, the prompt adherence is better than any other z-image model ive tested. very much appreciate the work youve done and i hope anyone reading this decides to give it a chance. well worth the download
you are talented on how to take a photorealistic model and tweak it to make it more fake, and you're number 1 in downlaods wow !!
Different eyes, different tastes. Glad it’s useful to so many people.
@Stable_Yogi ignore the haters. Keep up the great work.
The Lab Rat 11GB version is pretty good, it lose a touch of prompt adherence but give much cleaner image than base model, I feel ZIT tune and merge have potential to completly make even top photoreal SDXL models 'obsolete' in a few months.
I still manage to make most of my cleanest ones with SDXL models still, but it's really getting close and with better prompt adherence it will be a pleasure to switch once the ZIT merges/trained models get to this level. well done
Is there any way to get it working on ForgeUI ?
Yes, Try Forge Neo.
THanks for your help, it worked
@stargazer786 I'm trying to get forge neo to work. What specific text encoder and vae should I be using? Could I get a github link?
@stargazer786 How do I get it to work on forge neo? What text encoder and vae I need?
@olternaut I'm unsing ae.safetensor and Quen3_4B_Q5_K_M.gguf
The Q8 checkpoint is really excellent :)
Trained, not lora merging? I thought ZIT couldn't be trained until the full model was out?
That's the case from my experience with every single ZiT Checkpoint on here. None of them work with other LoRAs very well. Charging 100 buzz aint that bad, I've seen others charging 7k for their busted ass models.
Anyone know when the full model is supposed to drop?
@TiffanyHeat try the kaiji version. works good with loras. Some loras might need a lower strength to keep a similar style to the checkpoint.
@Melodic_Possible_582589 Where might I find this? Looking up 'zimage turbo kaiji version' wasn't of much help.
@TribalDiffusion sorry. mixed up the letters. it's kijai. you will also need the text endcoders and vae. check the description and/or the comments. "some" loras will need a lower weight like 0.5-0.8 if it changed the style too much. https://civitai.com/models/2170391/z-image-turbo-fp8-kijai
I tested Xmas FP8 and in my opinion, unfortunately, like almost all FP8s, it has too little color and contrast (with the exception of your Lab Rat v0 FP8, which has very good color and contrast). I tested Euler a/beta and 8 steps; the results were very interesting and almost flawless.
For me a very good checkpoint... with a bit more color and contrast i think the best one :) thank you for it
dont forget to try 13 / 16 steps and up the cfg to 1.5/2 , i found boosting the config can add some more colour and contrast to images. also try dpmpp sde/ beta, yes it doubles the generation times but might give you a more pleasing result 😊