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
FP8 - Experimental Build, please share feedback, drop images.
Get the GGUF versions Q4 (3.3 GB) and Q5 (4.1 GB) Get the Pro version of my models here (free for all)
Thanks
FAQ
Comments (13)
How exactly are you training the model? I've been under the impression that it wasn't worth it until the base was released, other than merging with LoRAs. Great job!
So Stable_Yogi, what is your opinion on the potential of Z-Image Turbo. I want to know if I should jump into using it above Illustrious or Pony models with their LoRAs.
Right now?
I’m sitting here full Doctor Strange mode, calculating like 14 million timelines trying to see which model wins 💀🔮( XL or ZIT)
We’ll know soon.
Your v0 fp8 takes very good pictures. Like all Zimage-Turbo fp8s, the color and contrast are slightly diminished compared to your BF16. (and yes, i tested it without LoRa) So that's the only thing I'd suggest you could change. Otherwise, feel free to include some LoRas, skinny, shiny, detail, and fetish content (but there isn't much good content created in that category yet) :)
Thank you for your checkpoints :)
Right now?
I’m sitting here full Doctor Strange mode, calculating like 14 million timelines trying to see if i could keep this model in my Pipeline. If yes, then a lot of Loras will be done soon.
@Stable_Yogi i know you will do a great job :)
It's truly amazing how good the quality is for this FP8 checkpoint; it's really fun.
Que maravilla tu modelo FP8!!! muchas gracias!!!
at least they have different breast shapes.
v0 Q8 I really like the Checkpoint. It has much better colors and contrast. I just needed a bit of skinny LoRa. I think it's a very good Checkpoint, thank you for it :)
why Zit having difficult time generating genitals been trying different model still the same
Because the base dataset is censored (No genitals, no nipples), and it is currently impossible to finetune or meaningfully merge the Z-Image-Turbo model. You can try lora but ymmv. Finetunes and merges are largely a grift (if they're not free) at worst and made in ignorance at best. The way Ai Toolkit (or any other trainer) trains values into a model makes the new values end up in a way where Z-Image-Turbo cannot access them during diffusion, the "changes" to output are due to introduced noise/dilution of the matrix. (So for the time being it is impossible to fix via tuning, either for this model or any other model, wait for Z-Image-Base to be released)
@InvictusAI wym? theres plenty of loras out there that fix it perfectly?
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.