A personal fine-tune of Z-Image Turbo, customized to my specific aesthetic preferences.
This is a highly subjective model tailored to my taste—it's not intended to replace or improve upon the original Z-Image Turbo, just a specialized variant for my own workflow. Results will vary depending on individual preferences.
Key characteristics:
Strong bias toward photorealistic, high-detail skin and lighting
Enhanced curve and body emphasis (baked-in style)
Performs best with provocative/glamour-style prompts
NSFW capability is present but limited—can produce suggestive content reliably, but extreme/explicit requests may degrade quality or fail
LoRA compatibility: Limited. External LoRAs can be applied, but work best at low strengths (0.2–0.4). Higher weights often cause artifacts or overpower the base style.
Recommended settings (tested in Forge Neo):
Sampler: DPM++ 2S a RF
Scheduler: Beta
Steps: 10
CFG: 1
No ComfyUI workflows provided—developed and used primarily in Forge.
Use at your own risk and expectation. Feedback welcome, but this is a personal model, not a general-purpose release. Enjoy responsibly!
Description
V 1.0
FAQ
Comments (5)
Can I run this on ComfyUI ?
Yes
Can you share a workflow for how to use this on ComfyUI and is compatible with the native ZIT text encoder and VAE ?
@akringerzekrom656 I don't use comfyui so I can't share a workflow. I use forge neo. The standard ZIT text encoder and VAE will work fot this in comfyui. I'm sure this will with any ZIT comfyui workflow you find
@akringerzekrom656 use the standard template for Z-Image in ComfyUI, then change the model