Z-Image is here, and it's pretty damn great. Let's make it more NSFW!
Still a work in progress. It can output good genitals, but not always. About 3/5 images are good, working on penis stability specifically for future versions.
ComfyUI: Models should be placed in diffusion_models, and uses the official workflow with the Load Diffusion Model loader, and loading the VAE/CLIP models separately. Models named "_AIO" should go in the checkpoints folder, and uses the normal checkpoint loaders.
Recommended Settings
Steps: 12
CFG: 1.0
Sampler: dpmpp_sde
Scheduler: simple
Description
FAQ
Comments (33)
great looking model! any suggestions for a good workflow ?
Each image comes with a workflow :)
LOVING 5.0! One note: I'm getting a lot of two-toned hair regardless of prompting.
Also, does this model work with negative prompts? I see some ZIT checkpoint makers have figured out how to implement that a bit. Would be nice, hopefully get rid of all the transgendered dudes :P
Cheers!
I haven't done any special work for negative prompts to work, the CFG is as it is.
incredibly hard to get pure blonde, yep
大神，你包涵了哪些明星，我测试有刘亦菲、迪丽热巴
model version named "_diffusion" should be placed in diffusion_models, and uses the official workflow with the Load Diffusion Model loader, and loading the VAE/CLIP models separately. Models named AIO should go in the checkpoints folder, and uses the normal checkpoint loaders
Pls explain, where are this folders in Forge?
I don't quite know I'm afraid. I use only ComfyUI at this point. Forge is too slow with updates and modern model support :)
hello, i download this model and what next?
goto official ComfyUI web-page and read manuals
is it worth using the AuraFlow Sampling node for the z-image turbo model ?
If you would like to use it I would totally recommend that you do.
However, if you are in the camp of people that prefer NOT to use the AuraFlow Sampling node, I think it's not worth your time.
I hope that helps.
@6tZ I looked at your chart and it doesn't exist.
@6tZ I really like your model, it's very cool, I don't know what you've added to it, but it works and generates very well, thank you!
@soyv4 Just use standard workflows, they are fine :)
Can you tell me what is added to version 5, how is it different from 4.5?
love v4.5 but v5.0 is a step back unfortunately. oversaturated and oversharpened... still, thanks for 4.5, my favorite ZIT checkpoint ATM
Yes, 4.5 is better in terms of realism, color, contrast, I agree with you, I stayed at 4.5
Thanks for the input! I'm personally bored with the muted "realistic" colors and prefer v5. But I'll tinker with a bit of a blend perhaps, as long as you can prompt for "colorful" :)
@6tZ a blend would be nice! i do like the different look it has but it's just a little bit too much :)
@6tZ i figured that with DPM++ 2s a RF Beta, 8 Steps and Shift 5 i quite like it!
@chgrau89578 Yeah merging was no good. 5.0 is so much better than 4.5 with stability, it's a big downgrade in terms of quality, so I won't be using 4.5 for anything than a guideline ^^
It is perfect for some Loras that may take out color and sharpness though!!
So if you want to do NSFW with z-image-turbo, this is basically the one to use right now.
It even fixes the contrast issue zit has, and almost removes the need for a LUT pass.
Thanks for the vote of confidence <3
If not mind asking, what you use to generate? If its a comfyUI flow, could you share? I tried a bunch of the ones in Civitai and all seems to break.
@NSFWSEGUNDO I use SwarmUI exclusively, which more or less is a UI wrapper around ComfyUI. My workflow is simple:
1) generate base image with this model
2) upscale 1.5x lanczos
3) refine with this model
4) upscale 1.5x with seedvr2
5) apply film grain and LUT
@clueless_engineer Thanks, never used SwarmUI, but will take a look.
Really amazing images @clueless_engineer <3
Thank you so much for sharing!