Z Image Turbo Versions
Edit: This is possibly going to be the last of this model range, and is very much ready for use as is. Making this has been a great learning experience, and a lot of what's been learned here will be carried forward into the next model.
I hate asking, but if you can spare some buzz for training, or add anything you generate with it to the gallery here, it'll be appreciated.
The current optimum (for my taste) generation parameters are 4-6 Steps, CFG 1.0, using DPM++ 2s a RF with Bong Tangent, Beta, Simple. This works well with Forge Neo. If you use Comfy, YMMV. It does seem to do well with other schedulers/samplers, but there, you may find yourself using 6-8 steps, with Euler, for example.
As always, feedback welcome.
Have fun using it! :)
The new LexiVision-Z has been trained on top of the base Z-Image-Turbo, an experiment with generation of synthetic data using SDXL LexiVision AIO VNSFW and training locally. Beta 2 has been trained for an additional 4000 steps. I'm still trying to tweak it stylistically to match a more analog aesthetic, but it's taking a lot longer than I thought it would. It can work with DPM++/Simple, Res Multistep/Simple or Beta, but I'd suggest playing around with your favourite combo.
SDXL Versions
This is a new merge of AIO-NSFW with the fantastic Photo1 by SubtleShader.
This is another merge of two private models that are being used to generate synthetic training data for other models. There are various 'flavours' of model, the first of which will be one focused (pardon the pun) on photo-realistic pictures of more mature models.
Prompts can be generic or more descriptive, and prompt adherence is ok. The model's fairly NSFW, so be warned. It can generate SFW or NSFW content equally well.
It's DMD2 based, and the VAE is baked in, so few steps are required for generation.
As always, my personal favourite generation parameters are in the showcase images.
Description
FAQ
Comments (13)
On Fooocus the images are ruined, with extreme sharp effect, dark spots on the skin and oversaturated with the basic settings, why? It's totally different from what you see in your screenshots, I think it can't work well, are you sure it's SDXL 1.0?
Experiment with lower CFG, fewer steps, etc. The settings work for my workflow, yours may vary.
what settings are you using on foocus ?
@mariedoll123 I open Fooocus realistic presets, go to models, call the checkpoint, put high quality as I do with all the other checkpoints that have never given problems. This one instead there is no way to make it work well, I lowered the CFG and other options in the debug options but nothing changes. I also tried disabling all styles but nothing.. It's very strange because DSXL 1.0 is the native supported format, it's the first time a checkpoint gives me this kind of problems.
@NoMarzipan8994381 this version has the DMD2 lora built in , try lowering config to 1 and steps down to 8, plus change the sampler to LCM / exponential . go into the advanced settings to change the settings.
a high cfg and number of steps will create a distorted image if DMD2 is used, check out the DMD2 lora on here to find out more.
@mariedoll123 you're right, the problem was the sampler, I had the same problem with another DMD2 based checkpoint, besides this one. You can solve it by changing the sampler to LCM and lowering the guidance scale a bit to not have too much saturation, now it works well! Thanks!
@NoMarzipan8994381 happy its worked 😊
NoMarzipan8994381 FWIW, I tend to create a Fooocus Preset when I download a new checkpoint. Saves me trying to remember what I need to change.
Nice model! Can you post a version without DMD2?
Sorry, I use models with DMD2 baked in as sources.
Oh, how sad :( Thanks for the reply
In some, you can "undo" DMD2 by applying the dmd2 lora with weight -1. In some, the weight might be slightly different, e.g. -0.75. In others you can't do it at all.
I think it depends on how the merge is done and if there's any subsequent refinement on the model or not. I don't know enough about how it works. But in my experience DMD2 sacrifices prompt adherence for speed, so I try to get rid of it when I can, and it often works.
TrueToLife_Fauxto GIGACHAD merging DMD2 so we dont have to download or waste time on finding that lora each time
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.