This model is a precision-tuned ZiT merge designed to push the boundaries of fine-grained texture, anatomical accuracy, and atmospheric lighting while maintaining compatibility with LoRAs trained on the base models. I've train my character and style LoRAs on vanilla Z-Image Base. They work fine on Turbo derived models and tend to be more stable during training.
I will upload future models to Hugging Face also, it's a lot faster than dealing with a VPN
https://huggingface.co/PsychoLogic/PsychoMix_ZiT
The Composition
The architecture of this model is a blend of checkpoints and custom refinements:
v1.x
Core Foundation based on Moody Porn Mix v10 for subject pose, with various other models merged for style and consistency (see 'About this version' for specifics for a particular release)
For a more raw/realistic look I recommend using z-image-turbo-flow-dpo with a strength of -0.5 to -0.75 (yes, negative). It was baked into the base model and many (myself included) think it's merely a sloppifier.
v2.x
Core Foundation: A balanced merge of Z-Image Turbo NSFW v7.2 for subject pose, CyberRealistic Z-Image Turbo Catalyst v0.7 for style and technical precision, and selective merge from Famegrid ZIB v1.0, focused on later layers
Feedback & "Show Your Work"
I am actively looking for constructive feedback to improve this and other models. If the model isn't performing as expected for you, please help me fix it!
The "Pics or it didn't happen" Rule: If you encounter issues (artifacts, anatomy failures, or poor prompt adherence), please post your generation with metadata (Seed, Prompt, Sampler). Negative comments without an image and settings typically don't help. If you want a fix, show the failure so I can tune it out in the next merge.
Description
Back after a couple of weeks' distraction working on other projects.
v5.0 is a combination of the best parts of v3.0 and v4.0, comprised of:
Moody Pro Mix v13 (finally no DPO!) x Sick Ollie v1 for the ZiT foundation
SeeSee21's Z-Anime on some early layers, and weaker than v3.0
Juggernaut Z v1.0 skewed towards the later layers
FAQ
Comments (8)
Psycho I love your models, some of the most interesting ones out there, would you mind doing an FP8 like before so those of us with lesser cards can try the latest version? I tried doing it but I don't think my 6GB VRAM is enough to convert.
Yes, I'll do it today / this weekend when I have time
Quants are uploading to HF, should be here within about an hour: https://huggingface.co/PsychoLogic/PsychoMix_ZiT/tree/main/v5.0.
Will then trickle through to here as the VPN allows
My testing for v5.0 shows the fp8 conversion was rather lossy. Results are OK, but they will be different. The Q8 and even Q6 GGUFs are closer to the original BF16
Great!
v5 quite good. because it has moody pro mix, it helps with the clean skin tone. what does the anime layer do to the checkpoint?
Why do faces always look so ordinary, not very attractive? Any key keywords?
I would say it's because I'm not loading it up with a heap of style changing LoRAs. Maybe add some - compatibility with the base models (I use ZiB for training) is one of my major goals.