If you want you can support my work to help fuel more experiments in AI and digital art!
https://buymeacoffee.com/dreammaking
Lora for Z Image trained on 138 italian influencers.
Specifically crafted for the Z Image diffusion pipeline. Trained over 35,000 steps on a curated dataset of 138 authentic Italian influencers, this model captures the essence of modern Italian aesthetics — from effortless chic to vibrant street style, all infused with Mediterranean charisma.
Whether you're generating portraits, fashion shots, or lifestyle content, this LoRA brings a distinctive Italian flair: sun-kissed skin tones, expressive eyes, dynamic poses, and that unmistakable dolce vita energy.
Description
Trained on ZBase with Prodigy optimizer
FAQ
Comments (3)
It looks really good! Would you mind sharing your learning rate and related stuff besides Prodigy optimizer? I've got similar size dataset and I'm struggling with training parameters, details look undercooked despite 24k steps 🫥
same for me Ostris dosent work for me
hi, sure. But be aware that sometimes i experience weird anatomy issues, so even my settings might not be the best ones https://pastebin.com/tr0R3hde