My first LoRA release - extremely experimental, trained on a curated dataset. Expect jank. Future versions will focus on expanding the dataset and actually fixing hands xD.
If you're a degenerate for atmospheric vibes and glowing eyes, this might be your thing. Feedback appreciated so I can make v0.2 less scuffed.
Z Image Turbo:
Disclaimer: If you're looking for NSFW anime lora, this one might not be what you want. In the current dataset there is no nsfw images. Together with pretty weak nsfw capabilities of base turbo model leads to very underwhelming nsfw gens with just this lora. It's something I would like to fix, but need to expand dataset for it.
Lite:
Strength: 0.6-1
Trigger: None needed
Full:
Strength: 0.4-1
Trigger: None needed
Extra instructions: This lora is kinda overcooked to really hone in on the style I wanted. In most usecases, the best way to use it is to run 1-2 steps of lite/no-lora and finish the rest with this for the style. All the images in main post use just this lora at all the way through.
Pony V7:
Trigger: style_cluster_fuhrriel
Strength: 0.8-1
Chroma:
Trigger: not needed
Strength: 0.6-1
Description
Initial ZIT release. Largely the same dataset, recaptioned with Qwen 3 VL 32B