V1.
Tinkerbell lora for Z-Image Base (Use CFG 4).
This lora also works with Z-Image Turbo (Set lora strength to 2 or more)

This model is experimental and is somewhat inconsistent

Activation text: tinkfairy

Prompt Structure:

1. tinkfairy (Activation Text)
2. blond girl in a green dress (English natural language text description)
3. 1girl, green, dress, fairy wings (Tags)


V2.

This version is trained on higher resolution inputs and fixes many of the issues with consistency. This version should be on parr with the illustrious version (from my limited testing).