Big Thanks to RunDiffusion for all the help :)
This LoRA was trained on 55 images to capture a clean and expressive caricature style. It’s my first LoRA test on Z-Image Turbo, created to see how well the model handles stylized exaggeration without losing structure.
Training was done with the AI Toolkit, keeping everything lightweight, consistent and easy to blend with normal prompts. Only trained on People Caricatures
Use it with strength: 0.8-1.0
Works best with Triggerword: caricature