A test style LoRA for Flux Dev built on 焦茶 (Cogecha)'s masterpieces. When using this model, you should be aware you're doing imitation on Cogecha's art style.

The captions are using natural language. The example pictures are generated using ComfyUI and flux_dev_fp8.

It was trained using CivitAI trainer but the job failed at the last epoch. Fortunately I have downloaded some results before it failed. I merged them locally and compressed it to dim=16.

About trigger: I didn't put a trigger in captions but the tagger was frequently using word "anime" so you may need to add anime style or anime character if the style doesn't show up.