This Lora is intended to create digital painting style fantasy generations.

The sample images are generated with strength 1, but experiment as you like. The workflow is available in the sample images in v1.

The v2 makes the images sharper and clearer, though it messes it up them sometimes. I'll see if I can do something about it in the next version.

The training was done in kohya-ss with the default flux settings. This is my first model published here, so I don't know much, but please try it and review it. Any donations are welcome as I had to train 10 different times with 3 epochs each to get something acceptable.

(All the images provided are generated with fp8 version of t5xxl, the fp16 version doesn't follow prompt very well (even without any lora), so I don't know what's going on with that. If you're generating locally, try the fp8 version of t5xxl if flux ignores your specific hair or skin color instructions.)