This lora is a test, improve slightly the face of the woman than the flux base model. The dataset is a collection of image that scammers send to my email.

v1 lora is generated from 7 image trained for 462 step

v1.1 lora are the same 7 image but captioned with florence 2 large trained for 462 step

v2 lora have an increased number of image, 34 captioned with florence 2 large trained for 1000 step