<p>LoRA trained on a female headshot images. All training data was generated with the same base prompt which aimed to promote diversity, and was then filtered by similarity in appearance. All source data was generated by Nano Banana -- no characteristics of any individual were stipulated, and no images were used as input. Stable diffusion version trained on autogenerated tags, which seemed not especially accurate, rather than captions. trigger mytypeone_btwo</p>

Files with SHA256: 007b926c61ef3eba64aa9cad1ad60bac14c6d95ec2b6a76d90b7e4522384a01f