This model try to replicate the style of the lexica aperture v2 model.It was train using about 5000 portrait images and prompts scrapped from the lexica APIV1.1 is trained after manually curating the same image set to keep only the best 3000 images

aperture-v2-style

this updated version was trained with a curated set of the previous 5000 images reducing the dataset to about 3000 manually selected images.Let me know what you think if you want to see more version in the future !

Files with SHA256: 99ccd3390248b061b93f05fbed8f3022743a6d1c5db8ea772cdd4e911036a3c3