This Concept LoRA captures a personal aesthetic preference for realistic female faces.

It was trained on a carefully curated dataset of different actresses and public figures rather than a single person. The purpose is not to reproduce any specific individual, but to bias the generated faces toward a consistent facial aesthetic, including subtle shared characteristics found across the dataset.

This LoRA works best with realistic image generation models and can be combined with character, clothing, style, or environment LoRAs without significantly affecting non-facial elements.

It is intended to influence facial appearance rather than identity, pose, clothing, or composition.

V2.0 used same dataset, but trained with more steps.