I was making a Lora for something else and this is what turned out. After testing the results were pretty interesting so I decided to share it. All sample images were just basic generation and not cherry-picked with no upscaling or face detailing (obviously) This lora tends to make images more realistic and brighter - it was '1girl' focused with training, so that will definitely influence the outputProbably best to add some nsfw tags in the negative

Files with SHA256: 512f114b20870162a26a9da6bd12bb28349005c315a027153893fe6d76284f1e