The 26 image + caption pairs used to train inlineresearch/skin-lora-krea-2-raw, a photorealistic skin-texture LoRA for Krea 2. Close-up and half-face portraits chosen for real, unretouched skin: visible pores, fine hair, natural texture and imperfection, across a range of skin tones, ages, and lighting.

Contents

data/ metadata.jsonl captions for the dataset viewer (one row per image: file_name + text) 0000.jpg… See the full description on the dataset page: https://huggingface.co/datasets/inlineresearch/krea2-skin-lora.