Refined model for my two OCs with better training data. Still haven't gotten it to fully associate the hair color with the trigger words, but now has somewhat works as a style LoRA for the facial structure. Theoretically "lunaviltaria" should produce blue hair, "solaviltaria" should produce red hair, both with similar facial features and blue eyes. In practice the LoRA has picked up more on the image styling and less on the distinctive character attributes.

Luna & Sola Viltaria V1.5

Trained with entirely 1024x1024 images with more curation to bake facial features into the model. Hair color still iffy.

Files with SHA256: 1031493d08abcbee7f61bff6cee84d92c71832102b84cac387436876e042bed9