Trained for PonyXL with various sources using kohya_ss. I recommend a weight around ~0.8 - 1.0.Since she hardly has any screentime in the tv series only 2 screencaps were used. I added the <code>screencap</code> tag as always but it won't replicate the tv series style that well ofc.Lilith Tags:<ul><li>lilith, black lips, wings, purple dress, strapless dress, purple elbow gloves, pearl necklace, half-closed eyes (her eyes are mostly half-closed in the data set)</li><li>black crown (works well), gold crown (low probability)</li><li>Since the source material has multiple eye colors use one of the following<ul><li>yellow eyes (has most images in the dataset), purple eyes (her actual eye color), white eyes</li><li>pink sclera, yellow sclera, red sclera, purple sclera, grey sclera</li></ul></li><li>demon horns, pink horns, purple horns, red horns, black horns</li></ul>Lucifer Tags (not the focus of this LoRA, has only a few solo images in the dataset but works ok):<ul><li>lucifer morningstar \(hazbin hotel\), sharp teeth, white top hat, white suit, formal, evil smile, half-closed eyes, black bow, pink vest, striped vest</li></ul>

Files with SHA256: 599dd0c00212274d5a9238594f4815ad0e2da895217be37780bd9040ecd86ce3