As the name suggests, this is LORA, which focuses on a beautiful female face with large eyes and long curly hair
This LORA is most suitable for creating a portrait image of a girl (it can also be used for full-length images).
This LORA is trained on her own artwork of girls created in Zbrush and Blender.
Checkpoint - ReV Animated and Balovine V1.L.a6 are best suited for this LORA.
Sampling Method: Euler A ; DPM++ 2M Karras and DPM++ 2M SDE Karras
Description
FAQ
Comments (3)
Very good Lora for more interesting faces, I expect it's not seen a lot of use because of the random text it generates but if you want to stop the random text appearing, you can put - (tattoos, body markings, text) in the negative prompt and it'll (almost) completely stop it. In realistic images typically makes characters skin darker, which helps when a lot of loras try to make characters paler. Edit: It would benefit a lot from being able to dial the on-site version's strength down, a -2 to +2 range is the norm and this is locked at 1+.
Lora should follow the prompt, not making skin darker just because. I have no other loras and all images are African females. While in prompt - pale skin Caucasian.
Typical - "not a bug but feature", excuse for poor job.
Lora author did poor tagging job, so even if you try to make pale skinned person it will output dark skin.