Download
1 variant available
V2: Used the generate-new-images method, which has improved it. However, also CyberRealistic's embeddings also really make a difference even with the original one.
V1: Another of my embeddings, converted into a lora. This one I used a different method - taking the basic training data of my old one, and upscaling, despite it being 512.
The resulting Lora functions, and gets a workable character, but it's one of the worst ones out of the set. I won't be using this method again, and hopefully will find time and buzz to go back and train another one.
Any Buzz tip for any of these is appreciated.