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: <a target="_blank" rel="ugc" href="https://civarchive.com/models/202380/crimson-notexist">Another of my embeddings</a>, 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.

Files with SHA256: 9461a27d132bb39fe0f74df933e77cd461c729f0c5a2e50db7856e1db17b45b3