Without further ado, this is a partially retrained model. Unlike all previous versions, it has 64 alpha X 64 dim.

With this version, I wanted to let myself understand in which aspects I made mistakes during the learning phase.

Creating Lora based on the artist's style is a very interesting experience for me. I learn something new every day. Did you know that the checkpoints that we all use are just LORA mixes based on a single model? Anyway, that's the picture in front of my eyes right now, as long as I know that I can mix lora and checkpoint together. The whole problem of creating your own checkpoint lies in the complexity, as this is an incredible level of task that requires a lot of time and experimentation. I am talking about all this because my task of copying the style of the artist Shenki implies an objective repetition of her brushes, strokes, layers of light, and so on. Do you understand what I'm saying? While I'm using other people's checkpoints, my LORA is not objectively disclosed to the full extent that I personally need.