2024.09.05：      

      Update 1.1, fine-tuned the new LORA, gacha test, will randomly generate a realistic human face shape, not fixed. In the future, I will further refine the materials for fine-tuning. Continue online testing for now. Starting from version 1.1, this model will no longer integrate VAE and CLIP files, reducing the file size by about 5GB. Players who use the flux model should already have VAE and CLIP files downloaded in their local folders. I will no longer occupy hard drive space. Also, the workflow needs to be changed, the previous 1.0 workflow cannot be used for version 1.1.

       Please use the workflow https://www.liblib.art/modelinfo/6277684d8aaf4ff8ad6c6e0d4b8fd301?from=personal_page

          Additionally, during the gacha process, random cartoon images may still be generated. I temporarily do not know how to resolve this, as I tested the original version and it was the same. It may be related to the clip node in the workflow. I'll explore further in the future.

+++++++++++++++++++++++++++++++++++++++