This is a LoRA that was trained with the "sd_xl_base_1.0_0.9vae" Base Model and works very nicely with it as far as generating. I have included a Comfy .json Workflow that creates wonderful images; all you need to do is enter your prompt on the last blank line of the positive prompt. This Lora has 398 images with a repeat of 4 for 1592 training steps. I have also included the entire image_text_dataset as thumbnails so you can easily read them to produce images used in the training. The Lora will work from -.2 to 2.3 and the nominal sweet spot it seems to be .68.