This is the first LoRA that I have created for a new base model I am making that is called "sd_xl_vae_v1.01.1.1" ; I have merged this LoRA with the "sd_xl_base_1.0_0.9vae.safetensors" Base Model and the result is the first step for the new base model, "sd_xl_vae_v1.01.1.1" with many additional steps planned for the future. This LoRA consists of 1680 transparency images and 1680 blank text files. The repeats are 1 for each file and there is only 1 epoch for the LoRA. I chose to start this way because more images and less repeats focuses on detail, where more repeats and less images focuses more on similarity to the images. The images are scaled at 512x640 which for the sd_xl platform is an aspect ratio of 8:10; respectfully using 64 as a value; 64x8=512 and 64x10=640 ; by doing this I have constructed the new base model as being optimum and nominal at 800x1000 resolution because of a multiple of 10 for the 8:10 aspect ratio. This LoRA must be used with prompts or it will generate a gray screen when strength is applied to the LoRA; it does this because images were used to create it so it does not know anything but that.