This workflow does the following:

• Load 1-100 random images from a selected directory

• Create a grid from the loaded images

• Using Gemini to create a prompt by using the grid as a single image

• Use an input image as style reference

• Creates 4 different images using Detail Daemon

• Using Pick Score to select the image that best reflects the prompt

• Upscale image using an upscale model

• Refines the image using SDXL

• Go through Face Detailer

• Saves the image to Google Drive or OneDrive