This workflow allows you to do the following:

1. Batch-process images in a directory.

2. Images are resized to squares. Any added margins are filled in by AI.

3. Generate captions. You can choose from four caption patterns: General, People-Focused, Clothing-Focused, and Style-Focused.

4. After the first round of caption generation, use an LLM to refine the captions.

5. Save the images and captions to a specified directory in the same location as the original images.