- Best of four ranking option instead of 1v1.
- Import corresponding text-prompts for each image.
- Regularly save the ranking file in case there is some error.
- Undo button.
- Add loading screen for when a directory with a lot of images is loaded in.
- Automatically annotate/describe/generate text prompts for the images using a (multi-modal) vision model.
- Tools to scale and crop an image.
- Tools to draw over labels/text on an image and autofill to delete it.
- Allow to write/edit text-prompts for each image.
- Pre-rank based on image quality.
- Possibly replace the Python TrueSkill with Javascript so as much as possible can run directly in the browser.