I was preparing some images for a training run today and though, rather than having to go into TagUI, load the directory, load the captioning model, choose the images I want and then caption them, wouldn't it be easier if I could just right-click an image and get a caption file for it?
So that what I made.
https://github.com/morphicschris/easycaption
I've put the code and the release files up on GitHub. Currently it only handles selecting single image files but it would be easy enough to add multiple file selection (Edit: added now!). The setup is a little tricky initially as you need APIs running for both Florence 2 captioning and optionally an LLM to tidy up the captions (removing some of the fluff), but once those are setup it's dead easy to just click an image and generate captions. I have both APIs set up on another PC on my network and I just point it to those in the configuration file. All setup steps are in the README on GitHub.
Hopefully this can be useful to some people. If you spot any issues or have any suggestions then let me know.
CrasH 🤘