I continue to be surprised that nobody has made a normal-user-friendly CLIP image search tool, except the one for iOS nobody seems to have heard of. It's been more or less possible since mid-2021, and clearly useful given the amount of people with somewhat janky implementations (like https://mse.osmarks.net/, my thing) which have proven quite helpful. I looked into it and determined that writing an actual desktop app which users can use is annoying, and so is doing inference of big models on random people's computers, but surely SOME people are okay with desktop app development even now.

I like desktop app development but I have no idea what this CLIP image stuff is about.

For Windows apps made with ancient tech, you could ask one of the Jases here.

CLIP is a machine learning model (technically a pretraining process used on several different models but whatever) which you can use to match images with text. It is very smart and general-purpose compared to previous ways of doing this. I use it for my meme library and such. The problem is that while I can get away with a lot of shortcuts when writing code for myself, something for the average person needs easy configuration options rather than a CLI and some variables in the source to edit, no requirement for a powerful GPU, and to be packaged as one thing rather than several Python scripts with a lot of weird dependencies and a web interface.

Other people have a greater tolerance for that kind of annoying thing, so it's odd that nobody's done it.