I have a piece of Elisp that checks if the last thing in clipboard is an image, then sends it to tesseract¹. It's not as accurate as I wish, but it's very fast and it works for cases when someone's screen sharing. I've been thinking of extracting it all into a proper package with some additional features.

On Mac, I combined it with Flameshot to select a region, and Hammerspoon that watches the clipboard, so it always automatically OCRs anything I screenshot and pops into a buffer. On Linux I haven't spend time figuring out how to hook it up, so after the screenshot, I just have to manually call the command.

¹ https://github.com/agzam/.emacs.d/blob/main/modules/writing/...