[Grom_PE]

On my Linux system, I hooked up xfce4-screenshooter's "custom action" to a shell script (ocr.sh %f) with tesseract like so:

    #!/bin/sh
    set -o pipefail
    lang=${2:-eng}
    if tesseract "$1" - -l $lang | xclip -selection clipboard ; then
      notify-send "Text copied"
    else
      notify-send "Could not copy text"
    fi

It works great most of the time along with the xfce4-screenshooter's ability to select a rectangle.

When the text is especially difficult for tesseract, I can use Gemma3-4B via llama.cpp's llama-mtmd-cli, but that takes a minute.

[frumiousirc]

https://0x0.st/K9hq.png