Screen Text Capture Shortcut on Wayland Linux (grim, slurp, tesseract)

Why set this up?

Select any region of your screen and get the text instantly on your clipboard
No extra apps to open — just a keyboard shortcut
Built entirely from small, composable CLI tools already common on Wayland

What you’ll need

Install these four tools:

grim — takes screenshots on Wayland
slurp — lets you select a screen region with your mouse
tesseract — runs OCR (text recognition) on an image
wl-copy (part of wl-clipboard) — copies text to the clipboard

# Arch
sudo pacman -S grim slurp tesseract tesseract-data-eng wl-clipboard

# Fedora
sudo dnf install grim slurp tesseract wl-clipboard

# Debian/Ubuntu
sudo apt install grim slurp tesseract-ocr wl-clipboard

The basic command

Keyboard shortcut daemons usually run commands directly, without going through a shell — so pipes (|) and command substitution ($()) need to be wrapped in sh -c to work at all. This bare version, with no notification and no backgrounding, is the minimum needed to get the pipeline running from a shortcut.

# This selects a region, OCRs it, and copies the text.
# No feedback, and can feel like it "hangs" briefly.
sh -c 'grim -g "$(slurp)" - | tesseract stdin stdout | wl-copy'

Note: pasting the un-wrapped pipeline (without sh -c) into a shortcut field will not work — it only runs correctly in an interactive terminal.

The working command

Running the pipeline in the background (&) keeps the shortcut from feeling stuck while tesseract processes the image, and adding notify-send gives you feedback that it actually ran.

# Backgrounded + notification on completion.
sh -c 'grim -g "$(slurp)" - | tesseract stdin stdout | wl-copy && notify-send "OCR copied" &'

Adding it as a shortcut

In your desktop environment’s keyboard shortcut settings, add a new custom shortcut:

Command: sh -c 'grim -g "$(slurp)" - | tesseract stdin stdout | wl-copy && notify-send "OCR copied" &'
Shortcut: pick any unused key combination, e.g. Ctrl+Alt+S

Press it, select a region containing text, and you should see an “OCR copied” notification with the text on your clipboard.

Improvement: fix false-positive notifications

There’s one rough edge left. The command above says “OCR copied” even when nothing useful actually happened — if you cancel the selection (hit Escape in slurp), or select an area with no readable text, it still reports success. That’s because && only checks that each command in the pipe exited without an error — not whether any real text was produced. wl-copy happily “succeeds” copying an empty string.

The fix is to capture the OCR result into a variable, strip blank output, and only report success when there’s actual text:

sh -c 'text=$(grim -g "$(slurp)") 2>/dev/null
[ -z "$text" ] && exit
text=$(printf "%s" "$text" | tesseract stdin stdout 2>/dev/null)
text=$(printf "%s" "$text" | sed "/^[[:space:]]*$/d")
if [ -n "$text" ]; then
    printf "%s" "$text" | wl-copy
    notify-send "OCR copied" "$(printf %.100s "$text")"
else
    notify-send "OCR failed" "No text detected"
fi &'

Now:

Canceling the selection exits quietly, no notification at all.
A selection with no readable text shows “OCR failed” instead of a false “OCR copied.”
A successful capture shows a preview of the copied text, so you can catch OCR mistakes at a glance.

Final command

Paste this as your shortcut command for the most reliable version:

sh -c 'text=$(grim -g "$(slurp)") 2>/dev/null; [ -z "$text" ] && exit; text=$(printf "%s" "$text" | tesseract stdin stdout 2>/dev/null); text=$(printf "%s" "$text" | sed "/^[[:space:]]*$/d"); if [ -n "$text" ]; then printf "%s" "$text" | wl-copy; notify-send "OCR copied" "$(printf %.100s "$text")"; else notify-send "OCR failed" "No text detected"; fi &'

Note: restart your keyboard shortcut daemon (or log out/in) if a newly saved shortcut doesn’t seem to fire.

Bind it to any free key combination, and you’ve got a reliable screenshot-to-text shortcut that only tells you “copied” when it actually copied something.

A note on built-in tooling

It’s worth pointing out that this whole workaround exists because newer desktop environments like COSMIC don’t yet ship a screenshot tool with OCR built in. KDE’s default screenshot utility, Spectacle, has had this kind of “extract text from screenshot” capability built directly into its UI for a while now — no piping together four separate CLI tools required.

As younger DEs like COSMIC mature, it’d be a genuine usability win if their default screenshot tools followed KDE’s lead here — baking OCR-to-clipboard in as a first-class option rather than leaving users to assemble it themselves from grim, slurp, tesseract, and wl-copy. Until then, though, the shortcut above gets you the same result.

👏 Nice work!

Don’t forget to follow 🌏 my socials and subscribe to my 💌 newsletter.