A Wayland utility for OCR and Google Lens search, built using Quickshell. It is compositor-agnostic and works on any wlroots-based compositor (Hyprland, Sway, River, Niri, MangoWC, etc.). It's meant to be fast, minimal, and stay out of your way.

• OCR: Select a region to extract text using Tesseract. It includes a cleanup script to fix the awkward line breaks and spacing that Tesseract usually spits out.
• Google Lens: Uploads a cropped JPEG directly to Lens using a form injection hack. This avoids using third-party image hosts and is significantly faster.
• Selection UI: Uses fragment shaders for background dimming and spring physics for the selection box. You get precisison crosshairs while hovering and fluid animations while dragging.
• Low Overhead: This isn't a daemon. The process only spawns when you trigger the keybind, performs the capture/OCR, and kills itself immediately after.
• Compositor Support: Works on any compositor that supports wlr-layer-shell and wlr-screencopy (standard wlroots protocols). This includes Hyprland, Sway, River, Niri, MangoWC, and others.

• Tab: Toggle modes (OCR / Lens)
• t: Switch to OCR
• g: Switch to Lens
• Escape: Quit (cleans up temp files)

QuickSnip works on any Linux distribution. You just need the following dependencies in your path:

• quickshell
• grim
• imagemagick
• tesseract + tesseract-data-eng
• wl-clipboard
• curl
• libnotify

Fedora:

For other distros, install the equivalent packages using your package manager (e.g., apt, zypper). For quickshell, follow the official build instructions if it's not in your repos.

Clone it into your config:

Add a keybinding to your compositor config to launch the tool.

Note on Scaling: If the selection area looks shifted or wrong, it's probably Qt scaling fighting with the compositor. You can force it to 1 like this: