Lightning-fast OCR and QR code reader for Wayland/Hyprland on Arch Linux. Select screen area → Extract text or decode QR → Copy to clipboard.
- 🎯 Simple - One command to capture and extract text
- ⚡ Lightning Fast - Tesseract by default, results in under 0.5s
- 🖥️ Wayland Native - Built for modern Linux desktops
- 📋 Clipboard Integration - Automatically copies extracted text
- 🎨 Flexible - Fast mode by default, --accurate flag for complex text
- 🔗 URL Detection - Automatically detect and open URLs in extracted text
- 📱 QR Code Detection - Decode QR codes automatically
# Install system dependencies (Arch Linux)
sudo pacman -S grim slurp wl-clipboard tesseract tesseract-data-eng zbar
# Install TRex
pip install trex-ocrTRex includes both text OCR (using Tesseract by default) and QR code reading out of the box.
# Default: Select area → OCR → Copy to clipboard
trex
# OCR image from clipboard
trex -c
# OCR image file
trex -f screenshot.png
# Output to stdout instead of clipboard
trex -o stdout
# Use different language (e.g., German)
trex -l deu
# Use accurate mode for complex/handwritten text
trex --accurate
# Automatically open detected URLs in browser
trex --open-urls
# QR code mode - decode QR codes, fallback to OCR if none found
trex --qr
# Decode QR code and open if it's a URL
trex --qr --open-urls- Default mode: ~0.3-0.5s (Tesseract)
- Accurate mode: ~2-3s (EasyOCR, downloads AI models on first use)
Tesseract (default): Uses standard 3-letter codes
eng- English (default)deu- Germanfra- Frenchspa- Spanishjpn- Japanese- Install more:
sudo pacman -S tesseract-data-[lang]
EasyOCR (--accurate mode):
en,de,fr,es,ja,ko,ch_sim, etc.- Full list: https://github.com/JaidedAI/EasyOCR#supported-languages
Optional config file at ~/.config/trex/config.json:
{
"language": "en",
"gpu": false,
"open_urls": false,
"qr_mode": false
}- Set
"gpu": trueif you have NVIDIA GPU with CUDA support - Set
"open_urls": trueto automatically open detected URLs by default - Set
"qr_mode": trueto use QR code mode by default
- Arch Linux
- Wayland compositor (tested on Hyprland)
- Python 3.8+
- CUDA (optional, for GPU acceleration)
For complex images, handwriting, or multiple languages:
# Install accurate mode dependencies
pip install trex-ocr[accurate] --extra-index-url https://download.pytorch.org/whl/cpu
# Or with GPU support (NVIDIA CUDA)
pip install trex-ocr[accurate,gpu]Then use trex --accurate for better accuracy at the cost of speed.
MIT