OCR Screenshot Tool

A macOS tool that captures a screen region, runs OCR via the native Vision API, and copies the result to clipboard. Integrates with Raycast as a Script Command.

How It Works

1. Press a hotkey in Raycast
2. Select a screen area
3. Text is recognized and copied to clipboard
4. A sound plays to confirm

Requirements

• macOS 13+ (Ventura or later)
• Python 3.10+
• Raycast

Installation

bash install.sh

This will:

• Create a Python virtual environment
• Install dependencies
• Register a LaunchAgent (auto-starts the OCR server on login)
• Wait for the server to be ready

Then in Raycast:

1. Open Settings > Extensions > Script Commands
2. Add this project directory
3. Assign a hotkey to "OCR Screenshot"

Architecture

OCR Server (ocr_server.py) — FastAPI server on 127.0.0.1:19876 using macOS Vision API. Runs as a LaunchAgent, auto-restarts on crash.

Raycast Script (ocr.sh) — Captures a screen region via screencapture -i, sends it to the server, copies the result to clipboard, plays a sound.

Supported Languages

Ukrainian, Russian, English — detected automatically by the Vision framework.

Manual Usage

# Start the server
.venv/bin/python ocr_server.py

# OCR a file
curl -s -X POST http://127.0.0.1:19876/ocr -F "file=@screenshot.png"

# Health check
curl http://127.0.0.1:19876/health

Uninstall

launchctl bootout "gui/$(id -u)/com.ocr.server"
rm ~/Library/LaunchAgents/com.ocr.server.plist