Kapture

A native macOS menu bar app that captures a screen region and extracts the text from it using on-device Vision OCR. Press ⌘⇧S, drag to select an area, and the recognized text is copied to your clipboard and shown in a floating window.

No network calls, no API keys — everything runs locally.

Features

Global hotkey ⌘⇧S works from any app
Region selection overlay (all screens, Escape cancels)
OCR for English, Chinese (simplified/traditional), Japanese, Korean
Text is auto-copied to the clipboard and shown in a floating window
Extract text from an existing image file via the menu bar icon
Menu bar icon with permission status

Build & run

./build-app.sh          # produces Kapture.app
open Kapture.app        # or drag to /Applications

For development without the bundle:

swift run Kapture       # runs from terminal (will appear in Dock)

First launch

Click the menu bar icon → Grant Screen Recording Permission… (required so the app can capture the screen; text extraction from files works without it).
Press ⌘⇧S, drag to select a region, release.
The extracted text lands on your clipboard and in the result window.

Note: macOS may require re-granting Screen Recording permission after rebuilding the app, since the binary identity changes.

Smoke test

swift run Kapture --test-ocr   # renders text, OCRs it, prints PASS/FAIL

How it works

Concern	Tech
Hotkey	Carbon `RegisterEventHotKey` (⌘⇧S, global)
Selection	Borderless `NSPanel` overlay per screen, dim + clear-cut rectangle
Capture	`CGWindowListCreateImage` (retina-aware)
OCR	Vision `VNRecognizeTextRequest` (accurate, multi-language)
UI	SwiftUI `MenuBarExtra` + floating result window

Layout

Package.swift          SwiftPM manifest (macOS 13+)
build-app.sh           builds a proper .app bundle + ad-hoc codesign
Sources/Kapture/
  App.swift            app entry, menu bar UI, permission helper, OCR smoke test
  HotkeyManager.swift  global ⌘⇧S hotkey
  OverlayView.swift    selection overlay panel + drawing
  CaptureController.swift  selection flow, screen capture, coordinate mapping
  OCRService.swift     Vision text recognition
  ResultWindow.swift   floating result window + image-file extraction

2.4 KiB Raw Blame History