HushSnap

HushSnap bridges screenshots and OCR into one fluid flow. Press a hotkey and a crosshair overlay appears instantly - select a region (or click for the full screen) and the shot lands on your clipboard right away. A thumbnail fades in at the bottom-right corner; left-click it and the recognized text pops up, already reformatted into a clean, readable layout you can edit on the spot. Everything runs locally - your screenshots and recognized text never leave your device. It lives quietly in the system tray, so there are no windows or menus to pre-launch: the UI surfaces only when you need it, then fades away. HushSnap Website Demo Video 爱发电 (Support)

Click thumbnail → OCR	Drag thumbnail → Save	Edit → Redact

After capture, a thumbnail fades in at the bottom-right corner of your screen. This thumbnail is the heart of HushSnap's post-capture flow:

• Hover to pause auto-hide and reveal an action pill with Edit, Pin, and Close buttons.
• Drag and drop the thumbnail anywhere to save the image - into a chat, an email, a folder, or another app.
• Left-click the thumbnail to run OCR. Recognized text opens in a floating popup where you can edit, copy, resize, or pin the window to keep it visible. By default the text is also auto-copied to your clipboard.
• Edit (the brush button on the action pill) opens the built-in image editor - see Image Editor below for the full toolset.
• Right-click the thumbnail for Copy Text from Image (silent OCR - text straight to the clipboard, no popup) or Save to Desktop.
• Optionally overlay a decorative vine ornament on the thumbnail’s top-left corner (enable it in Settings - Capture). It is purely cosmetic; it does not change the thumbnail’s hit area or any behavior.
• Enable Auto OCR After Capture in Settings to have text recognized and copied to the clipboard automatically right after capture — no thumbnail click needed.

Release Notes

See what's new.txt for the full changelog (Simplified Chinese, Traditional Chinese, English, Japanese — newest first).

As the project continues to evolve, some parts of this README may occasionally lag behind the latest behavior or features.

Image Editor

The built-in image editor opens from the thumbnail's Edit button, or from the right-click menu on a pinned image. It's a lightweight, dark-themed window for touching up a capture before sharing - annotate, redact, crop, rotate, resize, then copy to clipboard or save. The window opens centered on the cursor's screen and remembers its size across sessions.

Tools: rectangle / ellipse / line (color, size, fill, optional arrowhead), text (font, size, color), brush, highlighter, mosaic (pixelate/redact), eraser, pan, and the crop / rotate / resize transforms. Plus undo/redo (Ctrl+Z / Ctrl+Y), fit-to-viewport (Ctrl+0), copy, and Save As… (Ctrl+S, PNG / JPEG / BMP). Transforms run as atomic sessions - Esc cancels - so the state stays unambiguous mid-edit.

OCR Engine

HushSnap uses PP-OCRv6 as its sole OCR engine:

• PP-OCRv6 (via RapidOCR): Runs PP-OCRv6 small ONNX models in-process via the rapidocr Python package (Apache 2.0). No external dependencies or language packs needed. Works offline. Uses a unified multilingual model covering 50 languages in a single 7.7M-parameter model - surpassing the accuracy of the previous-generation v5 server model at a fraction of the size.

The engine supports the following 50 languages out of the box:

Core: Simplified Chinese, Traditional Chinese, English, Japanese

Latin-script (46): French, German, Italian, Spanish, Portuguese, Dutch, Polish, Romanian, Czech, Swedish, Norwegian, Danish, Finnish, Hungarian, Turkish, Vietnamese, Indonesian, Malay, Azerbaijani, Afrikaans, Bosnian, Croatian, Welsh, Estonian, Irish, Icelandic, Kurdish, Lithuanian, Latvian, Maltese, Māori, Occitan, Slovak, Slovenian, Albanian, Swahili, Tagalog, Uzbek, Latin, Serbian (Latin), Catalan, Basque, Galician, Luxembourgish, Romansh, Quechua

Minimal cv2 build

The shipped MSIX is slimmed down: HushSnap uses a purpose-built 24.8 MB static cv2.pyd instead of the official 82 MB opencv-python wheel — 70% smaller, OCR output byte-identical (this is a size optimization only; the standard pip wheel works too). The trimmed package is committed in the repo (third_party/ binary + the cv2/ package at repo root, which dev and tests import directly). Rebuild when OpenCV or Python changes:

pwsh scripts/build/build_minimal_opencv.ps1 -NoIPP -ForceClean

Third-Party Acknowledgment

Design references: Text-Grab (MIT, OCR workflow), Pinta (MIT, image editor toolset), ShareX (GPL-3.0, capture workflow). Built with PyQt6 (GPL-3.0-only, © Riverbank Computing Limited) — see THIRD_PARTY_NOTICES.md for all attributions.

License

HushSnap is distributed under the GNU General Public License v3.0 (LICENSE.md). See THIRD_PARTY_NOTICES.md for third-party attribution.

Copyright © 2026 HushSnap.

Development & Debugging

Install dependencies:

pip install -r requirements.txt

Run from source:

python HushSnap.py

To enable debug mode, set debug = true in hushsnap_config.toml (the --debug CLI flag was removed because MSIX packages cannot receive command-line arguments).

Key Features of Debug Mode:

• Isolation: Running from source uses %LOCALAPPDATA%\HushSnap_Dev, ensuring your production settings remain untouched.
• Traceability: Sets log level to DEBUG and opens the log folder immediately upon startup.
• Live Output: Real-time logs are streamed to the terminal via the logging console handler (StreamHandler).
• OCR Inspection: Debug mode saves detection-box images to the data directory. Right-click the tray icon → "Config Folder" to open this location directly.
  • ocr_debug_words.png - raw PP-OCR detector word boxes (red)
  • ocr_debug_lines.png - post-clustering line boxes (green, L0/L1/… badges)
  • Source run: %LOCALAPPDATA%\HushSnap_Dev\
  • Packaged run (MSIX): %LOCALAPPDATA%\Packages\<PackageFamilyName>\LocalState\
  • Packaged run (PyInstaller standalone): %LOCALAPPDATA%\HushSnap\

Building (MSIX)

build_msix.bat              # build unsigned MSIX; version auto-resolved from git tag
sign_for_local_test.bat     # self-sign the package for local install testing

The build requires HEAD at a git tag (e.g. v0.3.0). For local testing, sign_for_local_test.bat auto-creates a self-signed certificate and trusts it - run as Administrator.

Every release is tagged in git — git checkout v1.5.4, run build_msix.bat, then sign_for_local_test.bat to produce a locally-installable MSIX for that version.

Release notes (four languages, newest first) are in what's new.txt.

---

Installation

HushSnap is distributed exclusively through the Microsoft Store.