Download and fully extract the Windows x64 package from GitHub Releases, then run SnapSharp.exe. The app starts in the system tray. Press Ctrl+Shift+A to capture; download the relevant models before using local AI features.

Download and launch

  1. Download and extract. Download the Windows x64 package from GitHub Releases. Keep the complete extracted folder and run SnapSharp.exe.
  2. Find the tray icon and capture. The app starts in the system tray. Press Ctrl + Shift + A to select an area, then add arrows, text, or pixelation.
  3. Copy, save, or take it further. Copy or export the capture, or pin it on your desktop. For recognition, open “Recognize clipboard image…” from the tray menu.

Prefer a self-contained release, which needs no separate .NET Desktop Runtime. Keep the complete extracted directory rather than moving only the EXE.

Capture shortcuts

Default shortcutAction
Ctrl + Shift + AArea capture
Ctrl + Shift + FFull-screen capture
Ctrl + Shift + WCurrent window capture
Ctrl + Shift + VPin the clipboard image
Ctrl + Shift + HHide / show all pins

Customize shortcuts in Settings. If a shortcut does not respond, check for conflicts with other capture tools or system shortcuts.

Set up local AI

Download the models you need in Settings, then click Save Settings. Configure features individually; you do not need all models or a separate Python installation. Initial downloads require internet.

FeatureSettings and notes
Chinese/English OCROCR and recognition → recommended ONNX OCR, about 14.2 MB for models
Tables and formulasOCR and recognition → local tables and formulas
Background removalBackground removal → download a model and choose the default
Local translationTranslation, APIs, and privacy → local engine, about 1.04 GiB for model and runtime

Sizes come from the current project guide and exclude the application download. Different model choices change download and hardware requirements. Optional online translation requires separate enabling or confirmation.

System and local data

  • Windows 10 2004+ or Windows 11, x64; macOS / Linux are not supported.
  • Framework-dependent releases need .NET 10 Desktop Runtime; self-contained releases do not.
  • Recording may require the VC++ x64 runtime and Windows media components. Use the recording environment setup first.
  • Settings, models, and history default to the application folder. Change this in Settings → General → Local Data; migration takes place at the next launch after confirmation.
  • Screenshot history is not encrypted. For sensitive material, review save locations, history settings, and file access permissions.

Common questions

Which systems does SnapSharp support?

Windows 10 version 2004 or later, and Windows 11, on x64. macOS and Linux are not currently supported.

Do I need to pay or install Python?

SnapSharp is open source under Apache-2.0, with free public release downloads. Local AI features do not need a separate Python installation. Download the relevant models first; third-party components and models have their own licenses.

Are my images uploaded to the cloud?

Local OCR, table and formula recognition, background removal, and local translation run on your device without uploading images or text. Optional online translation and image search involve network requests and require enabling or confirmation.

Can I use it offline?

Capturing and annotating work offline. Local AI features work after the relevant models and runtime components are downloaded and configured. Initial downloads and optional online services need internet access.

Why is there no window after launch?

The app starts in the system tray. Find the SnapSharp icon in the notification area, or press Ctrl+Shift+A to capture. If a shortcut does not respond, check Settings for conflicts with other apps.

Should I check the AI results?

Yes. OCR can make text, spacing, and punctuation errors. Check table structure and cell values, validate formula LaTeX, and inspect hair, transparency, and complex cutout edges. Results depend on the input, model, and device.

Read the full user guide →