Privacy and data

What data SigXor handles, when it connects to the network, where data is stored, and the website's privacy practices.

Updated 2 min read Edit this page on GitHub
All documentation
On this page

SigXor is designed local-first. File preview, speech recognition and OCR all happen on your machine, and the app contains no telemetry or behavioral analytics.

Local processing#

Capability Where it runs Uploaded
File preview and editing On-device No
Speech recognition (SenseVoice) On-device ONNX inference No
OCR (PaddleOCR) On-device ONNX inference No
Screenshot and color picking On-device No
Office web rendering Local WebView2 No
Update checks and model downloads Official release sources Download only; no content upload

Network access#

SigXor connects to the network only in these cases:

  1. Update checks and downloads: reading the version manifest from GitHub Releases and downloading the update package.
  2. Model downloads: fetching SenseVoice and PP-OCR models from their official model sources the first time voice or OCR is enabled.

Outside of these cases the app makes no outbound requests and never sends file contents, audio, screenshots, clipboard data or usage statistics.

Where data is stored#

All user data lives in:

Documents/FNSoftware/SigXor/
├─ settings.json
├─ Backups/
├─ Logs/
├─ Models/
├─ Screenshots/
└─ Audio/
  • Settings: stored only on this machine.
  • Backups: written to Backups/ before text or binary files are saved.
  • Logs: unhandled exceptions are written to Logs/ for local diagnosis.
  • Debug recordings: stored in Audio/ when enabled and can be deleted at any time.
  • Models: stored in Models/; deleting them means they download again on next use.

Deleting this folder removes local data. There are no cloud accounts and no cross-device sync.

Clipboard#

Voice input uses clipboard paste by default. SigXor restores the previous clipboard content after pasting; if restoration fails, the recognized text remains on the clipboard.

Permissions and system integration#

  • Global hotkeys use low-level keyboard hooks and only react to the configured keys. Other keystrokes are not recorded.
  • Region screenshots read screen content only when you trigger them.
  • Typing into elevated windows may be blocked by Windows UIPI.
  • The app installs no browser extensions and injects into no other processes.

Website privacy#

This website is fully static:

  • No cookies, analytics, advertising or third-party fonts.
  • The theme preference (dark / light) is stored in browser localStorage and never uploaded.
  • No third-party CDNs are used.
  • External links point to public resources such as GitHub; following them is governed by those sites' privacy policies.

Third-party components#

SigXor uses open-source components including AvalonEdit, CommunityToolkit.Mvvm, Markdig, NAudio, sherpa-onnx, RapidOcrNet, SharpCompress, WebView2 and the Office rendering libraries. See NOTICE in the repository for licenses and attributions.

Contact#

Questions about privacy? Open an issue on GitHub.