Support & FAQ

After required AI models have been downloaded, Notiary can record, transcribe, and organize notes without sending your note content to a server. An internet connection is required for the initial on-device model download, optional Pro model downloads, and App Store purchases. The default extraction model is approximately 444 MB.

When you record a voice note, Notiary transcribes your words using Apple Speech or the optional enhanced WhisperKit engine. The transcript is passed locally to a quantized Qwen3 model that structures your thoughts. If you mention actionable items like "buy coffee tomorrow at 10 AM," Notiary parses the result into exact dates and shows everything for confirmation before saving to Apple Calendar or Reminders.

Your audio is recorded into a temporary file in your app's local sandbox directory. Once transcription is complete, the temporary file is deleted. At no point is the audio or transcription sent over the network or saved in a cloud database. Your voice is completely private.

When you confirm items for export, iOS asks for write-only Calendar access and Reminders access. Calendar access only lets Notiary add the events you confirm. Apple requires full Reminders access for apps that add reminders, but Notiary does not read or modify your existing reminders. If you deny a permission, you can enable it in Settings > Privacy & Security > Calendars or Reminders.

Notiary requires iOS 17.0 or newer. For best performance, we recommend devices with Apple Neural Engine support (iPhone 11 or newer) and at least 4GB of RAM to comfortably load the local LLM. Large speech-to-text models require more processing time on older devices.