Privacy Policy

What Uttra processes

While you dictate, Uttra processes microphone audio in memory on your Mac. The bundled speech model transcribes that audio locally and cleans the result. Copy to Clipboard, the clean-install default, leaves the text there for you to paste manually. If you explicitly select Automatic Voice Input, Uttra briefly stages the text on the macOS clipboard and sends Paste. Audio is cleared when the active workflow ends or is cancelled.

A transcript and any confidence-review candidate are held in memory only for the active workflow. The Mac App Store build does not read text from other applications.

The macOS general clipboard can participate in Apple’s Universal Clipboard when you enable Handoff in system settings. That is an operating-system feature outside Uttra; disable Handoff if you do not want clipboard content synchronized by macOS.

Data stored on your Mac

Uttra stores app preferences and any vocabulary corrections you choose to save. Dictation history and local diagnostics are off by default and are stored only if you enable them. Diagnostics contain event names and technical metadata, not audio or transcript content.

Stored data remains inside Uttra’s macOS app container. You can inspect, export, or delete it from Uttra’s Privacy settings. “Delete All Data” removes vocabulary, history, diagnostics, and resets preferences.

What we do not collect

• No audio recordings or transcripts
• No account or identity information
• No analytics, advertising identifiers, or tracking data
• No surrounding text or saved vocabulary
• No telemetry or crash uploads

Uttra does not sell personal information and does not share data with third parties because the app does not collect data from you.

Permissions

Uttra requests microphone access only when you start dictation. Copy to Clipboard is selected by default and requires no Text Insertion permission. Only after you explicitly select Automatic Voice Input can Uttra request macOS’s limited Post Event permission, shown under Accessibility, solely to send Paste after you finish speaking. This does not grant full Accessibility access: Uttra cannot read another app’s interface, selection, or surrounding text, and it does not monitor general keyboard input. Shortcuts use registered key combinations and never request Input Monitoring.

Network access and third-party software

The baseline speech model is included with the app. Normal onboarding, dictation, transcription, cleanup, and text delivery do not require a network connection. Uttra contains no analytics, advertising, cloud-inference, or telemetry SDK.

Whisper Large v3 Turbo and Large v3 are optional. Uttra connects only after you press Install or Repair, stores the model locally, verifies it, and performs later transcription offline. Download hosts receive ordinary connection metadata, but never your audio, transcript, vocabulary, or settings.

Children’s privacy

Uttra does not knowingly collect information from children—or from anyone else. The app has no account system or cloud data service.

Changes and contact

Material changes will be posted on this page with a revised effective date. For privacy questions or support, email Uttra support.