Privacy & Permissions
Last updated: July 23, 2026 · Applies to ListenIn for Mac v0.1+
A dictation app hears the most sensitive thing you own — your unfiltered voice.
So instead of a wall of legalese, here is exactly what moves where, and why.
Your audio
- Audio never touches our servers. Speech-to-text runs on Apple's speech engines — on-device when your macOS version supports it, otherwise via Apple's speech service under Apple's privacy policy.
- Crash recovery is local. During an active dictation, ListenIn writes a temporary audio copy and rolling raw-transcript journal to Application Support on your Mac. A normal finish or cancellation deletes that draft. After an interruption, recovery audio stays local; on launch, ListenIn prunes recovered recordings to the five newest.
- ListenIn never uses audio for training. Our cleanup server doesn't receive it.
Your text
- When AI cleanup is on, the recognized text — plus optional personal-dictionary terms, recent text context, and cleanup settings — is sent over HTTPS to our cleanup API, which forwards it to an AI provider — DeepSeek primarily, with Anthropic and Google as fallbacks — under their respective API terms. No audio is included.
- We don't store your transcripts. Our server keeps only anonymous word-count totals for quota (see below). No transcript database exists.
- Bring your own models: add any OpenAI-compatible or Anthropic API in Settings (local Ollama included) and your text goes directly from your Mac to that provider — our servers are bypassed entirely.
- Dictation history and your personal dictionary live only on your Mac. Delete them anytime from the app.
Identifiers & payments
- Free-tier and trial quotas are keyed to a random anonymous device ID — no account, no email, no name.
- Payments are processed by Polar as merchant of record. We never see your card details. Your license key is validated against Polar to unlock Pro.
- The website itself sets no tracking cookies.
macOS permissions, explained
| Permission | What it's for | What it's not |
|---|
| Microphone | Hearing you while you hold the trigger key. | Never records in the background; capture runs only during a dictation. |
| Speech Recognition | Apple's engine turns your speech into text. | — |
| Accessibility | Catching your trigger key (fn by default) system-wide, pasting cleaned text at your cursor, and — for a short window after each dictation — re-reading the text it just inserted so it can learn your corrections. | Only the trigger key is acted on — doesn't log keys, doesn't read other windows, doesn't watch your screen outside that window. Not a keylogger. |
Updates & crash reports
- Auto-updates use Sparkle with signed releases hosted on GitHub; the updater checks a public feed and never sends personal data.
- If crash reporting ships in a future version, reports will be scrubbed of all text content (no transcripts, no clipboard, no field contents) before leaving your Mac, and this page will say so explicitly.
Questions
Ask anything, publicly or privately: GitHub support. If this page and the app ever disagree, the app's actual behavior is a bug — tell us and we'll fix it.