The short version
Audio is processed on your Mac while you hold the dictation key. There is no Murmur account, no Murmur server, and no telemetry. The audio buffer is discarded when you release the key. If you choose a cloud engine, audio is sent directly to that provider using the key you supplied. Murmur never receives it.
What Murmur processes on your Mac
Microphone audio. While you hold the Right Option key, Murmur reads audio from your selected microphone. In Apple On-device mode, Apple SpeechAnalyzer transcribes that audio locally. Murmur does not create a recording, a temporary audio file, or a recovery cache. The buffer is held in memory and discarded when the session ends.
Transcripts. The finished text is inserted at the cursor in the app you are using. Murmur has no account system and no Murmur server that receives transcripts.
Settings and dictionary. Preferences, language settings, and project words are stored locally on your Mac. If you connect a cloud engine, the key is stored in the macOS Keychain, not in Murmur's preferences file.
Cloud engines are optional
Apple On-device is the default. If you add your own key for Google Gemini or Meta Muse, the audio for the current session and the custom vocabulary hints you enabled are sent directly to that provider. The provider processes the request under its own privacy policy and bills you directly. Murmur does not proxy, store, or mark up that traffic.
If you enable final polish with OpenRouter, only the finished transcript text is sent. Audio is not. Remove the key from Murmur to stop using that provider.
History and local files
In the current build, transcript history is enabled by default. It is stored as a local JSON file inside Murmur's application data folder and contains transcript text, the engine and app name, and session timing. Audio is never part of the history.
You can turn history off, search it, or delete individual entries and the whole history from Murmur's Settings. History is protected by your macOS account's file permissions; Murmur does not add a separate encryption layer. If you use a shared Mac, turn history off or remove the app data when you are done.
Permissions
Murmur asks for three macOS permissions, and only uses them for dictation:
- Microphone — to hear the words you are dictating.
- Input Monitoring — to notice the Right Option key while Murmur is not the active app.
- Accessibility — to insert the finished text into the app where your cursor is.
You can revoke any of these in System Settings → Privacy & Security. Without them, dictation will not work.
This website
The Murmur website uses PostHog for aggregate product analytics: page views, referrers, browser and device information, and a download_clicked event when you press a download button. That event records the button location, product version, platform, and file size. PostHog is configured to use localStorage rather than cookies, with autocapture, heatmaps, performance capture, and session replay disabled. It does not create a cross-site advertising profile. Audio, transcripts, API keys, and email addresses are never sent to analytics.
The site stores your light or dark theme preference in your browser's local storage so the page does not flash on your next visit. Email links open your default mail app. The hosting provider may process standard request logs for security and reliability.
Your choices
- Keep Apple On-device selected to keep audio entirely local.
- Turn history off or delete it from Murmur's Settings.
- Remove a cloud key at any time to stop that path.
- Revoke microphone, Input Monitoring, or Accessibility access.
- Delete Murmur's local application data folder to remove history and settings.
Changes to this policy
If Murmur's data handling changes, this page will be updated with a new date. Material changes will also be noted in the app's release notes.
Contact
Questions about privacy can go to contactvishaw@gmail.com. Murmur is made by Vishawjeet Sidhu.