Open Whisper transcribes on your computer. This page lists what stays there, the only things that use the internet, and how to watch it happen with your own network monitor.
What stays on your computer
- Your voice. The audio is held in memory while you dictate and discarded after the transcription. It is never written to disk.
- The transcription. It runs on your computer, with no network at all. Dictation works the same with the internet off.
- The text. It goes where your cursor is, or to the clipboard. Open Whisper keeps the last dictation in memory so Copy Last Transcription and Retry Last Transcription work. It is gone when you quit. There is no history.
- Video files on your computer. A file you transcribe is read where it is. Nothing is copied, and no network is used.
The only things that use the internet
Nothing on this list happens during a dictation.
| What | When | Where it goes |
|---|---|---|
| Downloading an optional model | Only when you choose a model that isn't bundled, in Settings or the video window. Once per model. | huggingface.co. The file is checked before it is used. |
| Transcribing a YouTube video | Only when you paste a video link and ask for it. | YouTube, to fetch the captions or the audio. The window says so before anything runs. |
| The Made by entringer.dev link in About | Only when you click it. | Your browser opens the page. The app itself sends nothing. |
| The Windows installer | Only during installation, and only if Windows lacks WebView2. | Microsoft, to download WebView2. |
For a video link, the audio is fetched by yt-dlp, a separate tool you install yourself. Once the audio is on your computer, the transcription itself uses no network. See Video and files.
No account, no analytics
- No account, no sign-up, no API key.
- No analytics, no telemetry, no crash reports.
- No update check. Open Whisper never calls home to look for a new version. New versions are announced on the Releases page.
- No log files. Open Whisper writes nothing about what you do.
The clipboard
To paste, Open Whisper puts the text on the clipboard and presses paste for you.
Mac
When the paste works, the text that was on the clipboard before is put back half a second later. Only text is put back: an image or other content you had copied may not come back.
When the text can't be pasted, it stays on the clipboard so nothing is lost, and a notice says Copied to clipboard — paste with ⌘V.
Windows
When the paste works, the text that was on the clipboard before is put back half a second later. Only text is put back: an image or other content you had copied may not come back.
When the text can't be pasted, it stays on the clipboard so nothing is lost, and a notice says Copied to clipboard — paste with Ctrl+V.
Every dictated text is kept out of the clipboard history (WinV) and out of cloud clipboard sync. That holds when the text stays on the clipboard too: it never appears in WinV, and it never reaches your other devices.
Transcripts you save
Dictations are never saved. A video transcript is saved only when you click Save or Save As… in the video window.
Mac
Save writes to ~/.open-whisper/transcripts/. That folder is hidden in Finder and is created the first time you save.
Windows
Save writes to %USERPROFILE%\.open-whisper\transcripts\. That folder is created the first time you save.
Save As… writes where you choose. The command line never writes a file. Every location the app uses is in Data and uninstalling.
What the permissions are for
Mac
- Microphone: used only while you dictate, never in the background.
- Accessibility: used only to send ⌘V to the app you're in. To decide whether to paste, Open Whisper asks macOS what kind of field has focus (whether you can type in it, whether it is a password field). It never reads your screen or what you type.
Windows
- Microphone: used only while you dictate, never in the background.
- There is no Accessibility permission on Windows. To decide whether to paste, Open Whisper asks Windows what kind of field has focus (whether you can type in it, whether it is a password field). It never reads your screen or what you type.
Check it yourself
You don't have to take this on trust. Watch the app's network traffic while you dictate.
Mac
- Open a network monitor, such as Little Snitch, or run
nettopin Terminal. - Find Open Whisper (its process is called
open-whisper). - Dictate a few sentences. No connection appears.
- For a second test, turn Wi-Fi off and dictate again. It works the same.
Windows
- Open Resource Monitor: type
resmonin the Start menu and press Enter. - Go to the Network tab and tick Open Whisper's process in the list, so only its traffic shows.
- Dictate a few sentences. No connection appears.
- For a second test, disconnect from the internet and dictate again. It works the same.