Skip to content

Speech recognition that does not leave your machine, unless you ask it to.

If you only want voice input when the recording stays on your device, WhisPaste is built for exactly that: the default mode runs offline on your machine, there is no account, recordings are not kept after transcription, and the transcript history lives in a local SQLite database on your disk. Usage statistics and crash reports are on by default, never contain audio or text, and can be switched off.

WhisPaste is built to keep your voice private. By default, transcription runs locally on your own device with Whisper or Parakeet. A cloud provider is used only if you choose one, there is no account, and by default the only network traffic is usage statistics, crash reports and the update check. None of them is meant to carry your audio or text, and all can be switched off in Settings. Because it is open source, anyone can check these claims in the code.

Offline by default

After install, WhisPaste runs entirely on your machine: whisper.cpp transcribes the audio locally, the transcript is delivered to the cursor, and nothing is uploaded. You actively pick a cloud provider in the settings if you want one. There is no silent fallback that would send audio off-device because the local model is "too slow".

Direct-to-provider, never via our servers

If you opt into a cloud provider like OpenAI or Deepgram, the audio goes directly from your machine to that provider with your own API key. WhisPaste does not proxy or buffer the recording on any server we operate, because there is no such server. The legal terms for the audio are then governed by the provider you chose. See the linked privacy policy for the full picture.

No account, statistics you can switch off, local history

There is no sign-up and no login. Usage statistics and crash reports are on by default and can be switched off any time in Settings → Privacy. The statistics are cookieless, count anonymous screen views and feature use in aggregate, report each settings change as the setting name only, never its value, and use a random ID that changes every week, so your usage cannot be followed over time. They are designed to leave out your audio, transcripts and history. The app also checks for updates (can be switched off) and downloads speech models when you ask for them. The transcript history itself is stored in a local SQLite database under your user profile, so a privacy review only has to inspect one local file. For the legal text, see the privacy policy.

Does any of my audio leave the machine in the default setup?
No. In the default offline mode, whisper.cpp transcribes the recording locally and the audio is then discarded (unless you turn on audio retention yourself). The transcript goes into the local SQLite history. Usage statistics, crash reports and the update check may use the network (all can be switched off), but they are designed to leave out your audio, transcripts and history.
Do I need an account to use WhisPaste?
No. WhisPaste has no user account system at all. You install the app, configure a hotkey, and start using voice input. The Microsoft Store uses its own platform sign-in for the installation, but the app itself never asks you to log in or link an identity.
Where is my transcript history stored?
In a local SQLite database (history.db). Where that is depends on how you installed WhisPaste: %APPDATA%\WhisPaste for the direct download on Windows, ~/Library/Application Support/WhisPaste on macOS, ~/.config/whispaste on Linux; Microsoft Store installs keep it in the app's package folder under %LOCALAPPDATA%\Packages. Nothing is synced to a cloud. Optionally a PIN can protect the history, with auto-lock after inactivity. If you uninstall WhisPaste or delete that file, the history is gone: the file on your disk is the only copy.
What is the most private speech-to-text option?
A fully local, open-source tool is the most private choice, because your audio never leaves the device and the code can be checked independently. WhisPaste is built exactly this way: local processing, no cloud requirement, and usage statistics you can switch off that never carry your audio, transcripts or history.
What's the difference between the speech input built into Windows and Word and WhisPaste?
The speech input built into Windows and Word can send your recording to Microsoft's online services, so your audio leaves the machine. WhisPaste processes recordings locally by default with Whisper, works across every app (not just Word), and is open source instead of a black box.