What you’ll have at the end
A small app in your system tray. Click into any text box (WhatsApp Web, Gmail, ChatGPT, Word), press a hotkey, speak, and your words are typed at the cursor. With a local model, your voice never leaves your computer and it costs nothing to run.
Written from OpenWhispr’s official docs (docs.openwhispr.com) and GitHub page, read on 17 Nov 2026. Menu names can change between versions.
What you need
- Windows 10 or later (64-bit), a Mac, or Linux.
- A microphone. Your laptop’s built-in mic works.
- About 5 minutes, plus the time to download a model.
Setup in 5 steps
Download the installer
Go to openwhispr.com, or the repo’s Releases page, and download the file for your computer:
- Windows: the
.exeinstaller - Mac: the
.dmg(works on Apple Silicon and Intel) - Linux:
.AppImage,.deb,.rpmor.tar.gz
On a Mac you can also install it from Terminal:
brew install —cask openwhispr- Windows: the
Install and open it
Run the installer and follow the prompts. On Windows it adds a desktop and Start Menu shortcut, then sits in the system tray (bottom-right, near the clock).
If Windows blocks itThe app is signed by Gizmo Labs Inc. New apps sometimes still trigger SmartScreen or antivirus warnings. Only continue if you downloaded it from openwhispr.com or the official GitHub Releases page.
Choose how it turns voice into text
On first launch, a short onboarding asks you to pick one of three options:
- Local (recommended for privacy and free use): download a model. Orukeet is the recommended one. Whisper, Parakeet, Nemotron and Cohere Transcribe are also there. After the download, it works with no internet.
- Bring your own key: paste an OpenAI, Groq or other API key. It’s faster on slow computers, but you pay that provider.
- OpenWhispr Cloud: sign in with Google, Microsoft, Apple or email.
Allow the microphone
Windows: there’s nothing to grant inside the app. If no audio is picked up, open Settings → Privacy → Microphone in Windows and allow it.
Mac: allow Microphone, plus Accessibility (so it can paste text for you). Add Screen recording only if you want meeting notes.
Press the hotkey and talk
Click into any text box, then:
- Windows: press Ctrl + Win, speak, and press again. You can also hold it while you speak and let go to stop.
- Mac: press Globe / Fn.
- Linux: Ctrl + Super or F8, depending on your desktop.
Your words appear where the cursor is.
Things to try first
Write a WhatsApp reply to a customer: we received your order, it ships tomorrow, thank you.
Say it in your own words. OpenWhispr types it, and you just fix the small mistakes.
- Meeting notes: it can record Zoom, Teams and FaceTime calls and turn them into notes. (The new Microsoft Teams client doesn’t always share its audio. OpenWhispr’s docs cover the workaround.)
- Voice assistant: a separate hotkey sends what you say straight to an AI (GPT, Claude, Gemini or a local model) and pastes the answer.
- Translate while you dictate: speak in one language and paste in another.
If something doesn’t work
| Problem | Fix |
|---|---|
| Nothing gets typed | Check the microphone permission (step 4). Make sure the cursor is inside a text box. |
| Wrong language or bad accuracy | Try a different local model, or a bigger one if your computer can handle it. |
| Slow on an old laptop | Use a smaller local model, or switch to a cloud model with your own API key. |
| Hotkey does nothing | Windows reserves shortcuts like Win+E and Win+L. Pick another one in Settings. |
Removing it
Uninstall it from Settings → Apps on Windows. The uninstaller also removes the downloaded models.
Back to the full list: 4 open-source alternatives to paid AI tools