OpenWhispr Setup: Free Voice Dictation for Any App

OpenWhispr is a free, open-source voice-to-text dictation app for Windows, macOS, and Linux that pastes your speech at the cursor in any app, using local Whisper or Nvidia Parakeet models so your audio never has to leave your machine. It is climbing GitHub's daily trending list as the self-hosted answer to paid dictation tools like Wispr Flow, and the whole install-to-first-dictation path takes about five minutes with no account required.

• Ships as a one-click installer for macOS, Windows, and Linux, with a four-command source build for anyone who wants to run it from a checkout instead.
• Transcription runs fully offline with local Whisper or Nvidia Parakeet models, or you can opt into OpenWhispr Cloud or your own OpenAI/Groq key for speed.
• A global hotkey dictates into any app on the system, plus a separate AI-agent hotkey that can edit selected text or answer a spoken command with GPT-5, Claude, or a local model.
• MIT licensed and free, with no telemetry claimed in the README; team spaces and cloud sync are the parts that need an account.

The exact steps, start to finish

1. Step 1. Check your machine is ready.

   # macOS
   sw_vers
   
   # Windows (PowerShell)
   [System.Environment]::OSVersion.Version
   
   # Linux
   uname -a
   
   # Only needed if you plan to build from source instead of using an installer
   node --version

   OpenWhispr ships prebuilt installers for macOS, Windows, and Linux, so most people do not need Node at all; it only matters for the source-build path in Step 7.
2. Step 2. Download the installer for your OS. Grab the right file from the GitHub Releases page: a .dmg for macOS, an .exe for Windows, or an .AppImage, .deb, .rpm, or .tar.gz for Linux.
3. Step 3. Install and launch it. Open the installer and follow the prompts. On first launch, OpenWhispr starts an onboarding wizard that walks through the rest of the setup below.
4. Step 4. Choose your processing method. The wizard offers three paths: Local (download a Whisper or Parakeet model once, then transcribe with no internet connection and no account, the option most tutorials like this one default to), OpenWhispr Cloud (sign in with Google, Apple, Microsoft, or email for instant transcription), or bring your own key (paste an OpenAI or Groq API key). Picking Local keeps the whole setup account-free.
5. Step 5. Grant permissions. Approve the microphone prompt, which is required on every platform. On macOS, also grant Accessibility (needed for OpenWhispr to auto-paste text at your cursor) and Screen Recording (only needed if you plan to use meeting transcription).
6. Step 6. Start dictating. Press your hotkey, Globe or Fn on macOS, Ctrl+Win or Ctrl+Super on Windows and Linux, speak, then press it again. Your words appear as text wherever your cursor is focused. That is the whole loop working end to end.
7. Step 7 (optional). Build from source instead.

   $ git clone https://github.com/OpenWhispr/openwhispr.git
   $ cd openwhispr
   $ npm install
   $ npm run dev

   This needs Node.js 24 or newer and skips the packaged installer entirely, useful if you want to inspect or modify the Electron app before trusting it with your microphone.

What is OpenWhispr and why is it trending?

OpenWhispr bills itself as the open-source and free alternative to Wispr Flow and Granola: a menu-bar app that turns speech into text anywhere on your desktop and, separately, into meeting transcripts and notes. It is built on Electron with a React and TypeScript front end, uses better-sqlite3 for local storage, and leans on whisper.cpp and sherpa-onnx to run Whisper and Nvidia's Parakeet ASR models directly on your CPU or GPU. The pitch that is driving stars is the same one that keeps recurring in this Tutorials series: a category of paid, closed-source, cloud-dependent tools now has a free, MIT-licensed, fully local alternative that a solo developer can audit and self-host. Beyond plain dictation, OpenWhispr adds a named AI voice assistant that can rewrite selected text or answer spoken commands using GPT-5, Claude, Gemini, Groq, or a local model, plus Zoom, Teams, and FaceTime meeting transcription with on-device speaker labeling.

RelatedVoicebox Setup: A Local AI Voice Studio in 15 Minutes

How do you install OpenWhispr on Windows, macOS, and Linux?

Every platform has a packaged build on the Releases page. Windows users want the NSIS installer, roughly 241 MB. macOS ships separate Apple Silicon and Intel builds as .dmg files, each a little over 330 MB; note that on Intel Macs, live speaker identification and voice fingerprinting are unavailable because they depend on ONNX Runtime, which stopped shipping macOS x86_64 binaries. Linux gets four formats to choose from, .AppImage, .deb, .rpm, and .tar.gz, all north of 400 MB since each bundles its own Electron runtime and ASR engine. If you would rather run from source on any OS:

# requires Node.js 24+
$ git clone https://github.com/OpenWhispr/openwhispr.git
$ cd openwhispr
$ npm install
$ npm run dev

Whichever path you take, the first launch always drops you into the same onboarding wizard: pick a transcription method, grant the microphone (and, on macOS, Accessibility and Screen Recording) permissions, and you are ready to press the hotkey.

Local or cloud: which processing method should you pick?

Local is the option this tutorial defaults to because it needs no account and no recurring cost: OpenWhispr downloads a Whisper or Parakeet model once, and every dictation after that runs offline, with GPU acceleration available on Metal, CUDA, and Vulkan. OpenWhispr Cloud trades that offline guarantee for speed and requires signing in with Google, Apple, Microsoft, or email. Bring-your-own-key sits in between: you paste an OpenAI, Groq, or other provider's API key and pay that provider directly, without creating an OpenWhispr account or routing audio through OpenWhispr's own servers. All three methods use the same hotkey and the same paste-at-cursor behavior once configured, so switching between them later is a settings change, not a reinstall.

How does it compare with Wispr Flow and Granola?

Wispr Flow and Granola both compete on polish and zero-setup cloud convenience, and neither publishes source you can audit. OpenWhispr's case is the same one Meetily made for meeting notes in this series: if you want dictation and meeting transcripts that never have to leave your laptop, or you simply do not want another subscription for a feature this small, OpenWhispr covers the same core loop for free.

RelatedMeetily Setup: Free Local AI Meeting Notes in 10 Minutes

What are the gotchas before you rely on it?

A few things to know before OpenWhispr becomes your default input method. First, local models are a real download and a real compute cost: bigger Whisper variants are more accurate but slower without a capable GPU, so thin laptops should start with a smaller model and upgrade only if accuracy disappoints. Second, on Intel Macs you lose live speaker identification and voice fingerprinting entirely, a hard limitation from ONNX Runtime dropping x86_64 support, not a bug in OpenWhispr itself. Third, the global hotkey, Ctrl+Win or Ctrl+Super by default on Windows and Linux, can collide with window-manager shortcuts on some Linux desktop environments, so check your DE's own bindings if the hotkey seems to do nothing. Fourth, features like team spaces, cloud sync, and enterprise SSO or SCIM provisioning sit behind an account even though core dictation does not, so read the settings screen before assuming every feature is fully local.

Our take

OpenWhispr is another entry in a pattern this Tutorials series keeps documenting: a paid, closed-source SaaS category getting a free, local-first, open-source challenger built on the same whisper.cpp and llama.cpp lineage that made Meetily and dozens of other tools possible this year. What makes it worth five minutes of setup is that the local path genuinely needs nothing, no account, no API key, no data leaving your machine, and it still gets you the same hotkey-to-text loop Wispr Flow charges a subscription for. The AI-agent hotkey and meeting transcription are the features to watch; if those mature at the same pace as the core dictation, OpenWhispr stops being just a free alternative and starts being the default recommendation.

Original analysis by GenZTech. Tool documentation: OpenWhispr/openwhispr on GitHub.