OpenWhispr and PipeVoice are both free, MIT-licensed voice tools that run on Windows. OpenWhispr is the broader app: dictation plus notes, meeting transcription and an agent. PipeVoice is a focused push-to-talk tray app for Windows that types into any app. Here is where each is stronger.
Quick take: Want notes, agent features and Mac or Linux support alongside dictation? Look at OpenWhispr. Want a small Windows-first tool built around talking into the app you already have open, with per-app profiles? Try PipeVoice.
What OpenWhispr is
From the OpenWhispr GitHub README: it is MIT-licensed, built with Electron and React, and supports macOS (Apple Silicon and Intel), Windows and Linux. It can run local models (Whisper, NVIDIA Parakeet and others) where "your audio never leaves your device", or use cloud speech-to-text. Beyond dictation with a global hotkey it lists AI agent integration (GPT-5, Claude, Gemini, Groq, local models), meeting transcription with speaker diarization, notes with semantic search, audio and video import, a public API and an MCP server. It showed about 8.8k stars when we checked on 2026-09-29.
What PipeVoice is
PipeVoice is a Windows 10 and 11 tray app, MIT-licensed. Hold a hotkey, speak, release, and the text lands in the focused window as real keystrokes (or a clipboard paste, if you prefer). Engines: Deepgram (streaming), OpenAI Whisper, and Local Whisper for fully offline use. It also records meetings as two separate streams (your mic and your computer's audio), and an opt-in Agent MCP server exposes two local tools, listen and transcribe.
OpenWhispr vs PipeVoice: comparison
| Feature | PipeVoice | OpenWhispr |
|---|---|---|
| Price | Free | Free (README lists optional cloud sync for signed-in users) |
| License | MIT | MIT |
| Platforms | Windows 10/11 only | macOS, Windows, Linux |
| Offline | Yes (Local Whisper + Ollama) | Yes (local models) |
| Cloud engines | Deepgram, OpenAI Whisper (your own key) | Cloud or local speech-to-text |
| Meeting transcription | Yes, mic and computer audio as separate streams | Yes, with speaker diarization |
| Notes and search | No | Notes with semantic search |
| MCP server | Yes, opt-in (listen, transcribe) | Yes |
| Per-app profiles | Yes | Not stated in the README summary we checked |
| Account | No account, no telemetry | Optional sign-in for team features |
Where OpenWhispr wins
- Scope. Notes, semantic search, audio and video import, and an agent are in the same app.
- Platforms. It runs on Mac and Linux as well as Windows.
- More local model options, including Parakeet, per its README.
Where PipeVoice wins
- Focus. One job: talk and it types, into a terminal, editor, browser or chat box.
- No account. Nothing to sign in to, and no telemetry.
- Per-app behaviour. Give the terminal raw text plus Enter, give chat auto-send, and give the editor no cleanup.
- Streaming. With Deepgram, words appear live in the overlay as you speak.
If dictating to AI coding tools is your main use, start with dictating into Claude Code and Cursor or voice coding on Windows.
Try it: Download PipeVoice for Windows, see the PipeVoice vs Wispr Flow comparison, or read the Wispr Flow alternative for Windows guide.