PipeVoice: Free Voice Typing for Windows, Cloud or Offline

Dictation software has a type: Mac-first, cloud-only, subscription-shaped. If you are on Windows and you just want to talk to your computer without renting the privilege, the shelf is surprisingly bare. I came across PipeVoice while digging through GitHub for free alternatives to the paid dictation apps, and it reads like someone built exactly the tool I had in mind: a quiet little tray app where you hold a key, talk, release, and your words land in whatever window you were using. No account. No subscription. Ever, says the README, in a tone that suggests the author has feelings about subscriptions.

Hold, talk, release

The core loop is deliberately boring, which is a compliment. PipeVoice sits in the Windows tray. You hold a hotkey, speak, let go, and the transcript is typed straight into the focused app: terminal, code editor, browser, chat box, anywhere. There is a toggle mode if holding a key feels too much like a walkie-talkie, and a clipboard mode that pastes instead of typing for apps that get nervous around synthetic keystrokes. A small overlay shows it is listening, with a live meter and the words forming as you speak.

It was clearly built by someone who dictates to AI coding agents all day, and it shows in the good ways. Per-app profiles let you say "Terminal: raw text plus Enter. Chat app: polished and auto-send. Editor: hands off my words." Spoken commands like "new line", "scratch that" and "send it" work out of the box, and you can teach it a vocabulary list so your project names stop coming back as interpretive poetry. Accent selection covers US, British, Australian, Indian and New Zealand English, which is more respect than most free tools pay to anyone outside California. A local history keeps your last fifty dictations with a copy button each, saved on your PC, for the moment you realize the best sentence of the day went into the wrong window.

Pick your privacy dial

Most voice tools force a choice: fast cloud or private offline. PipeVoice lets you pick per moment. Bring your own Deepgram or OpenAI key for fast cloud transcription, with Deepgram streaming words live as you talk. Or clean up the transcript for free using Google's Gemini free tier or free models on OpenRouter. Or go fully offline with a local Whisper model on CPU plus Ollama for cleanup, and nothing leaves your PC at all. Notably absent: the usual GPU and CUDA obstacle course that "local AI" tools love to inflict on Windows users. API keys stay in a local file on your machine, and there is no telemetry.

Bonus depth if you want it: a meeting recorder that captures your mic and your computer's audio as two separate streams, then turns the conversation into notes with named speakers and action items.

Where to be careful

Windows 10 and 11 only; everyone else is out. The cloud engines are bring-your-own-key, which means a signup and a meter running somewhere else, even if the tab is pennies. The free Gemini cleanup leans on Google's free tier continuing to exist, which is a strategy, not a guarantee. And the offline path means downloading a model and accepting CPU speed: private, free, and calm, but not instant. It is also a young project with a small GitHub following, so expect sharp edges between releases. The auto-updater is built in, at least, so keeping current is painless.

The verdict

PipeVoice is the rare dictation app that treats Windows as a first-class citizen and your data as your business. Free, open source under MIT, genuinely offline if you want it, and thoughtful in the small places that matter when you dictate fifty times a day. If you type for a living and your wrists have opinions, this is worth the three-minute demo on the site.

Official website: https://pipevoice.app/

Source code: https://github.com/SignalEngine/PipeVoice