VoxType turns your voice into polished text in any app — on Mac and Windows. Hold a key, speak naturally, release — done. Free, open source, and private by design.
macOS 13+ · Windows 10 & 11 · Free · No account · No subscription
Hold fn + ⌥ and think out loud — messy, unfiltered, half-finished. VoxType instantly turns it into a clear, structured AI prompt, ready for ChatGPT, Claude, or Cursor. Free, no API key needed.
Works everywhere you type — email, Slack, browsers, code editors, documents.
Hold fn + ⌥ and ramble your idea — VoxType turns it into a clear, structured AI prompt for ChatGPT, Claude, or Cursor. Free, no API key needed.
VoxType inserts text right at your cursor in whatever app is frontmost. No window switching, no copy-paste.
Filler words removed, sentences capitalized, punctuation added automatically. Say "press enter" to submit.
Select text, hold fn+Ctrl, and say "make this more formal" or "translate to Spanish." Or generate text from scratch.
Plug in your own OpenAI key and an LLM rewrites your rambling into clean prose. Entirely optional.
Teach it names and jargon so they're always spelled right. Say a cue like "my email" to expand full snippets.
Dictate in English, Spanish, French, German, Japanese, Chinese, Hindi, Arabic, and more.
Uses Apple's speech recognition — with an on-device option so audio never leaves your Mac. No accounts, no telemetry.
Every dictation saved locally so you can copy past ones. Stored only on your Mac, clearable anytime.
No subscription, no word limits, no upsells. Read the code, build it yourself, make it yours.
From install to first dictation in under two minutes.
Download the DMG, drag VoxType to Applications, and open it. It lives quietly in your menu bar.
Allow Microphone, Speech Recognition, and Accessibility — the onboarding window walks you through it.
Click into any text field, hold fn, and talk. A small HUD shows your words live as you speak.
Let go and polished text appears at your cursor. That's it — now go write that email.
Free forever. macOS 13 Ventura or later.
xattr -dr com.apple.quarantine /Applications/VoxType.app
Same voice typing, now on Windows 10 & 11 (64-bit). Speech recognition runs 100% offline on your PC via whisper.cpp — no audio ever leaves your computer.
VoxType.exe. If SmartScreen says "Windows protected your PC," click More info → Run anyway — VoxType is a free community app, not code-signed, and you only do this once. On first launch it downloads its speech model (~150 MB, one time); a small bar at the bottom of the screen shows progress — when it disappears, you're ready.
Quick answers to common questions.
Yes. It's open source (MIT). No account, no subscription, no word limits. The optional AI features use your own OpenAI API key, so you pay OpenAI directly only for what you use — everything else costs nothing.
To Apple's speech recognition service, the same one built into macOS — or nowhere at all if you enable "prefer on-device recognition" in Settings. VoxType has no servers and collects nothing. Transcripts stay in a local file on your Mac.
Two reasons: to detect the fn key being held anywhere in macOS, and to insert text into the app you're using. Both happen entirely on your Mac.
Go to System Settings → Keyboard → "Press 🌐 key to" and choose Do Nothing. You can also switch the trigger to Right ⌘ or Right ⌥ in VoxType's Settings.
Hold fn+Ctrl and speak an instruction. If text is selected, VoxType rewrites it ("fix the grammar," "make it shorter"). If nothing is selected, it generates new text ("write a polite decline to this invite"). Requires an OpenAI API key in Settings → AI.
VoxType works with a hold-to-talk key in any app, auto-removes filler words, expands snippets, enforces your custom dictionary, supports voice commands like "press enter," and adds AI-powered editing. Built-in dictation does none of that.
Please do! It's a small, readable Swift codebase (~2,000 lines, no dependencies). Open issues and PRs on GitHub.