Hold ⌥ option + ⌃ control, talk, let go. WhisperTool types it at your cursor, in any app.
Your cursor is here.
Hold the keys and watch. Your own ⌥ + ⌃ work here too.
Filmed with version 1, which ran Whisper on the Mac's own chip. Version 2 keeps the same keys, adds the live island, and is far lighter:
A real capture. While you hold the keys, a black island drops from the notch and shows your words as you say them. If the internet fails, it pastes what your Mac heard; if everything fails, the audio is kept for a retry.
A beep tells you when it's listening. The microphone is only on while you hold.
Mail, Slack, Notion, your code editor, even apps in full screen. The text lands where your cursor is.
Spanish stays Spanish, English stays English, switched per phrase automatically.
With your free Groq key, an AI pass removes ums and self-corrections, fixes punctuation, and can format a dictated email.
Apple's on-device speech recognition. Nothing leaves your Mac.
On your Mac works offline
The audio of each dictation goes to Groq's Whisper (large-v3-turbo). The text comes back in about 0.3 s, then gets tidied.
0.3 s per phrase
No account with me, no server of mine.
About a minute. Step 2 happens only the first time. For the best accuracy, paste a free key from console.groq.com in Settings.
Open WhisperTool.pkg and click Install. It goes into Applications and opens by itself.
macOS asks for Microphone (to hear you), Speech Recognition (for the live text) and Accessibility (for the hotkey and to paste). A waveform icon appears in the menu bar. Click into any text box, hold ⌥+⌃ and talk.
Signed and notarized by Apple (Developer ID Ignacio San Benito), so it opens with a double-click, no warnings. The full source code is on GitHub, and you can check that your download is exactly the file I built. In Terminal:
shasum -a 256 ~/Downloads/WhisperTool.pkg
should print d0ef18a7f8f1a5aa5d7af9ef400b4e544c506b39f9690c65faa006b1919478c1
git clone https://github.com/nachoie123/whispertool.git && cd whispertool/mac && ./build.sh