Push to talk
A physical keypress makes the recording boundary explicit. Capstan listens only when you ask it to.
macOS preview available
Hold a key, speak, and release. Capstan turns your voice into cleaned-up text and places it wherever your cursor already is.
A physical keypress makes the recording boundary explicit. Capstan listens only when you ask it to.
Speech recognition can run locally with Apple SpeechAnalyzer or Parakeet, without sending every utterance to a server.
The result lands at the current cursor, so your documents, messages and tools remain where the work lives.
Formatting and personal dictionary corrections turn raw recognition into text that is ready to use.
No new editor, inbox or account-shaped workspace. Capstan is an input instrument, not another destination.
Recent dictation can be replayed or repaired without retaining a permanent archive of your microphone audio.
Choose any editable field in the application you already use.
Keep the push-to-talk key down for exactly as long as the utterance lasts.
Capstan transcribes, cleans up and inserts the finished text at the cursor.
The macOS app is used in daily work. This preview disk image is signed with a Developer ID, notarized by Apple and stapled for offline verification.
PLATFORM 1
PLATFORM 2
MACOS RELEASED 30 August 2026 · 9.8 MB · SHA-256499c6637cafeda863908e67f7db26ff7f05a06e7075861eed82405cfb55e77f0
Windows is implemented and exercised in CI, but nobody has yet held the key and spoken into a real Windows microphone. It will not be presented as a working download until that pass is complete.