Skip to content

Meet VocalFlow

VocalFlow is a native macOS voice input tool. Press a global hotkey in any app, speak, and a small floating window at the bottom of your screen shows your words transcribed in real time. Press the hotkey again, and AI-refined text appears at your cursor.

It turns speaking into a system-level input method. You do not switch apps, you do not copy and paste, and your workflow is never interrupted.

  1. Press the hotkey (default: a single tap of the right ⌘ key) in any text field — Slack, your code editor, a browser form, anywhere.
  2. Speak. A floating window appears above the Dock with a scrolling waveform and live transcription. Confirmed text shows in regular style; text still being decided shows in italic gray.
  3. Press the hotkey again. VocalFlow sends your speech through AI refinement — it removes filler words, repetitions, and slips, then fixes grammar and formatting.
  4. Text appears at your cursor, injected in one clean paste. Your document never flickers or shows half-finished text.

The VocalFlow floating window showing live transcribed text above a voice waveform The floating window during a dictation: live transcription above, waveform below. (App UI is currently Chinese; localization is planned.)

VocalFlow lives in your menu bar (no Dock icon). It stays silent in the background until you summon it with a hotkey.

VocalFlow does not run a small local speech model. Audio streams in real time over an encrypted connection to the VocalFlow service, which combines a state-of-the-art speech recognition engine with LLM-based text refinement. The result is much higher accuracy than local models can offer — especially for mixed Chinese-English speech and technical vocabulary.

Three principles protect your words along the way:

  • The first word is never lost. Audio capture starts the instant you press the hotkey; the network connection catches up in parallel.
  • Recording never stops because of network problems. If the connection drops, capture continues, and VocalFlow falls back to injecting the raw live transcript instead of making you wait.
  • Your cursor sees exactly one injection. No type-then-delete-then-retype churn in your document.
RequirementDetail
macOSmacOS 14 (Sonoma) or later
NetworkInternet connection required for transcription
AccountFree account, sign-in via email verification code

VocalFlow needs two low-level macOS capabilities to do its job: listening for global hotkeys system-wide, and injecting text at the cursor of other apps (via the Accessibility APIs). Both are incompatible with the Mac App Store sandbox. VocalFlow is therefore distributed directly as a signed and notarized DMG — see Installation.