Speakhold

Hold a key, talk, let go. Your words land wherever your cursor is. Free, open source, and everything runs on your Mac.

macOS 14+ · Apple Silicon · MIT licence · no account, no subscription

Speakhold pill: live waveform while listening, then the transcribed sentence pasted

brew install --cask sturzael/tap/speakhold

Why people switch from Willow, Wispr Flow and Superwhisper

100% on-device

Speech-to-text runs with WhisperKit on the Apple Neural Engine. Works offline. Audio never leaves the machine.

Any hotkey chord

Left and right modifiers are different keys. Default is left ⌃ + left ⌥, which collides with nothing. Hold to talk or press to toggle.

Pastes anywhere

Slack, Gmail, Xcode, Terminal, Notes, the browser. Wherever the cursor is. Your clipboard is restored afterwards.

Native and tiny

One Swift binary, about 2 MB. No Electron, no Tauri, no background helpers. Menu-bar only.

Optional AI polish

Add a Claude API key and Speakhold removes fillers, fixes punctuation and spells names from your personal dictionary. Off by default.

Open source

MIT licensed. Read every line that touches your microphone. Issues and pull requests welcome.

Compared

	Speakhold	Willow Voice	Wispr Flow	Superwhisper
Price	Free	US$12–15/mo	US$12–15/mo	US$8.49/mo
Audio stays on device	Yes	Cloud by default	Cloud	Yes
Native Swift	Yes	—	—	Yes
Left/right-specific chords	Yes	No	No	Partial
Open source	Yes	No	No	No

Prices and features as of October 2026. Corrections welcome via pull request.

How it works

Speakhold captures your microphone only while the chord is held, converts the audio to 16 kHz mono, runs it through Whisper Large v3 Turbo in-process, and types the result into the frontmost app with a synthetic ⌘V. The model (about 630 MB) downloads once on first launch. The only network request the app can make is the optional Claude cleanup, and only with a key you provide.