Developers
Dictate comments, commit messages, and prompts without leaving the keyboard home row.
v1.0 · macOS & Windows$0Completely free, forever — no subscription
Talk. It types.
Nothing leaves your Mac or PC.
Hold a key, speak, release — your words are transcribed locally and pasted right at your cursor, in any app. No account, no network, no audio ever sent anywhere.
100% on-device No account No telemetry Open source
Hold fn on Mac · Left Ctrl on Windows. Free & open source.
Cursor — untitled
Hold to talkon-device · 0 ms uploaded
Types straight into the apps you already use
VS CodeSlackGmailNotionTerminalGoogle DocsFigmaChatGPTObsidianLinearMessagesWordCursorDiscord
Three seconds, start to cursor.
No window to focus, no “record” button. It lives in your menu bar / tray and listens only while you hold the key.
- 01
Hold the key
Press and hold fn on Mac or Left Ctrl on Windows. The mic opens instantly — only while held.
- 02
Speak naturally
Talk at a normal pace. A waveform shows it’s hearing you. Every sample stays on the device.
- 03
Release to insert
Let go. Whisper transcribes locally and the text is pasted right where your cursor is — formatted and ready.
One small tool. A lot of range.
Click through what it actually does.
main.pyREADME.mdnotes.txt
def summary():
"""Return a short, plain-language recap of the run."""
...The whole point
Your voice never leaves the machine.
There’s no server to send audio to, because there is no server. Transcription runs on your own CPU/GPU with Whisper. No account, no API key, no telemetry. The only thing that ever touches the network is an optional update check (version & OS — nothing else), which you can turn off.
No network callsNo accountNo telemetryOpen source
Faster than your fastest typing.
Speaking runs about 150 words per minute; sustained typing about 40. Transcription is near-instant on Apple Silicon.
Typing
~40 wpm
Dictation
~150 wpm
0–0×
faster than real time (Apple Silicon)
0
bytes of audio uploaded, ever
0+
languages, auto-detected
∞
free — open source, no subscription
Apple Silicon runs whisper-large-v3-turbo on the GPU via Apple MLX. Intel Macs and Windows use faster-whisper on the CPU (with a snappier default model). The model is warmed at launch, so your first take is already fast.
Built for people who’d rather talk.
Writers
Draft at the speed of thought, then edit. Spoken lists come out as real lists.
Researchers
Capture notes and ideas the moment they land — privately, with searchable history.
RSI & accessibility
A genuinely fast typing alternative that never sends your voice to the cloud.
How it compares.
Most dictation either ships your audio to a server or only half-works outside one app.
| VocaYou’re here | Cloud dictation | Built-in OS | |
|---|---|---|---|
| Audio never leaves your device | varies | ||
| Works fully offline | partly | ||
| Pastes into any app | partly | ||
| No account or sign-in | |||
| Remappable push-to-talk key | varies | ||
| Spoken lists → formatted lists | |||
| Private on-device history | |||
| Free & open source | — |
Does my audio leave the device?
Never. It’s transcribed locally and discarded. The only network call is an optional update check (version + OS).
Which key do I hold?
fn on macOS, Left Ctrl on Windows (fn can’t be detected there). Both are remappable.
Why the security warning on launch?
Builds aren’t signed yet. Right-click → Open on Mac; “More info → Run anyway” on Windows.
Stop typing what you could say.
Free, open source, and private by design. v1.0.1.