Free & open source · Local Whisper · MIT
Press a hotkey. Speak. Your words land wherever the cursor waits — transcribed on your own hardware. No cloud. No subscription.
Linux · GNOME on Wayland — native. Windows 10/11 — beta. More on the way.
A quiet card shows what the model hears — without stealing focus. Release the key: the text is at your cursor.
Three times faster than typing.
Full context in one breath —
say everything you are thinking.
LINUX — HOTKEY, LIVE PREVIEW, PASTE INTO WHATEVER HAS FOCUS
Anywhere a cursor blinks. The same hotkey, every application.
Draft three replies in the time typing one used to take.
Slack, WhatsApp, Discord — answer at the speed of speech.
Born in a terminal: the author dictates prompts to coding agents all day.
Reports, essays, meeting notes — flow first, polish later.
Ideas evaporate faster than you can type. Not faster than you speak.
On Discord, your mic is muted to the call and the call is turned down — they never hear your prompts.
You speak about three times faster than you type.
One microphone key instead of a keyboard.
Your vocabulary steers the model — names land right, intent stays yours.
Audio and text never leave the phone — Whisper runs on it.
It is a system keyboard — it works in every single app.
THE ANDROID KEYBOARD — TAP, SPEAK, SENT
Other tools ship your voice to a server and bill you monthly. voiceflow keeps everything on your machine — by architecture, not by toggle.
That is the whole pitch: your thoughts reach the screen before they evaporate.
| voiceflow | typical paid dictation app | |
|---|---|---|
| price | free, MIT | ~$15/month |
| audio leaves your machine | never | always |
| Linux / GNOME / Wayland | native | unsupported |
| latency after you stop speaking | ~0.1 s on a GPU | network round-trip |
| works offline | yes | no |
| source code | all of it, on GitHub | closed |
Environment, hotkey, service, model — when the installer says done, dictation works. Updates: the same command again.
One step asks for sudo (input tooling and a udev rule — the script explains why). Then press Super + G, speak, press it again.
Prefer to read it first? View install.sh on GitHub.
Works in PowerShell and cmd, from any directory. Then press Ctrl + Shift + Space, speak, press it again.
No terminal? A double-clickable installer:
voiceflow-install.batA native menu-bar app, not the Linux daemon in disguise. Hold right ⌥, speak, let go — the text lands in whatever window has focus. Whisper runs on the machine; nothing is sent anywhere.
There is no installer, and that is not laziness: macOS ties the Accessibility permission this needs to a code-signing identity, and only you can supply yours.
Then pick your team under Signing & Capabilities and press ⌘R. Requires macOS 26. Full steps, and the two settings that stop a first build: macos/README.md.
Requirements, details and honest limitations: README · Windows notes · all releases
Good dictation should not cost fifteen dollars a month, nor require shipping your voice to a stranger's server.
voiceflow stays free and open source on every platform we can reach. The issues above carry implementation notes — grab one and go.