FREE FOREVER · 100% LOCAL AUDIO · OPEN SOURCE

Speak. It types.
Your voice stays on your Mac.

OpenVoiceFlow is a free push-to-talk voice dictation app for macOS. Hold a key, talk, release — polished text lands in any app.

Download for Mac See how it works
Signed & notarized · macOS 14+ · Apple Silicon & Intel · no account, ever

Fifty-two seconds. The whole idea.

Speak at 150 words a minute. Type at 40. Watch what that gap is worth.

Original score. On-screen text mirrors real app behavior — on-device Whisper, cleanup Off by default.

USP 01 · FREE FOREVER
$0

No subscription. No account. MIT open source.

USP 02 · EVERYTHING LOCAL
0 bytes

of audio ever leave your Mac. Airplane mode? Fine.

How it works

01 · Hold fn Globe
Hold, speak, release. One key, any app.
02 · Speak
Whisper transcribes on-device.
03 · Release
Polished text at your cursor
Full pipeline →
Your voice never leaves your Mac.
Not “encrypted.” Not “anonymized.” Not sent. Audio is transcribed on your Mac and discarded the moment text exists.
Not a promise — an architecture. The code is MIT-licensed open source.
YOUR MAC — AUDIO NEVER LEAVES THIS BOX
mic→ Whisper→ cleanup→ your cursor
With Ollama or cleanup off, no byte crosses this border at all.

Everything you need. Nothing you don't.

Live HUD
See it listening
A waveform tracks your voice while you hold the key.
Profile
Know-Me interview
Learns your name, team, and jargon.
Dictionary
Your words
Teach it the names and jargon it should never misspell.
Context
Per-app style
Casual in Slack, formal in Mail.
Docs
A real manual
Twenty-five pages covering every setting, model, and fix.
Snippets
Say it, expand it
Voice-triggered text expansion.

Free forever

Wispr Flow is $144 a year for someone else's cloud.

$0
Annual cost
$144
Saved every year
MIT
License
∞
Cleanup models via OpenRouter
OpenVoiceFlow
Wispr Flow
Price
$0, forever
$12/mo — $144/yr
Transcription
On this Mac
Their cloud
Works offline
Yes
No
Audio leaves the machine
Never
Uploaded
Source code
Open — MIT
Closed
Account required
No
Yes
AI cleanup
Your choice
Their pipeline

Competitor details per their published pricing and docs, July 2026.

Talking is how humans work.
Typing is the workaround.

People speak at roughly 150 words a minute and type at roughly 40. A Stanford study measured speech input at about three times the speed of typing on mobile keyboards. Your voice has always been the faster interface — the tools just hadn't caught up.

~150
Words per minute, speaking
~40
Words per minute, typing
~3×
Speech vs. keyboard, Stanford study

Our mission

In the AI era, talking to your computer is a basic need — and basic needs should never be gated by payment. Don't pay companies $144 a year for your own voice: dictation here is free for everyone, on-device, MIT-licensed, forever. The only optional extra, AI cleanup, is bring-your-own-key — pay the model provider cents for tokens, not a middleman for permission.

Honestly, we think every operating system should ship this as a default feature, and the day that happens well enough, apps like this one become unnecessary. Until then, we build it in the open — MIT-licensed and free for everyone, forever.

Read the full mission →

Pick your backend.

Fully local to ultra-fast API. You decide what runs where.

OpenRouterCLOUD · RECOMMENDED
One key, any modelaudio stays local
One OpenRouter key reaches any model it hosts — the simplest way in. Bring your own key.
OpenAI · Claude · Groq
Your own keyaudio stays local
Prefer a provider directly? Point cleanup at OpenAI, Anthropic, or Groq instead.
OllamaON-DEVICE
$0 / forevernever leaves Mac
Fully offline cleanup with any local model.
OffRAW TRANSCRIPT
$0 / forevernever leaves Mac
Raw on-device transcript, pasted as-is. The default.

All options: audio transcription is always on-device. Only cloud cleanup touches an API — and only if you turn it on.

Up and running in 60 seconds.

1

Download the DMG

Notarized — no Gatekeeper warnings.

2

Drag to Applications

That's it. No installer wizard.

3

Say hello

One ~5-minute setup, then talk. Quickstart →

From the blog

Guides and honest comparisons from the people who build this.

COMPARISONS · 9 MIN The best dictation software for Mac in 2026, honestly compared Six real options — including ours — sorted by where your audio goes, how you pay, and what happens after transcription. HOW-TO · 8 MIN How to dictate on a Mac — and when you'll outgrow the built-in way Turn on macOS dictation in two minutes, learn the commands worth memorizing, and see where it runs out of road. WORKFLOWS · 7 MIN Voice typing for developers: prompts, PRs, and reviews AI pair programming turned your main output into prose. Speak the English; keep the keyboard for the code.

All posts →

Questions

Really free? What's the catch?

MIT-licensed, no account, no telemetry. Your Mac provides the compute — there is nothing to sell you.

How accurate is it?

Whisper on-device, models from tiny to large. Your personal dictionary fixes names and jargon once, forever.

Does AI cleanup send my text to the cloud?

Cleanup is Off by default, so the raw on-device transcript is pasted as-is. Turn on cleanup and pick OpenRouter (one key, any model), a direct provider — OpenAI, Anthropic, or Groq — with your own key, or local Ollama for fully on-device. Audio never goes anywhere, regardless.

Which Macs?

macOS 14 Sonoma or newer. Apple Silicon flies; Intel works with smaller models.

Where does my data live?

Settings and history stay in Application Support on your Mac, and API keys live in the macOS Keychain. Uninstall the app and remove those to erase every trace.

Start dictating for free.
Right now.

No account. No credit card. No cloud.

Download for Mac — Free Install guide

macOS 14+ · Sonoma → Sequoia · Apple Silicon & Intel