Hold a key. Speak.
Your words appear.
Without your voice ever leaving your computer.
Dictovox is press-to-talk dictation for Mac and Windows. Hold ⌘⇧Space, speak, release — your words are typed into whatever is focused, in any app. Speech-to-text runs entirely on your machine, in English or any of 100 languages.
No account. No card. No cloud.
Private by architecture
Your voice stays on your machine.
Not as a promise — as a design.
Dictovox runs OpenAI’s Whisper model on your own CPU and GPU, via whisper.cpp. There is no server to trust, because there is no server.
No cloud
Transcription runs on your CPU and GPU. Turn on airplane mode — Dictovox keeps working.
No account
Download and go. No sign-up, no email, no telemetry, no analytics.
No subscription
Pay once. $15, one-time, per machine. Free updates, no renewal, nothing to cancel.
The only network requests Dictovox ever makes are one-time downloads of the speech models you choose.
How it works
Three seconds, start to finish.
Hold the hotkey
⌘⇧Space on Mac, Ctrl⇧Space on Windows. Change it any time in Settings.
Speak
A small floating pill shows Dictovox is listening. Say what you want to write.
Release
Your words are typed into the focused text field — right where your cursor is.
Works in every app: email, docs, chat, code editors, browsers. If you can type there, you can dictate there.
100 languages
Dictate in your language.
Or all of them.
Switch to a multilingual model in Settings and Dictovox understands all 100 languages Whisper supports — still entirely on your machine. Pick your language, or let auto-detect figure it out per dictation.
Choose the model that fits your machine: Base (English, fastest), Small (English or multilingual, more accurate) or Turbo (multilingual, most accurate). Models download once, on demand, and swap in without a restart.
Details
A small utility, done carefully.
Your hotkey, your choice
⌘⇧Space by default. Set any combination in Settings.
~0.3 s to text
On Apple Silicon, a spoken sentence appears about 0.3 seconds after you release the key. Metal GPU-accelerated, ~18× realtime.
Always know when it’s listening
A floating “Recording / Transcribing” pill appears while you hold the key — and warns you if the mic delivers no audio, instead of silently pasting nothing.
Clipboard preserved
Dictovox pastes your words, then puts your clipboard back exactly as it was.
Start at login
Optional. Dictovox waits quietly in the menu bar or tray until you hold the key.
Guided macOS setup
A built-in wizard walks you through the microphone and accessibility permissions.
Pricing
One price. That’s the pricing page.
- 7-day free trial — full features, no card, no account
- One license activates one machine
- Free updates
- 14-day money-back guarantee
Payments, VAT, invoices and refunds are handled securely by Stripe, our merchant of record.
Download
Free for 7 days. Fully offline after one small download.
On first run, Dictovox downloads the default English speech model once (141 MB). Multilingual models are optional downloads in Settings. After that, it never touches the network.
The trial is the full app — nothing held back.
macOS build is unsigned (no Apple Developer account — this is a small,
independent app). On first launch, right-click the app → Open →
Open. If macOS instead says the app "is damaged," open
Terminal and run:
xattr -dr com.apple.quarantine /Applications/Dictovox.app
FAQ
Questions, answered plainly.
Is my audio really never uploaded?
Yes. Dictovox runs the Whisper speech model on your own machine — audio is processed locally and never sent anywhere. You can verify this with a firewall: the only network calls Dictovox ever makes are the one-time downloads of the speech models you choose.
What happens after the 7-day trial?
Dictation pauses until you enter a license key. Nothing is deleted and your settings stay put — buy a license, paste the key, and continue.
How do refunds work?
14 days, no questions asked. Refunds are handled by Stripe, our merchant of record — just reply to your receipt email.
Which languages does Dictovox support?
All 100 languages Whisper supports. The default model is English-optimized (smallest and fastest); switch to a multilingual model in Settings, then pick your language or use auto-detect. Every language runs locally — nothing changes about privacy.
Do you offer a subscription?
No. Dictovox is $15, one-time, per machine. There is nothing to cancel.
What hardware do I need?
An Apple Silicon Mac, or a Windows 10/11 PC (x64). About 500 MB of disk space with the default English model — more if you add the larger multilingual models (up to 1.6 GB for the most accurate one) — and a microphone. Linux is not supported.