Hold a key, talk, release — your words appear right where your cursor is. Email, Slack, your IDE, your browser. Cleaned up by AI before they land.
Free · no account · $8/mo for our cloud, if you want it
A native app for Windows 10 and 11. Windows first, on purpose — macOS is next.
The built-in local recognition engine runs fully offline, and it is free without any limit. In cloud mode, audio is processed and not stored afterward.
The whole application is free — on the on-device engine or your own API key, for as long as you like. Would rather not hold a key at all? $8 a month puts our cloud behind it instead, and you can stop whenever you like. Either way: no sign-up, no password, no email in anyone's database. There's nothing to leak.
One key of your choosing — hold it to talk, or tap it to toggle.
VoxDrot listens while the key is down.
The text is already in your app, right at the cursor.
Prefer hands-free? Tap once to start recording, tap again to stop.
VoxDrot lives in the system tray. The window is where you pick a mode, edit its prompt, choose the key and read back what you said — the dictation itself happens wherever your cursor is, and the window never takes focus from it.
Native .NET, not a browser in disguise — and it follows your Windows theme, light or dark.
The prompt is the slow part. Hold the key, describe what you want, release — it lands in the terminal, the chat box or the editor, punctuation included.
Raw mode keeps your exact words, so nothing "helpful" rewrites a command. Clean up removes the umms. And your clipboard is put back afterwards, so the thing you copied a minute ago is still there.
Speaks your language: the on-device Parakeet engine covers 25 European languages, whisper.cpp covers 99 — from German, French and Spanish to Chinese, Japanese and Arabic.
Types into any Windows application — not into its own little window. If you can put a cursor there, you can dictate there.
No app to open, no record button to find. Hold-to-talk or tap-to-toggle, over any program, from the system tray.
AI post-processing cleans your words before they land: fix grammar, strip the "umms," translate to another language, soften the tone, or run your own prompts ("format as a business email").
Free, no card and no account. Download and dictate — the on-device engine needs nothing set up, and with a subscription there are no API keys either.
Pasting with Ctrl+V normally eats whatever you had copied. VoxDrot snapshots the clipboard, pastes the dictation, and puts the original back the moment the paste has landed.
Already have an API key? Plug in OpenAI, Groq, or xAI (Grok) and pay the provider directly — your key stays encrypted on your device.
$0 forever
Nothing to accept but the EULA, which the app shows once.
$8 / month
or $69 a year — 28% less
Checkout opens soon
By subscribing you accept the EULA and the Terms of Service. Billed monthly or yearly by Polar, our merchant of record. The founder price stays while your subscription renews without a break. Cancelling drops you to the free tier; nothing is deleted.
Twenty minutes of our cloud a week is about 2,500 words — a few long emails. The on-device engine and your own keys have no limit at all, and never will.
VoxDrot is built so that trust doesn't depend on a policy page. The architecture does the promising.
Bring your own key: OpenAI, Groq, or xAI (Grok) — your choice.
No. The local engine works fully offline. Cloud mode and AI post-processing need a connection.
In local mode — nowhere; recognition happens on your machine. In cloud mode, audio is sent for processing and is not stored afterward. With BYOK, audio goes directly to the provider you chose.
Yes — anywhere a cursor blinks: Windows Terminal, PowerShell, VS Code, Cursor, Claude Code, JetBrains IDEs. Use Raw mode there so nothing rewrites your words.
No. VoxDrot is a native .NET application. It lives in the system tray, starts with Windows only if you let it, and does not run a browser.
The on-device Parakeet engine covers 25 European languages: English, Dutch, French, German, Italian, Polish, Portuguese, Russian, Spanish, Swedish, Ukrainian and 14 more. The whisper.cpp engine covers 99, including Chinese, Japanese, Korean, Hindi, Arabic and Turkish. The cloud engines cover at least as many.
Nothing breaks. The twenty free minutes a week are our cloud's, and they come back every Monday. The on-device engine and your own API keys are never limited — switch to either and keep dictating. Subscribe and our cloud is unlimited too.
The free app runs on as many machines as you like. A subscription covers up to 5 devices; a machine you no longer use can be released from Settings, so a full list never needs a support ticket.
Full refund within 30 days, no questions asked. Email us or use the order link in your Polar receipt.
Windows 10 and 11 (64-bit).
Because that is where the good native dictation app was missing. macOS comes after Windows is solid — email us from the Contact page and you will hear when it is there.
The installer is signed by Kazkai Tech AB — right-click it, Properties, Digital Signatures. SmartScreen warns about any new publisher until enough people have installed its software; click "More info", then "Run anyway". In local mode the app makes no network calls at all, which any network monitor can confirm.
Yes — OpenAI, Groq, or xAI (Grok). Your key is encrypted locally with DPAPI and never sent to our servers.
Free · no account · our cloud $8/mo, 30-day money-back guarantee