Most voice-to-text tools send your microphone audio to a company's servers, require an account, and charge a monthly fee. You don't need any of that. Thanks to OpenAI's Whisper model running locally, you can get accurate dictation on Windows that is 100% offline, completely free, and private — your voice never leaves your PC.

This guide uses Sotto, a free, open-source (MIT) dictation app that runs Whisper on your CPU. No GPU required. Here's the whole setup.

Why offline (local) voice-to-text?

Privacy: audio is transcribed on your machine — nothing is uploaded, logged, or sold.
Free forever: no per-minute pricing, no subscription.
Works without internet: on a plane, in a secure environment, anywhere.
No account: nothing to sign up for; nothing to cancel.

What you need

A Windows 10 or 11 PC (a normal laptop CPU is fine — no GPU needed).
A microphone (a built-in laptop mic works).
About 300 MB of disk for the one-time model download.

Step-by-step

Download Sotto

Grab Sotto-windows-x64.zip from the latest release and unzip it anywhere.

Run Sotto.exe

Double-click Sotto.exe. Because the app is unsigned (code-signing certificates cost money; Sotto is free), Windows SmartScreen shows a warning — click More info → Run anyway. A small icon appears in your system tray.

Download the local model (one time)

Open Settings from the tray, choose a Whisper model, and click Download (~300 MB). This is the only network request Sotto ever makes. After it finishes, unplug the internet if you like — everything runs locally.

Dictate anywhere

Place your cursor in any app — Word, a browser, a terminal, chat, your IDE — then hold Ctrl + Win, speak, and release. Your words are typed at the cursor. Prefer hands-free? Turn on toggle mode to type live as you speak. 🎉

Tips for best accuracy

Use a decent mic and reduce background noise. Sotto has a voice-activity gate that ignores static, but a clean signal always helps.
Pick the model that fits your machine: smaller = faster, larger = more accurate. The default balances both on a typical CPU.
Speak in natural phrases rather than one word at a time — Whisper uses context to get punctuation and homophones right.
For English + Hindi/Hinglish, leave language on auto, or lock a language in Settings. See the Hindi voice typing guide.

Is this really free and private?

Yes. Sotto is MIT-licensed and open source — you can read every line on GitHub. There's no telemetry and no account. The transcription happens entirely on your CPU via OpenVINO-optimized Whisper. The only time it touches the network is the one-time model download you start yourself.

Get offline dictation now

Free, open source, 100% on your device.

⬇️ Download for Windows ★ Star on GitHub

How to Get Offline Voice-to-Text on Windows (Free, No Cloud)