Molvi
Don't type — speak
Hold a key, say it out loud — and the text appears right where your cursor is. Whisper large-v3 recognition runs right on your computer: no cloud, no subscription, no data leaks.
Windows 10/11 · macOS on any M-chip (M1, M2, M3, M4…) · open source (MIT) · no sign-up · guide: Windows · Mac
Your voice is processed on your PC and never sent anywhere.
Free and open source. No accounts, no payments.
The same model top-tier services use — but entirely yours.
Types into an editor, browser, messenger — anywhere.
Three moves, and you're done
No windows or buttons while you work. One key, your voice, and text.
Hold the key
Left Ctrl by default. You can assign any combination in settings.
Speak
A pill-shaped overlay shows a recording is in progress. Speak naturally, at your own pace.
Release
Molvi recognizes your speech and types it straight into the active window, wherever your cursor is.
Thought through down to the details
Everything you need for everyday dictation — and nothing you don't.
Privacy
Audio never leaves your computer. Works even without internet.
Any hotkey
The dictation key is configurable — from a single Ctrl to your own combination.
Microphone and language
Choose your recording device and language: Russian, English, or auto-detect.
GPU and CPU
With an NVIDIA GPU — maximum quality; without one — a fast model on the CPU.
Overlay and sounds
See and hear when recording is happening and when text is being recognized.
Launch at startup
One checkbox, and Molvi is always ready in the tray once your PC boots.
Built for Claude Code and Codex
Long prompts are faster to say than to type. Hold the key, describe the task in your own words — Molvi types it for the agent.
Works right inside the terminal: Molvi detects the console window and types character by character — even where a regular Ctrl+V paste breaks.
Molvi vs. cloud dictation
Molvi does what Wispr Flow and Aqua Voice charge a subscription for — locally, privately, and for free.
| What matters | Molvi | Wispr Flow, Aqua Voice |
|---|---|---|
| Where voice is processed | On your PC | On the company's servers |
| Price | Free | Subscription |
| Audio privacy | Never leaves your device | Uploaded to the cloud |
| Works offline | Yes | No |
| Open source | Yes, MIT | No |
Start dictating in a couple of minutes
Windows 10/11 x64 (maximum quality with an NVIDIA GPU with 6+ GB) or a Mac with any M-series chip: M1, M2, M3, M4 and newer — recognition runs through Metal, even a base M1 with 8 GB is enough. Intel Macs are not supported. First launch detects your hardware and downloads the model automatically.
Molvi runs on enthusiasm
The project is free — no ads, no subscriptions, no fees. If Molvi saves you time, support its development with a donation: it helps ship updates.
Payments via Tribute · one-time or monthly