hushkey Download free
Push-to-talk for your AI

Talk to your AI.
Stop typing to it.

Hold a key, say what you need, let go — your words land in ChatGPT, Claude, Cursor or whatever window has focus. Speaking is about three times faster than typing, so your prompts get the detail that makes answers good.

  • Free & open source
  • Runs offline
  • No account
  • Windows · macOS · Linux
Why voice

Better answers start with better prompts.

AI does its best work when you give it context: what you tried, what you want, what to avoid. Typing all that is tedious, so most prompts end up short. Saying it takes seconds.

~3× faster

Say more, get more

People speak around 150 words a minute and type around 40. A detailed prompt is half a minute of talking, not five of typing.

Works where your cursor is

Chat apps, coding assistants, the terminal, your editor, email — hushkey types into the focused window. No new app to switch to.

Your voice stays on your machine

Speech recognition runs locally. Nothing you say is uploaded — only the finished text goes where you put it.

  • ChatGPT
  • Claude
  • Gemini
  • Copilot
  • Cursor
  • Claude Code
  • VS Code
  • Terminal
  • Slack
  • Teams
  • Email
  • any app with a text cursor
How it works

Hold. Speak. Release.

No window to open, no button to click. hushkey sits in your tray and waits for one key.

  1. Hold your key

    Caps Lock by default — or Right Alt, Right Ctrl, Scroll Lock and others, picked in the tray menu.

  2. Speak naturally

    A small overlay shows hushkey is listening. Ramble if you like — over 100 languages are understood, and the language is detected for you.

  3. Let go

    The text is pasted into the focused window, and your clipboard is put back the way it was.

Private by design

No cloud. No subscription. No account.

hushkey runs OpenAI's open Whisper speech model on your own computer, on the graphics card when there is one. After the one-time model download it needs no internet at all.

Offline

Works on a plane, behind a firewall, with the network cable pulled.

Fast on your hardware

Uses NVIDIA, AMD and Intel graphics on Windows and Linux, and Apple Silicon on the Mac — the CPU otherwise.

Small and self-updating

One program, no Python, no setup. It keeps itself up to date — or not, if you switch that off.

Recent dictations at hand

The last transcripts stay in the tray menu, ready to copy again.

Download

Get hushkey — it's free.

One download, about a minute to set up. The speech model (about 1.6 GB) is fetched on first start.

Your system
Windows

Windows 10 or newer, 64-bit. No admin rights needed; it sets up start at login itself.

Download for Windows
Your system
macOS

macOS 13 or newer, Apple Silicon. On the first start, right-click the app and choose Open (see FAQ), then allow Microphone and Accessibility.

Download for macOS
Intel Mac
Your system
Linux

Ubuntu 22.04+, Debian 12+ and alike; X11 or Wayland. Open the package, or sudo apt install ./hushkey_*.deb.

Download .deb (64-bit PC)
ARM & plain binary

On macOS and Linux the one-line install saves a step — it clears macOS' quarantine flag and sets up keyboard access on Linux:

curl -fsSL https://raw.githubusercontent.com/aignermax/hushkey/master/install.sh | bash

Every file with its checksum: latest release on GitHub.

Questions

Good to know

Does anything I say leave my computer?

No. Recognition runs locally. The only downloads are the speech model on first start and update checks against GitHub, which you can switch off in the tray menu.

Is it really free?

Yes. No trial, no subscription, no account. hushkey is open source under the MIT license — the code is on GitHub.

Does it only work with AI chats?

No — it types into any window with a text cursor: chat apps, editors, terminals, browsers, email, office documents. AI prompts are simply where talking instead of typing pays off most.

Which languages?

Everything Whisper understands — over 100 languages, detected automatically, or fixed to one in the tray menu. German and English are tested the most.

Do I need a graphics card?

No. A graphics card makes it faster and is used when present; otherwise the processor does the work.

Can I change the key?

Yes, in the tray menu. Caps Lock is the default because few people miss it; Right Alt, Right Ctrl, Scroll Lock and a few more are offered too.

macOS says the app cannot be opened — what now?

macOS quarantines anything downloaded from a browser, and hushkey is not signed with an Apple developer certificate. On the first start, right-click (or Control-click) hushkey in Finder, choose Open and confirm — after that it opens normally. The one-line install avoids this. macOS then asks for Microphone and Accessibility access: one to hear you, the other to see your key and type for you.

I used the older hushkey — do I need to do anything?

No. The older version updates itself to this one and brings your settings and history along. Afterwards the tray menu offers Remove the old hushkey version, which clears out what the old one left behind (a few gigabytes).

How do I remove it?

Windows: through the app list. Linux: sudo apt remove hushkey, or ~/.local/bin/hushkey --uninstall for a plain-binary install. macOS: hushkey --uninstall. Your settings and the downloaded model stay unless you delete them.

Your next prompt could take ten seconds.

Install hushkey, hold Caps Lock, and just say it.

Download free