Tagara

Speak. I’ll type.

Voice dictation for Mac. Hold ⌥, speak, release: the text appears wherever your cursor is. Runs entirely on your computer. Free, no account.

Download for Mac
curl -fsSL https://raw.githubusercontent.com/mateusfonsek/tagara/main/scripts/get.sh | bash

macOS 15 or later · Apple Silicon Mac (M1 and up)

Three steps. None of them is typing.

Works in the terminal, Slack, WhatsApp, your browser and any text field.

  1. Hold ⌥

    The right Option key becomes a talk button. Shortcuts with ⌥ keep working.

  2. listening…typing…

    Speak

    Your way, in English or Portuguese. You can even mix both in one sentence.

  3. Speak. I’ll type.

    Release

    In about 0.3 seconds the text shows up at your cursor.

Small on the outside. Smart on the inside.

Works offline

Once the model is downloaded, dictation runs on your Mac’s Neural Engine. On a plane or in a café with no Wi-Fi, it just works.

Actually fast

A sentence becomes text 0.2 to 0.4 seconds after you let go of the key.

English and Portuguese

The model detects the language on every dictation, so mixed sentences come out right.

Hands-free

Double-tap ⌥ to dictate long texts without holding the key. One tap finishes.

In any app

Types wherever your cursor is and puts your clipboard back the way it was.

History that stays yours

Your last 50 dictations stay on your Mac, searchable, so you can copy back what got lost. You can turn it off.

Nothing leaves your Mac.

Audio stays in memory while you speak and is gone right after. No server, no account, no telemetry. And the code is open, so you can check every line.

Read the code on GitHub →

Your Mac

  1. ⌥ key
  2. Microphone
  3. Neural Engine
  4. Text at your cursor

No catch.

$0

No account. No subscription. Open source, MIT license.

What Tagara doesn’t do

  • It doesn’t rewrite what you said with AI. You get what you said.
  • It doesn’t record meetings or transcribe files.
  • It doesn’t send anything to the cloud. Not even if you ask.

Made for people who live in the terminal.

Install with one line, update with the same line, and dictate straight into Claude Code. Native Swift, no Electron.

curl -fsSL https://raw.githubusercontent.com/mateusfonsek/tagara/main/scripts/get.sh | bash

Downloads the latest version, installs it in Applications and opens it.

View on GitHub →Build from source →

FAQ

Will it run on my Mac?

You need an Apple Silicon Mac (M1 or newer) with macOS 15 or later. Intel Macs aren’t supported, because the model runs on the Neural Engine.

Why does it ask for Microphone and Accessibility?

The microphone is to hear you while the key is down. Accessibility is to notice ⌥ and type into the app in focus. None of it leaves your Mac.

macOS says it can’t open the app. Now what?

Tagara isn’t signed with an Apple developer certificate yet. If you install from the Terminal, the warning doesn’t show up. From the DMG, allow it in System Settings → Privacy & Security → Open Anyway.

How much space does it take?

About 650 MB, almost all of it the voice model. It’s downloaded once, the first time you open the app.

How do I update?

The app tells you when a new version is out and updates itself in one click. You can also run the install command again.

Is there a Windows, Linux or iPhone version?

Not for now. Tagara is built for the Mac and uses features that only exist there.

How much does it cost?

Nothing. Tagara is free and open source, under the MIT license.

Hold ⌥ and go.

Free, light, no account. Takes a minute to install.

Download for Mac
curl -fsSL https://raw.githubusercontent.com/mateusfonsek/tagara/main/scripts/get.sh | bash