Trie

Arrive as you finish speaking.

On your Mac, press the right Command key to start speaking, then press it again to stop. Trie transcribes and refines your words on your Mac, then pastes the result into the active text field.

Install Trie

The name

A trie is a tree of words. In it, how fast you reach an answer depends only on the depth of the question.

We want writing to work the same way: between a thought and its text, only the sentence you have to say.

How it works

Three steps

  1. 01

    Press

    Tap the right Command key (⌘) on its own and release to start recording.

  2. 02

    Speak

    Say what you want to write.

  3. 03

    Press again

    Press the right Command key (⌘) again. The refined text is pasted into the active text field.

You stay in control

You decide when to send.

Trie never presses Return for you and restores your previous clipboard contents after automatic pasting.

Switched windows
If you switch windows after stopping, or a password field is active, the text is copied to the clipboard instead of pasted automatically. If text editing cannot be completed, Trie keeps the original transcription.
To cancel
Press Command + period (⌘ .) while Trie is recording or processing your words.
Limits
Trie cancels if no speech is detected within 8 seconds of starting. Each recording is limited to 3 minutes; if you reach the limit, press the right Command key again to paste.

Privacy

It stays on your Mac.

  • Speech recognition and text editing happen on your Mac. Trie does not upload your audio or text.
  • Dictate offline once the models are ready. You initiate the initial model downloads; Trie does not download models automatically while you dictate.
  • Trie does not save recordings or transcripts. Diagnostic logs do not contain your text. Text is passed to the current app through the clipboard; that app controls how it saves or syncs the content.

Install

Installation and requirements

Mac
Apple silicon, macOS 14 or later.
Models
Before first use, download two models yourself, approximately 7 GB in total; running them needs at least 16 GB of memory. Downloading requires an internet connection; recognition and text editing work offline once the models are ready.
Permissions
Microphone and Accessibility permissions are required.

Version 0.3.2 Run the command below in Terminal to install. The app is signed with an Apple Developer ID; notarization is still being processed by Apple, so installation is offered from the command line only for now.

curl -fsSL https://trie.ai/install.sh | sh

Read the installation guide