MacWhisper Launches CLI Tool! One-Click Automated AI Speech-to-Text in Terminal

MacWhisper, the acclaimed macOS AI speech recognition app widely used for subtitling and transcription, today released features that many advanced users have been eagerly awaiting:Command-line tool (CLI)The new MacWhisper version 13.20 includes a built-in mw a command line tool that allows users to directly perform speech-to-text in Terminal, manage models, and even integrate transcription results into automation scripts. This update means MacWhisper is no longer just a pretty GUI app, but further becomes a powerful AI transcription “engine” in the macOS ecosystem that can be integrated into various workflows by developers, podcasters, journalists, and researchers.

What is MacWhisper?

MacWhisper is a native macOS speech-to-text application developed by independent developer Jordi Bruin, powered by OpenAI’s Whisper model (via whisper.cpp and WhisperKit) and NVIDIA’s Parakeet model. Its greatest advantage is:

• Fully offline operationAll speech recognition is performed locally — no files are uploaded to the cloud.
• Apple Silicon OptimizationFully utilizes Mac’s Neural Engine (ANE) and CoreML for acceleration.
• Multi-model supportCan freely switch between Whisper models of different sizes, NVIDIA Parakeet, and the built-in Apple speech models on macOS 26+.

MacWhisper 13.20 is out now with full CLI support!

Transcribe and stream media files using any of the local or cloud models, connect to agents and scripts, and use LLMs to refine your transcripts and pass them into your workflows.

Full documentation: https://t.co/ngDkhAOMCu pic.twitter.com/iDnuzWKFRH

— Jordi Bruin (@jordibruin) April 23, 2026

Install mw CLI tool

Installation is very simple – you can do it directly through MacWhisper’s graphical interface:

1. Open MacWhisper → Settings → Advanced
2. at Command-Line Tool Paragraph Click Install
3. MacWhisper will automatically mw Binary file placed to /usr/local/bin/mw
4. After installation is complete, the screen will display a green checkmark: “installed at /usr/local/bin/mw”

When first installing, macOS may ask you to enter a password, as it needs write permissions to /usr/local/bin Menu. To remove it, go back to the same settings page and click Uninstall.

Verifying the installation is simple. Open a new terminal window and enter:

$ mw version
MacWhisper 13.19.2 (1409)

If you see a version number, it means the installation was successful.

Core Instructions Overview

mw Provide four main sub-commands, covering the core requirements of transcription work:

just input mw Shows the complete help summary; add after any subcommand --help You can view more detailed options (such as mw transcribe --help）。

Admin: List and Switch Models

mw Supports multiple speech recognition engines, including OpenAI Whisper, NVIDIA Parakeet, and Apple’s native speech model:

$ mw models list
  ID                                        NAME                      SIZE
  whisper-cpp:ggml-tiny.en                  Tiny (English Only)       80 MB
  whisperkit:openai_whisper-small           Small                     483 MB
▸ parakeet-pro:nvidia_parakeet-v2_476MB     Parakeet v2 (476MB)       476 MB
  apple:en-GB                               English (United Kingdom)  -

• ▸ The marker indicates the model currently in use
• ID field (such as whisperkit:openai_whisper-smallThat’s you --model or mw models select The name to pass to the command
• apple:<locale> Items are Apple on-device speech models available on macOS 26 and later, with storage managed by the system.

Switching models also only requires one line of code:

$ mw models select whisperkit:openai_whisper-small
Selected: whisperkit:openai_whisper-small (Small)

The switch takes effect immediately, affecting both the CLI and MacWhisper’s graphical interface.

Transcribe audio files: the core feature

The most basic usage is to directly pass an audio or video file:

$ mw transcribe ~/Desktop/meeting.m4a
Hello everyone, welcome to the meeting...

The default for transcription results is Transient:Output only to the terminal and then discard, without storing in MacWhisper’s history. This is very useful for processing large amounts of temporary files or privacy-sensitive content.

Practical Parameters

mw transcribe Supports multiple useful flags:

Pipeline Integration

The true power of CLI lies in its ability to chain with other Unix commands:

# 轉錄後直接統計字數
mw transcribe meeting.m4a | wc -w

# 轉錄後進行翻譯（需先安裝 translate-cli）
mw transcribe meeting.m4a | trans -b :zh-TW

# 轉錄後用 Glow 渲染 Markdown
mw transcribe meeting.m4a | glow -

Underlying Technology and Operating Principles

mw The CLI tool communicates with the running MacWhisper application via a local socket, so it uses exactly the engine, model, and parameters you have already configured in the GUI. If called mw If MacWhisper isn’t already running, the CLI will automatically launch it for you.

This means:

• You don’t need to install Python, PyTorch, or any other dependencies.
• The transcription quality is completely consistent with the graphical interface
• Model management (downloading, updating) still needs to be done through the MacWhisper UI

Celebrate Our Launch: 10% Off Pro Version

To celebrate the launch of the CLI tool, developer Jordi Bruin is offering a 10% discount on MacWhisper Pro (there’s also a free version). The Pro version offers batch transcription, higher accuracy models, and longer audio support, making it ideal for professional users.

MacWhisper Official Website

Source: KOCPC Chinese