How to run Whisper on a Mac: free command-line tools and apps

You can run Whisper on a Mac in three ways: install OpenAI's free Python package, build a Mac-optimised version such as whisper.cpp or WhisperKit, or download a Whisper Mac app that hides the setup. All three can run the speech model on your own computer. Before you pick, decide which job you have. Transcribing an audio file and dictating live into other apps are different jobs, and most tools are built for one of them.

Publisher note: SuprFlow publishes this guide and is one of the products discussed. Competitor details come from each company's own pages, checked on 1 October 2026. Prices and features change; confirm before you buy.

We have not run a controlled benchmark of any option here, so this guide makes no speed or accuracy claims of its own.

Which job do you have?

A file tool will not type into your email for you. A dictation tool is not a workspace for a two-hour recording.

Option 1: How do you install OpenAI's Whisper from the command line?

This is the original, and it is free. The openai/whisper README says the code and model weights are released under the MIT License. You need to be comfortable in Terminal.

1. Install Python. The README says the codebase is expected to be compatible with Python 3.8 to 3.11.
2. Install ffmpeg. The README's line for macOS is:

brew install ffmpeg

3. Install Whisper:

pip install -U openai-whisper

4. Transcribe a file:

whisper audio.mp3 --model turbo

The README lists six model sizes (tiny, base, small, medium, large and turbo), four of them with English-only versions, "offering speed and accuracy tradeoffs". For speech that is not English, pass a language, for example --language Spanish.

The README has no Apple Silicon instructions beyond the ffmpeg line, and its relative speed figures were measured on an A100 GPU, so they do not tell you how fast it runs on your Mac. This tool works on files; it does not type into other apps.

Option 2: Is whisper.cpp the better choice on an M1 or later Mac?

For a command-line user on Apple Silicon, it is the option written with your hardware in mind. The whisper.cpp README describes a "plain C/C++ implementation without dependencies" and calls Apple Silicon a "first-class citizen - optimized via ARM NEON, Accelerate framework, Metal and Core ML". It also says "On Apple Silicon, the inference runs fully on the GPU via Metal." Supported platforms include "Mac OS (Intel and Arm)", so Intel Macs are covered as well as M1 and later. It is MIT licensed.

The README's quick start:

git clone https://github.com/ggml-org/whisper.cpp.git
cd whisper.cpp
sh ./models/download-ggml-model.sh base.en
cmake -B build
cmake --build build -j --config Release
./build/bin/whisper-cli -f samples/jfk.wav

What the README tells you to expect:

• WAV only. The whisper-cli example "currently runs only with 16-bit WAV files". Convert first: ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wav.
• Optional Core ML. The encoder can run on the Apple Neural Engine. The README says this needs extra Python packages and Xcode command-line tools.
• Microphone input is a demo. The whisper-stream tool is described as "a naive example of performing real-time inference on audio from your microphone". It is not a dictation app.

Option 3: What is WhisperKit?

WhisperKit is a Swift framework from Argmax that runs Whisper models on Apple devices. Its GitHub repository is now named argmax-oss-swift; the old WhisperKit address redirects there. The README describes it as "speech-to-text with OpenAI Whisper", lists macOS 14.0 or later and Xcode 16.0 or later as prerequisites for developers, and says the project is released under the MIT License.

Developers add it to an app as a Swift package. Command-line users can install its tool with Homebrew:

brew install whisperkit-cli

The README shows the command-line tool transcribing an audio file (wav, mp3, m4a or flac) and, with a --stream flag, transcribing from the microphone. It also says WhisperKit "automatically downloads the recommended model for the device if not specified." The README's examples are written for running from source, so check the tool's help for the installed syntax.

Mac apps are built on these projects. VoiceInk's README credits whisper.cpp, and SuprFlow uses WhisperKit for live dictation.

Option 4: Which Whisper Mac app should you use instead?

If Terminal is not for you, a packaged app handles the model download and setup.

Choose by job:

• You have recordings to transcribe: MacWhisper is the packaged choice. Superwhisper vs MacWhisper compares the two.
• You want free live dictation: start with Handy. Open-source Wispr Flow alternatives lists more, with licences and release dates.
• You want live dictation with extra modes, or a Windows or phone version: look at Superwhisper.

Where does SuprFlow fit?

SuprFlow is a Mac menu-bar app that uses a WhisperKit model for live dictation. Hold the Fn key, speak, release, and the text is inserted into the app you are using. It has a personal dictionary for names and acronyms, and voice snippets. Setup includes a one-time model download; see https://suprflow.app/requirements.

It does not transcribe audio files, so it is the wrong tool if you have recordings. It is proprietary, so you cannot inspect the source as you can with Handy or VoiceInk. It needs an Apple Silicon Mac. It costs $125 once, with no free tier or trial. If a free tool above does your job, use it. Best dictation app for Mac compares paid and free options, and Free Wispr Flow alternatives covers the no-cost ones.

How can you test your choice in 15 minutes?

1. Record yourself reading a made-up paragraph with a name, a date, an amount and a negation: "Jordan Chen confirmed the 14 May review. The fee is 2,750 dollars, not 2,570."
2. For a file tool: transcribe the recording with a small model and a larger one. Compare the name, the numbers and the "not".
3. For a dictation app: dictate the same paragraph into a note and into the app you write in most.
4. Turn off Wi-Fi and run it again if working without a connection matters to you.

Questions people ask

Is Whisper free on a Mac?

The model and OpenAI's code are free under the MIT License, and so are whisper.cpp and WhisperKit. Handy is a free app that runs Whisper models, and MacWhisper has a free version.

Does Whisper work on an M1 Mac?

Yes. The whisper.cpp README calls Apple Silicon a first-class citizen and lists both Intel and Arm Macs as supported. Individual apps set their own requirements: VoiceInk's README lists macOS 15.0 or later, and SuprFlow needs Apple Silicon and macOS 14 or later.

Can Whisper do real-time dictation on a Mac?

The command-line tools are built for files. whisper.cpp and the WhisperKit command-line tool include microphone streaming, but they print text in Terminal. For typing into other apps, use a dictation app such as Handy, VoiceInk, Superwhisper or SuprFlow.

Does running Whisper locally keep audio on my Mac?

The command-line tools run the model on your computer. With an app, check its settings: MacWhisper, for example, offers both local and cloud models, and some dictation apps have optional cloud formatting. Where does my dictation go? lists the questions to ask.

Related guides

• Superwhisper vs MacWhisper
• Open-source Wispr Flow alternatives
• Offline voice to text on Mac
• Best dictation app for Mac
• SuprFlow vs MacWhisper
• How to use voice to text on a Mac

Sources

• OpenAI — whisper GitHub repository and README
• ggml-org — whisper.cpp GitHub repository and README
• Argmax — argmax-oss-swift (WhisperKit) GitHub repository and README
• MacWhisper — official site
• MacWhisper Support — MacWhisper and Whisper Transcription Difference
• Handy — GitHub repository
• VoiceInk — GitHub repository
• Superwhisper Docs — Introduction
• Quickstart
• SuprFlow — System requirements