MacWhisper

At a glance

MacWhisper is worth considering if you want audio or video turned into editable transcripts and subtitles on a Mac, with local models as an option. Its value goes beyond dictation: you can review text, export captions and automate files through a command-line interface. Choose the model deliberately, because optional cloud processing changes where the audio goes.

Good fit

Mac users working with interviews, lectures, podcasts or recorded meetings who want text they can correct and reuse. The direct-download edition combines the desktop editor with mw, a CLI that talks to the Mac app over a local socket. That can fit an agent or script workflow without turning the product into a standalone Linux transcription server.

Consider something else when

You need a cross-platform, headless deployment or assume all features are offline. Whisper.cpp offers a source CLI for Linux and other platforms, with its own build and model setup. MacWhisper’s cloud transcription, AI services and optional export integrations can send content outside the Mac.

Start here

Start with the free edition and a non-sensitive short recording. Choose a downloaded local transcription model, drag the file into the app, then compare the text with the audio before exporting TXT or subtitles. For automation in the direct-download edition, the support guide installs mw from Settings → Advanced. mw version checks connectivity; mw models list shows installed models before mw transcribe example.wav produces text.

Cost and requirements

The official site lists Free at €0 and a direct-download Pro license at €64 with lifetime updates, checked October 31, 2026. The storefront lists macOS 14 or later, recommends an M-series Mac and also supports Intel Macs. Larger models need more memory. App Store licensing is separate; cloud/API services can add recurring costs. Structured JSON output through mw requires Pro.

Limits to know

The CLI uses the app’s configured engine and can launch the Mac app automatically; it is not an independent inference service. CLI transcriptions are transient unless --persist is supplied. The older privacy page says no audio or text leaves the device, while current support explicitly describes cloud uploads. Apply the local-processing claim only to the chosen local workflow. No app, model, transcription, speaker detection or export was run for this assessment.

Related work

• Whisper.cpp: a source CLI for local transcription across platforms (alternative)