MacWhisper for Mac: Features, Pricing, Honest Review

MacWhisper for Mac: Features, Pricing, Honest Review

You've got a 90-minute interview recording sitting on your desktop. The client wants a transcript by tomorrow, and the idea of typing it out by ear makes you want to close the laptop and go for a walk. Someone says "just use MacWhisper," and now you're wondering whether that's the right call.

This guide covers MacWhisper for Mac from top to bottom: what it actually does, how the free and Pro versions differ, the two confusingly similar versions that exist, what it costs, how accurate and fast it is, where it falls short, and when a different kind of tool (like voice dictation) will serve you better.

What MacWhisper is, in one paragraph

MacWhisper is a Mac app that turns audio and video into text using OpenAI's Whisper speech models, and it can do that entirely on your own machine. You drag in a file, pick a model, wait a bit, and get an editable transcript with timestamps. That's the core of it.

It was built by a solo developer, Jordi Bruin, and it grew from a small experiment into one of the best-known transcription apps on the Mac. If you want the very short version of the concept, we've written a plain-English answer to what MacWhisper does. And if the name "Whisper" itself is fuzzy, there's a separate explainer on what Whisper is.

After a few weeks of using it, one thing stood out: MacWhisper for Mac is primarily a file transcriber. It also does dictation and live captions, but I kept reaching for it when I had a recording, not when I wanted to write an email. If you picture yourself talking into your Mac all day, that's a different job, and I'll come back to it.

What you can do with MacWhisper for Mac

MacWhisper handles a wider range of inputs than the name suggests. The Mac App Store listing (last updated November 5, 2026) says it supports audio and video imports, live captions, YouTube and media links, batch transcription, transcript search, translation, and exports to DOCX, PDF, SRT, VTT, CSV and JSON. That's a long list, so let's go through the parts you'll actually touch.

Transcribing files

This is the bread and butter. Drop in an MP3, WAV, M4A, MP4 or MOV and you get text. Interviews, lectures, podcasts, voice memos, old Zoom recordings. It works with all of them.

Speed depends on your Mac and the model, and I'll give real numbers in the speed section below. For now, the practical advice: start with a mid-sized model, check the result, and only go bigger if the errors annoy you.

Recording and transcribing meetings

MacWhisper's support documentation says it can record audio from individual apps or from the Mac's system audio, and that it can detect meetings in Zoom, Teams, Webex, Skype, Slack and Discord. So you can capture a call without a bot joining and announcing itself to everyone.

That's a real selling point. Plenty of people hate the "Notetaker has joined the meeting" moment, and a local recorder sidesteps it. A quick courtesy though: tell people you're recording. Depending on where you live, it may be a legal requirement too (some US states require every party's consent, and most of the EU is stricter still).

One practical note from testing: system audio capture only grabs what your Mac plays. If you're on a call with headphones, your own voice comes from the microphone, so you'll want both sources selected. Forget that and you get a transcript of everyone except you.

Paste a link and MacWhisper pulls the audio and transcribes it. Handy for research, for grabbing quotes, or for turning a 50-minute conference talk into notes you can search. Check that you have the right to use the content before you republish anything, obviously.

Subtitles and exports

Exporting SRT and VTT is what makes MacWhisper popular with video editors and creators. You get a subtitle file you can load into Final Cut, Premiere, DaVinci Resolve or upload straight to YouTube. For documents, DOCX and PDF cover most needs, and CSV or JSON help if you want to process transcripts elsewhere, say in a spreadsheet or a script that tags quotes by speaker.

Subtitle timing is good, not perfect. On a 12-minute video I had to nudge about half a dozen lines where a long sentence got split awkwardly across two cards. That's a two-minute fix, but it's a fix.

Speaker labels, filler words and editing

The MacWhisper website advertises speaker labeling, automatic removal of filler words, and transcript search and editing. Speaker labeling (diarization) is the difference between a wall of text and something you can actually read after a two-person interview. It isn't perfect, so expect to fix a few "Speaker 2" mix-ups, especially where people interrupt each other, but it beats doing it all by hand.

Filler word removal is the sleeper feature. Stripping out the "um" and "you know" from a transcript makes quotes usable without you touching each one.

Dictation

The website lists system-wide dictation among its capabilities, and the Gumroad version includes it. It works. But it's a feature of the app rather than the reason the app exists. More on that below.

AI features

MacWhisper also layers AI on top of transcripts: summaries, questions about the transcript, rewriting and custom prompts. The company says you can use its optional Assistant, connect an AI provider with your own API key, or run a compatible model locally. The local route keeps everything on your Mac. The cloud routes, naturally, send text to a provider.

I used the summary feature on a 40-minute team call and got a tidy bullet list of decisions. It missed one action item that was mentioned in passing. That's typical of AI summaries: good for orientation, not a replacement for reading the parts that matter.

The two versions of MacWhisper for Mac (and why people get confused)

There are two products from the same developer: MacWhisper on the developer's website (sold through Gumroad) and Whisper Transcription on the Mac App Store. They overlap heavily, but they aren't identical, and the pricing models differ.

The developer's documentation has a page explaining the difference between MacWhisper and Whisper Transcription, and it's worth a skim before you buy anything. The short version:

MacWhisper (website) Whisper Transcription (App Store)
Name you'll see MacWhisper Whisper Transcription
Pro pricing One-time €64 Subscription or lifetime in-app purchase
Where you get it Developer's site Mac App Store
Updates Direct from developer Through the App Store
Best for People who hate subscriptions People who prefer App Store billing

The App Store listing shows the app is free to download, with Pro in-app purchases at $6.99 (monthly), $29.99 (yearly), and lifetime options of $89.99 or $99.99. Exact prices can vary by purchase option or storefront, so check your own region before you commit.

Which should you pick? If you want to pay once and be done, the website version at €64 is the cleaner deal. If you like having billing, refunds and family sharing handled by Apple, the App Store version is easier. I lean toward the one-time purchase for a tool you'll keep using for years, because the math favours it quickly (more on that in a moment).

One more wrinkle. Because they're separate apps, a license bought on one doesn't carry to the other. If you start on the App Store free tier and later buy on the website, you're installing a second app, not upgrading the first. Delete the one you stop using so you don't wonder which has your old transcripts.

Free vs Pro in MacWhisper for Mac: what you actually get

The free version is genuinely useful, and the website advertises free use of local models, which surprises people. You don't have to pay just to try it on a real file. Pro is for people who transcribe often or need the workflow features.

Here's where I can be specific about what I checked. In the app I confirmed batch queueing, SRT and VTT export, speaker labeling and the larger model downloads, and each of those sat behind the Pro prompt in the version I tested. I couldn't confirm exactly where system-wide dictation and meeting detection are gated, because that seems to differ between the website build and the App Store build. The release-notes page on Releasebot shows updates through November 5, 2026 but doesn't itemise the changes, so I can't tell you what moved between releases. Check the paywall screen on your own copy before assuming a feature is free.

When free is enough

If you transcribe a couple of voice memos a month and just want clean text, stay free. Seriously. Don't pay for features you won't open.

When Pro makes sense

Upgrade when you hit one of these walls:

  • Volume: you're transcribing several files a week and batch processing would save you hours.
  • Video work: you need SRT or VTT subtitle files regularly.
  • Interviews: speaker labels matter because you quote people.
  • Meetings: you want to record calls and get notes without a bot.

The break-even math

Say you'd otherwise pay a cloud transcription service around $10 to $20 a month. A €64 one-time license pays for itself in roughly three to six months. Compared to the $6.99 monthly App Store plan, the lifetime options pay back in 13 to 14 months. If you're a heavy user, the lifetime route wins; if you need Pro for one project, monthly is cheaper.

A cloud service also charges per minute at the top end, so a researcher with 30 hours of interviews a month can easily burn through that $20 figure. For that person the one-time license is almost an accident of arithmetic.

Is it worth it compared with cloud transcription?

For most people with regular recordings, yes: a local app wins on price over a year and on privacy from day one. Cloud services still win in a few spots, and it's only fair to say where.

Cloud tools tend to bundle things MacWhisper for Mac doesn't try to be: shared team workspaces, comment threads on a transcript, and a link you can send a colleague who doesn't own a Mac. If five people need to review the same transcript, a web tool is simpler. If it's just you and a folder of recordings, a local app is cheaper and quieter.

There's also the matter of waiting. A cloud service uploads your file, queues it, and emails you. A local run starts the second you drop the file in, and on a modern Apple silicon Mac the small model is quick enough that you'll often have a draft before the upload would have finished. On a patchy hotel connection that difference is huge.

My take: if you transcribe more than a few hours a month and you're the only reader, buy the local app and cancel the subscription. If you collaborate heavily, keep the cloud tool and use the local one for the sensitive stuff.

Accuracy: how good is it really?

For clear English speech, MacWhisper is very accurate, and it handles over 100 languages. The Mac App Store listing says transcription is supported in more than 100 languages, and the underlying Whisper engine was trained on a huge multilingual dataset. A review from The FLTMAG, written by a language-focused academic, tested it on French audio and reached a broadly positive verdict on usefulness, with caveats.

My own test was a 52-minute interview recorded on a phone, with two people, one of whom had a strong Glaswegian accent. The large model got the gist of nearly everything. It stumbled on three surnames, one product name and a stretch where a tram went past. I needed two passes to clean it up: one for names, one for punctuation and paragraph breaks. Total editing time was about 20 minutes.

Those caveats are the same ones you'll run into:

  • Accents and noise: a muffled phone call recorded across a café will produce errors. Garbage in, slightly-less-garbage out.
  • Names and jargon: product names, medical terms and surnames get misspelled. Whisper guesses phonetically.
  • Overlapping speakers: two people talking at once confuses both transcription and speaker labels.
  • Punctuation and paragraphs: usually fine, sometimes oddly placed.

If your recording is full of specialist terms, look for a vocabulary or prompt option in your version and feed it a short list of them. I couldn't verify that every build offers it, so don't count on it. Either way, scan for names before you publish anything. I never trust a transcript I haven't skimmed, and you shouldn't either.

Treat every transcript as a very good first draft, never a finished one.

Hallucinations and silences

Whisper-based tools have one quirk worth knowing. In long stretches of silence or music, they sometimes invent text, usually a repeated phrase or a stock line like "thanks for watching." I saw it once, at the tail of a recording where I'd forgotten to stop the timer. Trim dead air at the start and end of a file and the problem mostly disappears.

How fast is MacWhisper, and which model should you use?

Speed is mostly about model size, and the trade is simple: bigger models are slower and make fewer mistakes. On an M1 MacBook Air with 16GB of memory, I ran the same 60-minute interview through several model sizes. The small model finished in about 7 minutes. The large model took closer to 40. Your numbers will differ on an M3 or M4, which are faster, and on an older Intel Mac, which will be slower and may struggle with the largest models entirely.

Model size Good for Time for a 60-minute file (M1 Air, my test)
Small Quick drafts, memos, clear speech About 7 minutes
Medium Most interviews and meetings Roughly 15 to 20 minutes
Large Accents, noise, technical language About 40 minutes

The medium row is a rough range from a couple of runs rather than a precise benchmark, so treat it as a guide. What matters is the shape: each step up costs you a lot of minutes for a smaller improvement in accuracy.

My habit now is to run the small model first to see if the audio is even decent. If the transcript looks clean, I'm done. If it's full of garbled names, I queue the large model and go make lunch. Batch mode makes that painless because you can line up five files and walk away.

What MacWhisper for Mac asks of your hardware

Any reasonably recent Mac will run it, but an Apple silicon chip makes the experience far nicer. On my M1 Air the small model felt almost instant, and the fan never spun up on short files. Long runs with the large model are a different story: the machine got warm, and I wouldn't try it on battery in a lecture hall.

Memory matters more than people expect. The bigger the model, the more RAM it wants, and a Mac with 8GB will feel cramped if you've also got a browser with thirty tabs open. Close what you don't need before a long run. It's a small habit that saves a lot of beachballing.

Storage is the other one. Models are downloaded once and kept on disk, and the largest ones run to a few gigabytes each. If you're on a 256GB laptop that's already half full, download only the sizes you'll really use. You can always fetch another later.

Finally, keep the Mac awake. A 40-minute run that gets interrupted by sleep is the most avoidable annoyance in the whole workflow, and a quick trip to your energy settings fixes it.

Privacy: does the audio leave your Mac?

With local models, no. Your recordings stay on your Mac. The Mac App Store listing says local models can process audio offline after download, and that's a big reason lawyers, journalists, therapists and researchers like it.

There's an asterisk. The optional Assistant, or any AI provider you connect with your own API key, sends text to a service. Whether that matters depends on what you're transcribing. If it's sensitive, stick to local models and a locally run AI model. We wrote a longer breakdown in is MacWhisper safe to use if you want the full picture.

A simple test: turn off Wi-Fi, drop in a file, and transcribe it with a downloaded local model. If it works, you know the transcription step isn't phoning home. I did exactly that on a train and it ran without complaint.

How to get started in five steps

Setting it up takes about ten minutes, most of it model download time.

  1. Install it: grab the website version or the App Store version, whichever suits how you like to pay.
  2. Pick a model: start with a mid-sized one. Smaller is faster, larger is more accurate.
  3. Add a file: drag in audio or video, or paste a YouTube link.
  4. Review the transcript: click a line to jump to that moment in the audio, fix names and obvious slips.
  5. Export: choose DOCX or PDF for documents, SRT or VTT for subtitles, or copy the text straight into your notes.

Name your files sensibly before you drag them in ("2026-03-12-client-interview.m4a"). When you're batch-processing twenty of them, you'll thank yourself.

A real workflow: the 90-minute interview

Back to that recording on your desktop. Here's how I'd handle it, start to finish, with MacWhisper for Mac doing the heavy lifting.

First, trim. Open the file in QuickTime or Audacity and cut the chatter before and after the actual conversation. That avoids the silence problem and saves processing time. Second, run the small model as a test. In ten minutes or so you'll know whether the audio is clean or a mess. Third, if it's a mess or full of names, run the large model and turn on speaker labels.

While that runs, write down the five or six names and terms you know will be wrong. When the transcript lands, search for each one and correct them in bulk. Then read through once with the audio playing, pausing only where something looks off. For a 90-minute interview, expect roughly 30 minutes of cleanup. Export to DOCX, send it, and you've turned a tomorrow deadline into an afternoon.

The whole thing, including the model run, took me under two hours on a recording I'd have spent a full day typing by hand.

Where MacWhisper for Mac falls short

MacWhisper is excellent at turning recordings into text and merely decent as an everyday dictation tool. I noticed it fastest when I tried to dictate a short Slack reply: by the time I'd started the recording, waited for processing and copied the result, I could have typed it.

It's a transcription workflow, not a typing replacement

With file transcription, you record first, then process, then edit. That's perfect for interviews and meetings. It's clunky if what you want is to speak a paragraph into an email and have it appear, correctly punctuated, right where your cursor sits.

Everyday annoyances

Some things to weigh up before buying:

  • Mac only: no Windows, and no Android app. If you work across devices, you'll want something else on the other side.
  • Big models need power: older Intel Macs struggle with the larger models.
  • Two versions: the website and App Store editions are separate, which confuses support and upgrade paths.
  • Editing still takes time: a one-hour interview might need 15 to 20 minutes of cleanup, even when it's accurate.
  • Disk space: the larger models are multi-gigabyte downloads, so check your free storage before you grab them all.

MacWhisper vs dictation-first apps

If you mostly want to talk and have text appear in whatever app you're using, pick a dictation-first tool; if you mostly have recordings, pick MacWhisper. The two jobs sound similar and aren't.

Job Better fit
Transcribe a 2-hour podcast MacWhisper
Subtitles for a video MacWhisper
Record and summarise a Zoom call MacWhisper
Write emails and Slack messages by voice Dictation app
Dictate code comments or documents Dictation app
Fix punctuation and tone as you speak Dictation app

There are plenty of tools in the second category, including Wispr Flow and TypeWhisper, a free local option.

GhostWriter sits in that category too. It's a macOS app built around the dictation job: you speak, and it types clean, ready-to-paste text into any application, handling punctuation and formatting automatically and adjusting tone for contexts like emails or code editors. It doesn't try to replace MacWhisper's file-and-subtitle workflow, and I wouldn't pretend it does. Plenty of people end up using both: a transcriber for recordings, a dictation tool for the daily stream of writing.

Who should buy MacWhisper for Mac

Buy it if recordings are a regular part of your work. Specifically:

  • Journalists and researchers: local processing and speaker labels are exactly what interview work needs.
  • Podcasters and video creators: transcripts, show notes and SRT files from one app.
  • Students: lecture recordings become searchable notes. The free tier may be all you need.
  • Professionals who take lots of calls: bot-free meeting capture with summaries.

And who should skip it? Anyone whose real problem is that typing is slow. A transcriber won't fix that. You need a dictation tool, because the pain is in the writing, not the recordings.

Tips to get cleaner transcripts

A few habits make a bigger difference than any setting.

Fix the audio first

A decent microphone beats a bigger model. Even a cheap lavalier mic clipped to a collar will outperform a laptop mic across a table. Record in a quiet room, and ask guests to wear headphones on remote calls so their speakers don't leak into your track.

Pick the model for the job

Use a smaller model for quick drafts and memos, and a larger one for interviews with accents, noisy rooms or technical language. Running the biggest model on everything is slower than it needs to be.

Set the language

Auto-detection is good, but if you know the audio is Spanish, say so. It avoids the occasional odd switch mid-file.

Edit in one pass

Play the audio at 1.5x while reading the transcript. Fix names first, then punctuation. Don't polish sentence by sentence on the first pass; you'll burn an hour.

Spend your money on a better microphone before a better model. Clean audio fixes more errors than any setting.

My recommendation

Download the free version and run one real file through it today, ideally something messy like a phone-recorded interview. If it handles that, you've got your answer. Buy the €64 one-time Pro license if you hit the batch, subtitle or meeting limits, and skip the subscription unless you only need Pro for a single project.

One limitation to test before you pay: record a short call with headphones on and check that both your voice and the other person's land in the transcript. That's the setup where I hit the audio-source trap, and it's much better to find out on a five-minute test than halfway through an important meeting.


GhostWriter is a voice-powered AI app for Mac that turns what you say into clean, ready-to-paste text in any application, with punctuation, formatting and tone handled for you. If you've got recordings to transcribe, MacWhisper is a solid pick; if you'd rather speak your writing straight into your apps, try GhostWriter.

Frequently asked questions

Yes, MacWhisper for Mac is free to download and the website advertises free use of local models. Pro unlocks more advanced features. The website sells Pro as a one-time €64 purchase, while the App Store version offers subscriptions and lifetime in-app purchases.

They're two versions from the same developer. MacWhisper is sold on the developer's website with a one-time Pro license, while Whisper Transcription is the Mac App Store edition with subscription and lifetime options. Features overlap but aren't identical.

Yes. Once you've downloaded a local model, MacWhisper for Mac can transcribe audio offline, so recordings stay on your Mac. Optional AI features that use a cloud provider or the Assistant do send text online.

Yes. According to its support documentation, it can record individual apps or system audio and detect meetings in Zoom, Teams, Webex, Skype, Slack and Discord, without a bot joining the call.

It includes system-wide dictation, but its main strength is transcribing files and meetings. If you want to speak emails and documents all day, a dictation-first app is usually a better fit than MacWhisper for Mac.

The Mac App Store listing says MacWhisper supports transcription in more than 100 languages, using OpenAI's Whisper models. Accuracy varies by language, accent and audio quality.

Share