How to Transcribe a Meeting on Mac (2026): Free Tools, Whisper, Parakeet and Automatic AI Methods
Every way to transcribe a meeting on Mac: built-in Dictation, QuickTime plus Whisper, cloud APIs and on-device AI recorders. Accuracy in English, Dutch and French, speed on Apple Silicon, privacy, and how to make transcription fully automatic.
If you have ever spent an hour after a meeting typing up what was said, you already know why transcription matters. The question in 2026 is not whether your Mac can transcribe a meeting but which of the half-dozen ways to do it fits your workflow: free and manual, paid and cloud-based, or automatic and on-device.
This guide walks through every practical method, from the Dictation feature already in macOS to AI meeting recorders that transcribe live on the Apple Neural Engine, and compares them on accuracy (including Dutch and French, not just English), speed, privacy and how much work is left for you afterwards.
Quick answer
- Free, occasional, do not mind manual steps: record with QuickTime, transcribe with a local Whisper app. Private and accurate; no summary, no speaker labels.
- Want it fully automatic on Mac without uploading audio: MeetMemo. It records, transcribes on-device with Parakeet, summarises on-device with Apple MLX, and files the result in Apple Notes.
- Happy with cloud processing and want team collaboration features: a bot notetaker like Otter.ai or Fireflies.
Why transcribe meetings in the first place?
Manually taking notes while staying engaged in a conversation is nearly impossible: when you type, you stop listening; when you listen, you stop capturing. Even diligent note-takers miss details, and typed notes are hard to search three weeks later. A transcript solves all three problems: it captures everything verbatim, it lets you participate instead of typing, and it is searchable.
Beyond convenience, transcription helps in specific situations:
- Follow-up clarity. When a decision was made, the transcript removes ambiguity about what was agreed.
- Stakeholder alignment. People who could not attend read the record instead of a reconstruction from memory.
- Compliance. In regulated industries, full meeting records may need to be retained.
- Multilingual teams. A Flemish, French and English mix in one call is normal in Belgium; a good transcript captures every language accurately.
How meeting transcription works on Mac
Under the hood every method does the same two things: capture audio, then run a speech-recognition model over it. The differences that matter are what audio is captured and where the model runs.
- Audio capture. Your Mac's microphone picks up the room. For video calls you also need system audio, the sound coming out of Zoom, Teams or Meet, otherwise remote participants are faint or missing.
- Speech-to-text. A model such as OpenAI Whisper, NVIDIA Parakeet or a vendor's cloud engine turns audio into text. On Apple Silicon Macs this can run on the Neural Engine, a dedicated AI chip that sits idle for most apps, which is why on-device transcription is now both fast and private.
- Speaker identification. Better tools label who said what and add timestamps.
- Summary. Optionally, a language model turns the transcript into key points, decisions and action items.
- Filing. The result lands in a file, a dashboard, or somewhere you actually work, like Apple Notes.
The big shift in the last two years is step 2 moving from the cloud onto the Mac itself. That change is what makes the privacy and offline differences below possible.
Method 1: built-in macOS transcription (Dictation, Voice Memos, Notes)
macOS Dictation. System Settings, Keyboard, Dictation. Enable it and a keyboard shortcut starts turning your speech into text wherever the cursor is. It is live, free and decent for your own voice. It does not record audio from a video call, does not label speakers, and stops being useful the moment two people talk.
Voice Memos and Notes. Recent macOS versions can transcribe recordings and audio attached to notes in supported languages. It is convenient for a quick voice memo, but the transcript is plain text with no speakers, no summary and no meeting context, and capture is microphone-only.
QuickTime Player. Records audio (File, New Audio Recording) or screen, producing a file you can transcribe with another tool. It does not transcribe by itself.
Limitation: all of these hear what your Mac's microphone hears. On a Zoom call that is you, clearly, and everyone else through your speakers, badly. They are a starting point, not a meeting workflow.
Method 2: free Whisper transcription on Mac
OpenAI's Whisper is an open-source speech model that runs locally. Several free Mac apps wrap it in a drag-and-drop interface:
- MacWhisper. Native Mac app, clean interface, free tier with paid extras. Drop in an audio file, get text.
- Whisper Desktop and similar. Same idea, fully offline.
Whisper's quality on clean audio is good, including Dutch and French, and it copes with multiple speakers reasonably well. Because it runs on your Mac, nothing is uploaded. Larger models are more accurate and slower; on an M2 or later a one-hour meeting takes a few minutes with a medium model.
The workflow: record with QuickTime or Voice Memos (add a virtual audio device such as BlackHole if you need the remote side of a call), save the file, drop it into the Whisper app, export the text.
Limitation: Whisper transcribes; it does not record your calls, label speakers reliably, summarise, extract action items or sync anywhere. Each meeting is three tools and a few minutes of manual work. Fine for the occasional interview; it does not survive ten meetings a week.
Method 3: cloud transcription APIs and bot notetakers
Cloud APIs (Google, AWS, Deepgram, AssemblyAI and others) are fast and accurate in English, but you upload the audio, you are billed per minute, and using them directly requires scripting. Most people encounter them indirectly, through a product.
Bot-based meeting recorders such as Otter.ai, Fireflies and Fathom are those products. You connect your calendar, a bot joins your Zoom, Teams or Meet call, the audio streams to the vendor's servers, and you get a transcript, a summary and a searchable archive. The experience is fully managed and the collaboration features (shared highlights, comments, CRM sync) are mature.
Limitations for Mac users who care about privacy: the audio leaves your device. Under GDPR the vendor is a processor, you need a Data Processing Agreement, and if the servers are in the US you need transfer safeguards. A bot is also a visible participant that hosts can block and clients sometimes object to, and nothing works offline. Accuracy in English is excellent; in Dutch, English-first services dropped to roughly 60 to 70% in our tests.
Method 4: automatic on-device transcription with MeetMemo
MeetMemo is a menu-bar app for macOS that makes transcription automatic without the cloud trade-off.
- Capture. When you join a call in Zoom, Teams, Google Meet or Slack, MeetMemo records system audio plus your microphone directly from macOS. No bot, no browser extension, nothing the host has to enable. In-room meetings and phone calls routed through the Mac work the same way.
- Transcribe. NVIDIA Parakeet runs on the Apple Neural Engine and transcribes live as the meeting happens, with speaker labels and timestamps. A one-hour meeting is finished in well under a minute after you press stop.
- Summarise. The transcript is summarised on-device with Apple MLX (Qwen 3 4B by default): key points, decisions, action items with owners, open questions.
- File. Summary and transcript go to Apple Notes in a MeetMemo folder, and to a local Markdown knowledge base that Claude Desktop or Cursor can search through a local MCP server.
Dutch, English and French are supported out of the box, including Flemish dialect and code-switching mid-sentence. In our Dutch test meetings Parakeet landed at 95%+ word accuracy; English around 97% on clean audio; French around 93%.
Because every step runs on the Mac, it works in airplane mode, needs no account, and there is no Data Processing Agreement to sign. Requirements: macOS 14 or later on Apple Silicon; summaries are faster with 16 GB of RAM or more.
Comparison: ways to transcribe a meeting on Mac
| macOS Dictation / Voice Memos | QuickTime + local Whisper | Cloud API / bot notetaker | MeetMemo | |
|---|---|---|---|---|
| Records video-call audio (both sides) | No | Only with a virtual audio device | Yes (bot) | Yes (system audio) |
| Runs on your Mac | Yes | Yes | No | Yes |
| Live transcription | Your voice only | No | Yes | Yes |
| Speaker labels | No | Limited | Yes | Yes |
| Dutch / French quality | Basic | Good | Varies, often weak | Strong |
| Automatic summary | No | No | Yes (cloud) | Yes (on-device) |
| Apple Notes sync | Partial (Notes) | No | No | Yes |
| Works offline | Yes | Yes | No | Yes |
| Bot joins call | No | No | Yes | No |
| GDPR posture | Local | Local | DPA required | Local, by design |
| Cost | Free | Free | Free tiers, then per seat | 3 meetings free, then €9/month |
Speed, accuracy and privacy on Apple Silicon
Three practical observations from testing on M-series Macs:
Speed. The Neural Engine is purpose-built for this. Parakeet transcribes faster than real time by a wide margin, which is why MeetMemo can show a live transcript during the call and finish seconds after it ends. Whisper on the GPU is slower but still comfortably faster than listening back. Cloud services are bound by your upload speed and their queue, not your hardware.
Accuracy. On clear English audio every serious tool is in the mid-90s. The separation shows up with accents, cross-talk, jargon and non-English speech. For Dutch and French, on-device Parakeet and local Whisper both beat English-first cloud tools clearly. Microphone quality matters as much as the model: a €50 USB mic in a meeting room is worth more than a bigger model.
Privacy. Where the model runs is where your audio goes. Voices are personal data under GDPR, and for professionals under confidentiality obligations (legal, healthcare, finance, HR) that is not a preference, it is a requirement. On-device processing keeps you as the only party handling the recording. Our local vs cloud transcription piece goes deeper, and the European recording laws guide covers consent rules country by country.
What to look for in a Mac transcription tool
If you are evaluating something not covered here, check these:
- Captures system audio so remote participants are transcribed cleanly.
- States where speech recognition runs. "Encrypted" says nothing about who processes it.
- Handles your languages, including switching between them in one meeting.
- Speaker labels and timestamps, or the transcript is a wall of text.
- Produces a summary you would actually send, not just a shorter transcript.
- Puts the result where you work: Apple Notes, Obsidian, a folder, your CRM.
- Works offline, because hotel Wi-Fi should not cost you a meeting.
Which method should you choose?
Choose built-in tools or Whisper if you transcribe occasionally and are happy to review long transcripts yourself. The cost is zero; you give up automation, speaker labels and summaries.
Choose a bot-based cloud recorder if you are comfortable with audio being processed on third-party servers and want a fully managed experience with team collaboration and integrations.
Choose on-device automatic transcription (MeetMemo) if you want transcription to simply happen, you prefer not to invite bots or install extensions, you want notes in Apple Notes, you meet in Dutch, French or English, or you cannot sign a DPA for meeting audio.
For the step-by-step recording setup see how to record a meeting on Mac; for a broader comparison of apps, best meeting recorder for Mac.
Frequently asked questions
How do I transcribe a meeting on Mac for free? Record with QuickTime or Voice Memos, then drop the file into a local Whisper app such as MacWhisper. Private and accurate; no summary or speaker labels.
Can my Mac transcribe a meeting automatically? Yes. MeetMemo records, transcribes on the Neural Engine and summarises on-device, then saves to Apple Notes. Bot notetakers also automate it, but in the cloud.
Does macOS have built-in transcription? Dictation transcribes your own voice live, and Voice Memos and Notes can transcribe recordings in supported languages. Neither captures the remote side of a call or labels speakers.
How accurate is it? Mid-90s percent for English on clear audio across good tools. For Dutch and French, on-device Parakeet and local Whisper lead; English-first cloud tools trail badly.
Is it private? Only if the model runs on your Mac. Whisper apps and MeetMemo do; cloud services and bots upload audio.
How long does a one-hour meeting take? Parakeet in MeetMemo: well under a minute after the call, with live transcription during it. Local Whisper: a few minutes. Cloud: depends on upload and queue.
Can I edit the transcript afterwards? Yes. MeetMemo transcripts and summaries are editable in Apple Notes and in the local Markdown files; Whisper output is plain text you can edit anywhere.
Which tool is best for Dutch and French? MeetMemo, with Parakeet running locally. Whisper is a good free alternative.
Try automatic transcription on your next meeting
Install MeetMemo, open it before your next call and press record. The transcript appears live, the summary lands in Apple Notes seconds after you stop, and nothing leaves your Mac. Your first three meetings are free with every feature, no credit card. Download for Mac.
