How to Record a Meeting on a Mac (Every Method, Honestly Compared)

The short answer

• Recording just yourself (a voice memo, a solo take): QuickTime Player or Voice Memos. Zero setup.
• Recording a call, and you don't mind everyone knowing: the built-in recorder in Zoom, Teams, or Google Meet. Easiest, but it announces itself and the file lands in the vendor's cloud.
• Recording a call quietly, with notes and a transcript: a system-audio notetaker like Humla - no bot in the participant list, and it can run entirely on your Mac.
• Recording a call with a bot (Otter, Fireflies, Read.ai): fine internally, awkward with clients, because a stranger appears in the participant list.
• Recording arbitrary computer audio into a file you control: a virtual audio device (BlackHole, Loopback) routed into QuickTime. Powerful, fiddly, and very easy to leave misconfigured.

Full disclosure before anything else: I make Humla, which is one of the five, so read me accordingly. What I can promise is that I've tried to be straight about the four times out of five when something else is the better answer. Starting immediately: if you just want to record your own voice, close this tab and open QuickTime. Don't install anything. Don't install MY thing.

First, the wall everybody hits

Every "how do I record a meeting on my Mac" question walks confidently forward and then collides with the same wall: you can record your microphone easily, and you cannot record what's coming out of your Mac easily.

Your mic is an input device, so any app can ask for it, politely, with your permission. But the audio being played - the other person's voice in Zoom, a video, a shared clip - is not an input. macOS simply doesn't route it back to apps as one. That's the whole thing. That single design decision is why QuickTime Player's "New Screen Recording" cheerfully records your voice and hands you a file where the other person is a silent, gesturing mime, and why approximately every guide on this topic ends with the words "…now install BlackHole."

There are exactly two ways around it.

1. A virtual audio device. Install a driver (BlackHole is free, Loopback is paid) that pretends to be a speaker, send your Mac's output into it, then record it as an input. It works. You are also now routing your audio through a fake speaker, which means you can't hear anything unless you ALSO build a multi-output device, and it stays that way until you remember to put it back.
2. Screen Recording permission. Apple's ScreenCaptureKit, the framework behind modern screen recording, can capture app and system audio, and it's gated behind the Screen Recording privacy permission rather than the Microphone one.

That second route is the reason a meeting notetaker asks for Screen Recording permission when it plainly isn't recording your screen. It's not being nosy. On macOS, Screen Recording IS the permission that governs system audio, and there's no narrower one to ask for.

Humla asks for exactly that, for exactly that reason, and captures no video at all. The video stream it opens is two pixels wide and thrown straight away, because ScreenCaptureKit insists a video stream exist before it will hand over any audio. Two pixels. Discarded immediately. That is the silliest sentence in this post and it is a completely accurate description of how system audio works on a Mac in 2026.

Anyway. The thing worth carrying with you when you evaluate anything on this list: granting Screen Recording is a much bigger grant than granting a microphone. It's the permission that could, in other hands, watch everything you do. Check what the app does with it, mine included.

The five methods

1. QuickTime Player or Voice Memos, for recording yourself

QuickTime → File → New Audio Recording, hit record. That's it, that's the method. Voice Memos is the same idea with a nicer list view.

Use it when: you're dictating, recording a talk in a room, or capturing your own half of something. For an in-person meeting where everyone's sat around one table this genuinely works, because the mic just hears the room like an ear would.

Don't use it when: you need the other side of a video call. It will not be there, and you will not find out until afterwards.

2. A virtual audio device, for control at a setup cost

Install BlackHole, create a Multi-Output Device in Audio MIDI Setup so you can still hear the call, point your system output at it, then record that device as an input in QuickTime.

Use it when: you want a raw audio file of arbitrary system audio and you're comfortable with audio routing. It's easily the most flexible thing on this list.

Don't use it when: you'll be doing this weekly. It's a configuration you have to set up and then remember to tear down, and the failure mode is genuinely miserable: an hour-long call, wrapped up, feeling good, and then discovering your output was still pointed at a virtual device nobody was listening to. Or the reverse. There is no version of that discovery that happens early. You also get a .m4a and nothing else - no transcript, no speaker labels, no summary, just a file called Audio Recording.m4a that will mean nothing to you in March.

3. The built-in recorder in Zoom, Teams, or Google Meet

They all record natively, and this is the most honest method here by a distance, because everyone in the call gets told loudly and unmistakably.

Use it when: the recording is for the group. A training session, an all-hands, a demo somebody missed. Nobody should be surprised it exists, and everyone's audio is captured cleanly at the source rather than through a laptop speaker.

Don't use it when: the recording is your own working memory. Two reasons, and the first is the one people underrate: the banner changes how people talk, and the "off the record" half of a client call is usually the half worth having. The second is more boring and bites later - the file lives in the vendor's cloud under the host's account, which may well not be you. Be a guest on someone else's call and you may never see a copy at all.

4. A bot notetaker

Otter, Fireflies and Read.ai send a participant into the call to record it. Zero setup, cross-platform, and it works even when your Mac is shut.

Use it when: your meetings are internal, everyone's used to the bot by now, and you want the transcript sitting in a shared cloud workspace where the team can find it.

Don't use it when: you're on client calls. An unfamiliar robot appearing in the participant list reads badly no matter how good the transcript is, plenty of companies flatly forbid unknown apps in confidential meetings, and some of these tools email a recap to every attendee by default, which is a sentence worth reading twice. We wrote up the alternatives in the best no-bot notetakers and the workflow itself in client meeting notes without a bot.

5. A system-audio notetaker

The fifth shape: an app on your Mac that captures your mic and your system audio, which is to say both halves of the conversation, and transcribes them, with nothing joining anything. From the other side the call looks completely normal, because it is completely normal. You're recording your own device, the same way you'd record a voice memo.

Here's what that looks like in Humla specifically, since it's the one I can describe down to the detail:

1. Grant Microphone and Screen Recording permission once. (You'll need to relaunch after granting Screen Recording. That's a macOS behaviour, not a Humla one, and no, I can't fix it.)
2. Join your call normally. Nothing to admit to the meeting, no plugin, no browser extension.
3. Hit Record. It captures your mic and the call audio as two separate streams, kept separate the whole way through rather than mixed.
4. Type your own notes while it transcribes.
5. Stop. It runs speaker identification on your Mac and labels each turn, and Summarize gives you structured notes fusing what you typed with what was actually said.

Step 3 is doing more work than it looks. When the mic and the call get mixed into one file, "who said that" becomes guesswork forever after - you're asking software to un-bake a cake. Kept apart, the mic stream is you and the system stream is them, before any speaker detection even starts. It's the cheapest correctness win in the whole pipeline and it's invisible in every feature list.

Because it's OS-level audio rather than a per-service integration, the app doesn't care what you're meeting in: Zoom, Google Meet, Microsoft Teams, Webex, or something nobody's heard of that your client insists on. No waiting for support to be added.

Two cases behave differently and I'd rather flag them than have you find out live:

• In-person meetings are mic-only, because there's no system audio when there's no call. Put the Mac on the table. Speaker detection still separates the voices in the room.
• Phone calls only reach your Mac if the audio does: a softphone or VoIP app, your iPhone handed off to the Mac, or plain old speakerphone next to the mic. There is no clever path to your carrier's audio and anyone claiming otherwise is doing something you should ask questions about.

Where the recording actually goes

Most guides on this topic stop at "here's how to press record," which is a shame, because this is the section that matters six months later when someone asks where a client's call is stored.

QuickTime and the virtual-audio-device route: the file is yours, on your disk, and nothing else happens to it. Built-in recorders and bot notetakers: the recording goes to the vendor's cloud, because that IS the product, and there's nothing sneaky about it as long as you know.

System-audio notetakers vary wildly, and here's the trap: "no bot" does not mean "private." Several perfectly good bot-free tools still upload every recording to a server to transcribe it. Quiet in the meeting, chatty afterwards. Those are separate properties and the marketing tends to blur them.

Humla's answer, since I'm the one making claims here: transcription can run entirely on-device with local Whisper, speaker detection is offline either way, and recorded audio isn't kept at all unless you turn on "Keep recorded audio." By default the audio sits in a temp directory during processing and is deleted once transcription and the speaker pass have finished. You keep the transcript and the word timings. There's just no playback file, because there's no file. Turn retention on and recordings land in your own Application Support folder at roughly 1 MB per minute per channel, which adds up faster than you'd guess on a week of calls.

The longer argument for why off is the right default is in local-first vs cloud meeting notes and on the privacy page.

The part that isn't technical

None of this is legal advice, and the rules genuinely do differ depending on where you and the other person happen to be sitting. But the practical version is one line long:

Say you're recording. Out loud. At the start.

Not because a statute compels it everywhere - in plenty of places recording a conversation you're part of is entirely lawful without asking anyone. Say it because it costs you one sentence and removes every single version of this going badly later. There is no scenario where you regret having said it. There are several where you regret not having.

Some jurisdictions require all-party consent and treat covert recording as a criminal matter. The EU and EEA add a second and completely separate question, which people constantly merge into the first: a recording of an identifiable person is personal data under GDPR whether or not making it was lawful. Lawful to record, still regulated to keep. We've covered recording client calls in English and the Norwegian rules specifically in Norwegian, where the criminal-law line is considerably narrower than almost everyone assumes.

Here's the uncomfortable bit for method five, which is my own: a bot notetaker announces itself and does some of this work for you. A system-audio app doesn't do any of it. The thing that makes it pleasant to use with clients is precisely the thing that puts the entire disclosure burden on you. I'm not going to pretend that's a feature. Say the sentence.

So which one?

• Recording yourself: QuickTime. Install nothing.
• A group recording everyone expects: the built-in recorder in whatever you're already in.
• A one-off capture of arbitrary computer audio: BlackHole into QuickTime, and set your output back afterwards. Write yourself a note. You will forget.
• Every client call, with notes you'll actually search later: a system-audio notetaker. Humla is free, open source, runs on macOS 13+ (Apple Silicon recommended), and can keep the whole thing on your Mac.

The honest limits of that last one, since it's mine and you'd be right to discount it: Mac-only, no mobile app, no CRM integration. If you need a recording that exists independently of one specific laptop, or you've got teammates on Windows, a cloud tool is straightforwardly the better fit, and I'd rather tell you that here than have you find it out in month three.