Can I get AI notes for an in-person meeting?

Yes — but nothing in the AI-notetaker category will do it, and it’s worth understanding why before you go looking for the tool that does.

Every capture method this market is built on assumes the meeting exists as software somewhere. A notetaker bot needs a link to join. The platform’s own AI needs a platform to live inside. A caption-scraping extension needs captions the platform already rendered. System-audio capture needs something playing through your speakers. An in-person conversation is none of those. There is no meeting object — no link, no roster, no stream, no audio on any machine. There’s a room, and there’s air.

That makes this the one setting where all four methods fail at once, and it collapses the whole question down to a single one: whose microphone is in the room?

What actually works in a room

Approach	In-person, no call	Hybrid: you’re in the room	Hybrid: you’re dialing in
Notetaker bot (Otter, Fireflies, Fathom, Read AI, tl;dv)	No — nothing to join	Only the call leg; hears the table through one shared mic	Yes — it joins the call
Platform-native AI (Zoom, Teams, Meet, Webex)	No — no platform	Yes for the call leg, if the host’s plan and admin allow it	Host- and plan-gated
Caption-scraping extension (Tactiq)	No — no captions	No	Only in a supported browser client with captions on
Dedicated recorder or phone app (Plaud, Limitless pendant)	Yes — this is the category that owns the room	Yes	Unnecessary — it’s already on your computer
System-audio capture (Canary)	No — nothing is playing	Partly — remote voices only; the table isn’t in the audio	Yes — the room arrives as ordinary computer audio

For a meeting with no call attached, the answer is a recorder, and we’d rather say so than pretend otherwise. A small dedicated device or a phone app sitting on the table is purpose-built for it: hands-free, no screen, no setup, and designed for exactly the interviews, workshops, site visits, and coffee-shop conversations that never touch a computer. Plaud is the clearest example of the hardware approach and Limitless of the always-on wearable one; the Plaud alternatives roundup maps that territory in more detail. Canary has no microphone-in-the-room story to sell you — it needs computer audio, and in a room there isn’t any.

The hard part in a room isn’t the software — it’s the acoustics

This is the part most tool comparisons skip, and it matters more than which app you pick.

A video call is, acoustically, the easy case — the case that made this category possible in the first place. Everyone speaks into their own microphone a few inches from their mouth. Each voice arrives as its own stream. There’s no room reverb, and the small lag on a call pushes people into taking turns rather than talking over each other.

A room inverts all four. One microphone, several feet away, picking up several people at once, plus chairs, HVAC, laptops, and the person who turns toward the whiteboard mid-sentence. There are no per-speaker streams, so speaker diarization has to separate voices acoustically instead of reading them off a roster — which is the same disadvantage system-audio capture already carries on calls, made considerably worse. The result is that the same tool you find accurate on your Zoom calls will be noticeably rougher on a conference-room recording, and it isn’t the model’s fault. How accurate are AI meeting notes covers what to expect and which failures you’ll see versus the ones you won’t.

Two things help more than switching apps: put one device near the middle of the table rather than four devices scattered around it, and get into the habit of saying names when it matters — which helps the humans dialled in, too.

Hybrid is the case people actually mean

Most people asking this question aren’t describing a purely offline meeting. They’re describing a hybrid one: three people in a room, two more on a screen at the end of the table. And hybrid meetings split sharply depending on which side of the screen you’re on.

In the room, you’re mostly back to the microphone problem. The remote participants’ voices are playing through the laptop, so system-audio capture gets them cleanly — but the four people around you are in the air, and a tool reading computer audio hears half the meeting. If the room joined through a dedicated room system rather than someone’s laptop, there’s no audio on your machine at all, which is the same honest limit noted in can I get a live summary during a Webex meeting. In the room, the platform’s built-in AI or a recorder on the table is the better answer.

Dialling in, everything changes. The whole meeting — room and all — comes out of your speakers as ordinary computer audio. There is nothing special to configure and no bot to send.

And this is the honest home of the wedge, because the remote seat in a hybrid meeting is the hardest call of the week to follow. One shared room mic flattens several distant voices into a single stream. People talk over each other the way they do in person, without the latency that keeps calls orderly. Side conversations happen two feet from the mic and reach you as an unparseable mumble. Someone reads off a whiteboard you can’t see. You often can’t tell who’s speaking. Drifting under those conditions isn’t a discipline problem — it’s what happens when the audio is poor and the visual handholds are gone, and it’s the same shape of problem that makes phone calls hard to stay with, arriving from a different cause.

Canary is a real-time, bot-free meeting summarizer. It captures your computer’s system audio (no bot in the call, no plugin, no virtual audio device) and shows a live, multi-resolution rolling summary — from what’s being said right now to the whole call — so you can catch up the instant your name is called. On a hybrid call that’s the whole point: when the room says your name and you’ve lost the thread of a conversation you could only half-hear, the rolling summary is the scrollback the room never gave you. It runs on macOS, Windows, and Linux, and it doesn’t care that half the participants are sitting in the same building.

Two caveats, stated plainly. A live summary is not a live transcript — it compresses rather than scrolling at speaking pace, which is what a listener who’s fallen behind actually needs. And whatever’s on the room’s screen or whiteboard isn’t in the audio, so it won’t reach your notes; screen shares and slides are audio capture’s standing blind spot, and a whiteboard is the most extreme version of it.

What about the laptop microphone?

Canary can fall back to the microphone when there’s no system audio, and other desktop tools can too. It works, in the same sense that a speakerphone recording works: a laptop lid on a conference table is a distant, off-axis mic pointed at the ceiling, in a reverberant room, mixing several people at once. Expect the accuracy drop described above, particularly on names and numbers.

It’s a reasonable stopgap for an unplanned meeting you’re already sitting in with a laptop open. If rooms are a regular part of your week — client visits, standups at a desk, interviews, field work — buy the thing built for it. Using a call tool as a room recorder is the same mistake as using a room recorder for your calls.

A room has no participant list

Recording topics on this site end with disclosure, and the in-person version is the strictest one outside phone calls — for a reason specific to the setting.

On a call, the software does some of the announcing for you. A bot appears in the participant list. A recording banner turns on. A notes indicator sits in the corner. Those are all forms of disclosure that happen whether or not you remember to speak up. A room has none of that. There is no participant list to be visible in, no banner, no roster — there’s only you saying it out loud. And a room’s membership is fuzzy in a way a call’s never is: someone leans in for two minutes, a colleague stops by the doorway, the meeting continues into the hallway. All of them are in the audio; none of them were ever in an invite.

So the practical rules are simple:

• Say it at the start, to the room, out loud. Not in the invite — in the room. The how to tell participants you’re using an AI notetaker guide has wording that doesn’t derail the meeting.
• Put the device where people can see it. A recorder on the table is self-disclosing in a way a phone in a pocket is not. Visible capture is easier for everyone to be comfortable with.
• Assume the stricter rule. One-party vs two-party consent varies by region, and an in-person meeting is exactly where “everyone present” is hardest to pin down. If people are joining and leaving, re-state it.
• In a hybrid meeting, say it twice. The room can’t see what’s on the remote attendee’s screen, and the remote attendee can’t see what’s on the table. Whichever side you’re capturing from, the other half doesn’t know unless you tell them.

That no notetaker can capture a room is a fact about how these products are built. What you owe the people in it doesn’t change with the tool — and bot-free capture never reduces that obligation, it only makes the capturing less visible, which is precisely why you have to do the announcing yourself.