Can I get a live summary during a meeting on Windows?

Yes, but not from Windows, and not from the two things on a Windows PC with “AI” in the name. Windows 11 ships the words: Live Captions can put every sentence of a call on screen as it’s spoken, from any app, processed on the PC. Microsoft’s meeting summaries live somewhere else, inside Teams, and they arrive after the meeting or when you type a prompt. What nothing built in does is the thing in between: tell you, mid-call, what the last five minutes were about, whichever app the call is in. That’s the gap a real-time meeting summarizer fills.

Windows has the words. Teams has the meaning. Neither has both, live.

On a Mac, both halves of a live summary ship in the operating system. On Windows they’re split across products, and that split is the whole answer.

The live half is in Windows, and it’s a transcript. Windows 11 has Live Captions (Settings → Accessibility → Captions, or Win+Ctrl+L). It captions whatever the PC is playing: a Teams call, a Zoom window, a Meet tab in Edge, a video. Processing happens on the device, so it doesn’t need a connection. It shipped in the 22H2 update, initially in English, with more languages added in later releases, so check what your build supports. What it shows is words at speaking pace. That’s a live transcript, not a live summary. If you’ve been in Slack for three minutes, the captions for those three minutes have scrolled past, and reading them while new ones arrive is the same problem again, with your eyes.

The summary half is in Teams, and it’s after. Microsoft’s meeting AI is a feature of Teams, not of Windows. Intelligent recap is written once the meeting ends. Copilot in a live meeting answers a question you type, once the license, the admin policy, and transcription are all switched on for that meeting. The Teams answer walks through each gate. Because it lives in the meeting, it belongs to whoever hosts the meeting. On a customer’s Teams call, or any Zoom or Meet call, your own Copilot license does nothing. Platform-native meeting AI explains why that’s structural rather than a missing setting.

The Windows-branded AI doesn’t listen to calls. The Copilot app on Windows is a chat assistant: you type or talk to it, and it doesn’t hear your meeting. Recall, on Copilot+ PCs, is easy to misread as a meeting memory. At the time of writing it’s opt-in, it takes periodic snapshots of your screen, and it captures no audio. A meeting is the one thing on your PC that happens mostly in sound, so a screenshot memory of a call is a folder of participant tiles with nothing anyone said. It’s the mirror image of every notetaker, which hears the call and sees nothing on screen. Microsoft changes these features between releases, so check your version.

So Windows gives you words now, and Teams gives you meaning later, on Teams. A live summary is meaning now, in any app: the conversation condensed continuously while it’s still happening, at more than one zoom level. That’s the combination nothing on the PC ships.

What each option shows you during the call

Here’s everything you can run on a Windows PC, scored on the one question this page is about: what’s on your screen while the meeting is still going?

Option on your PC	What you see during the call	Live summary?	The catch
Windows 11 Live Captions	Word-for-word captions of any app’s audio, on-device	No, it’s a live transcript	Windows 11 22H2 or later; languages vary by build; moves at speaking pace
Copilot app / Recall (Copilot+ PCs)	A chat window; on Copilot+ PCs, searchable screen snapshots	No, neither hears the call	Recall is opt-in, screen only, no audio (at the time of writing)
Microsoft Teams (captions, transcript, Copilot, recap)	Live captions and a speaker-labelled transcript, if enabled	Only on request, if licensed; the recap comes after	The host’s tenant, license, and admin decide; Teams only
Zoom / Meet / Webex built-in AI	Captions, plus an on-demand catch-up on some	Only on request, one answer at a time	The host’s plan and admin decide; one platform each
Microsoft 365 Word Transcribe	A recording in progress	No, the transcript comes after	Records through the mic; on headphones it hears only you; plan-dependent
Notetaker bots (Otter, Fireflies, Fathom)	A live transcript or highlight buttons, depending on the tool	No, the summary comes after the call	A bot joins the participant list; your PC isn’t doing the listening
ChatGPT record mode	A recording in progress	No, notes arrive once you stop	Desktop app and plan gated; built for the note afterward
Bot-free Windows notepads (Granola, Jamie, Krisp)	Your own typed notes, if you take any	No, the AI note is written after the call	Built for the document, not the moment
Canary	A rolling summary at four horizons: now, 2 min, 5 min, full call	Yes, continuously	Processed in the cloud; hears audio, not the screen; English-first

Two rows deserve a second look.

Teams is the strongest built-in row on this table, and on Windows it’s the one most people already have. If every meeting you attend is hosted inside your own tenant and the licenses are in place, its recap is the sanctioned default, and Canary vs Microsoft Copilot says so plainly. The row fails on the calls that cross a company boundary, which for most Windows-heavy organizations is exactly the customer, vendor, and interview calls that matter most. The not-the-host answer covers what that means in practice.

The bot-free notepads are where Windows has caught up with the Mac. Granola, Jamie, and Krisp all publish Windows builds (platform support changes, so check their download pages), and all three are built for the document you read afterward. Does Granola show a summary during the meeting? explains why that’s a design choice rather than a missing feature.

On Windows, listening is the solved half, and it doesn’t even ask

The hard-sounding part, getting a call’s audio into an app, is something Windows has done natively since Vista. WASAPI loopback lets an app open your output device, whatever Windows is currently mixing everything to, and read back the exact stream being sent there. Live Captions is the everyday proof: it captions any app because the OS can hear every app.

What that means for a meeting tool:

• Nothing to install. No VB-Cable, no Voicemeeter, no virtual audio device, and no Stereo Mix to hunt for under Recording devices. How to record system audio on Windows without a virtual device covers why those two are the wrong answers on a modern laptop.
• Nothing to reroute. You keep hearing the call exactly as before. The app reads a copy of the stream that’s already going to your ears.
• Any output device. USB headset, Bluetooth earbuds, HDMI, laptop speakers. System-audio capture reads the digital stream before it reaches your ears, which is why headphones don’t break it.
• No driver, so no admin. A virtual cable needs driver-install rights; loopback is part of the OS audio stack. On a managed work laptop that’s the difference between working and being blocked, the same wall people hit when their company blocks meeting bots. Not blocked isn’t approved, and that page covers how to ask.

Here’s the part that’s specific to Windows, and it cuts both ways. A Mac asks before any app can hear other apps, and lists the permission in System Settings. Windows doesn’t. There’s no prompt for loopback capture, no privacy page that lists which apps hold it, and, as far as Windows shows you, no indicator while it runs. For setup that’s convenient: press Start and it works. For transparency it means the OS won’t do any of the telling for you, which the last section comes back to.

Canary uses WASAPI loopback on Windows. It opens whichever output device was the default when you pressed Start Listening. If loopback isn’t available for some reason, it falls back to your microphone, which works in a room but on a headphones call hears only your side. And in the current version it stays on the device it started with, so if you switch from speakers to a headset mid-call, stop and start listening so it follows you.

The Windows-specific half: where the summary sits

A live summary only helps if you can read it without leaving the call. The moment you need it is the moment your name was just called, and Alt-Tab costs exactly the seconds you don’t have. A full-screen Teams or Zoom window covers everything, and a summary buried in the taskbar is a summary you’ll never look at in time.

So the shape that works is a glance, not a switch: a small panel that floats above your other windows, sits beside the call, and never takes focus from it. That’s what meeting situational awareness means in practice, knowing where the room is without going to look.

That’s how Canary is built on Windows:

• It lives in the system tray, not the taskbar. Closing its window hides it; the tray icon is where it lives.
• The panel stays on top. It’s compact and opens at the top-right of your screen. Drag it beside the call or onto your second monitor.
• It never takes focus. Your meeting window stays active, and you can keep typing in Slack while the panel updates. It doesn’t take a taskbar slot either.
• The panel is the resolution matrix: now, the last 2 minutes, the last 5, and the full call. Each row is a rolling summary that re-condenses as the conversation moves.
• You can trim it. Hide the rows you don’t use in settings, or collapse the panel to a slim bar that shows just the latest line.
• Global keyboard shortcuts control it, and you can remap them in settings.

Two practical tips for a Windows desk. First, Windows tucks new tray icons into the overflow menu by default; drag Canary’s icon out onto the visible part of the tray once, so it’s always a click away. Second, if you run the call full screen, snap it to two-thirds of the screen instead, or put the panel on the second monitor. Catching up should be a glance, never an Alt-Tab.

What Canary does on Windows

Canary is a real-time, bot-free meeting summarizer. It captures your computer’s system audio (no bot in the call, no plugin) 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 Windows that’s three steps:

1. Install it on Windows 10 or 11. It’s an ordinary desktop install: no audio driver, nothing added to your sound devices.
2. Click Start Listening. There’s no permission to grant for system audio, because Windows doesn’t require one.
3. Optionally connect your calendar so it can offer to start when a meeting begins. The calendar is optional, and whether a notetaker needs your calendar explains what that grant does.

Because it reads what the PC is playing rather than joining a platform, the same panel works on Teams, Zoom, a Meet tab in Edge or Chrome, Webex, a Slack huddle, or a Discord call. It also runs on Linux and macOS, so a mixed-OS team gets the same panel everywhere. It’s free for 5 meetings a month, and Pro is $15/mo.

When someone says “…what do you think?”, glance at now, widen to last 2 min if you need the thread, and answer. What did I miss in the meeting is the whole reason the panel exists. Canary vs Teams Live Transcription covers how that differs from the live text most Windows users already have, and Canary vs Granola covers the best-known bot-free notepad.

What it won’t do

Credit where it’s due, because the built-in options win some of these outright:

• It isn’t on-device. Live Captions transcribes on the PC; Canary doesn’t. Audio is streamed in short chunks for transcription and then discarded; the only things stored are the text transcript and the rolling summaries, encrypted at rest. If “nothing leaves the PC” is a hard requirement, Live Captions meets it and Canary doesn’t. On-device transcription explains what that property does and doesn’t buy you.
• It hears the call, not the screen. A number on a slide that nobody reads out isn’t in the summary. See do AI meeting notes capture screen shares?
• It starts when your PC starts hearing. It can’t summarize the ten minutes before you joined. Joining late covers what can.
• It’s English-first. For a call held in another language, captions or a multilingual tool will serve you better. See the languages answer.
• It’s for calls on your computer. In a room with no call attached, your laptop’s microphone is the wrong input for any of these tools, Canary included. See in-person meetings.

Nothing stops you running both: Live Captions when you want every word, the panel when you want to know where the conversation is.

Windows never asks, so you have to

On a Mac, the permission prompt is at least a moment where capture becomes visible to you. Windows has no such moment for loopback audio. No prompt, no list of apps that can hear other apps, no indicator. Combine that with no bot in the participant list and no banner in anyone’s window, and bot-free capture on Windows is about the least visible way there is to capture a call. The OS’s silence is a fact about Windows. It isn’t consent from the people whose voices come through your headset.

So say it. A one-line heads-up at the top of the call is enough in most meetings, and here’s a short script. One-party vs two-party consent rules vary by region, and some places require everyone’s agreement. Nothing to install and nothing to approve is a smaller footprint. It isn’t a quieter obligation.