Will Your AI Notetaker Transcribe the People Talking Near You?
You take a client call from a cafe or a coworking space, and the person at the next table is on a call of their own. A notetaker that listens through your Mac can hear both of you, and in many tools everything on your microphone carries your name. Here is what Granola documents, what Apple's mic modes cover, how Driffle labels a second voice on your mic, and a two minute test to run before your next call from a shared space.
Your microphone does not know who is talking
A notetaker that runs on your Mac usually takes two feeds: your microphone and your computer's audio. The people on the call reach it through the computer's audio. You reach it through the microphone, and the microphone takes whatever arrives. In a cafe that includes the barista, the grinder and the founder at the next table pitching their seed round.
Noise is the easy part. Hum, clatter and traffic are what headsets and noise suppression are designed to remove, and transcription models cope with them reasonably well. A clear voice two seats away is a different problem. It is speech, and speech is exactly what a transcriber exists to keep.
The stranger's words can land under your name
Granola's documentation is unusually plain about this. Where its speaker tags are not available, the transcript shows Me and Them, which it says correspond to your microphone input and your system audio. Its speaker tags page explains that names come from the participants your meeting app shows, and that when several people speak through one shared device, Granola cannot tell them apart. A stranger near your laptop is not in the call at all, so there is no name for them to receive.
On a microphone based notetaker, then, the default answer to who said a line from your side is you. That matters more than a messy transcript, because the summary is written from the transcript. If the person next to you says they will send the revised contract by Friday, a summary model has little reason to doubt it was your promise to your client. It can turn up in your action items, or in the recap you forward.
The same thing happens at home when someone else is talking in the room. Cafes, coworking spaces and open offices just make it routine.
Voice Isolation helps your call. Your notes may not get it.
Macs have a Mic Mode control. Apple's support page describes Voice Isolation as giving priority to your voice and reducing background noise, and you choose it from a menu that appears while a video call is in progress. It shapes the audio your call app sends. Whether it also cleans the audio your notetaker hears depends on how the notetaker opens the microphone, and most vendors do not say.
Driffle's answer, so you do not have to guess: it opens the microphone directly, outside the call app's voice processing, so do not count on a mic mode you chose for the call to clean what Driffle hears. And if you are wearing AirPods, Driffle listens through the Mac's built in microphone instead of the AirPods mic, which keeps the AirPods in their higher quality playback mode. The trade is real. A laptop's built in mic is the one most likely to hear the next table. A wired or USB mic that you have selected as your input is used as it is.
How Driffle labels a second voice on your mic
Driffle transcribes your microphone as its own stream and separates the voices on it. While it hears one voice on your mic, it labels that voice You. As soon as a second distinct voice turns up on the mic, it stops calling any mic voice You and labels them Speaker 1 (in room), Speaker 2 (in room) and so on, including lines it had already shown as You. The rule written into the code is that under labelling is acceptable and labelling someone else as you is not.
For a voice on your mic that it cannot name, Driffle shows a one tap Who's this? prompt. It offers the people on the meeting invite who do not appear as remote in the call window, plus Someone else. A stranger at the next table is Someone else.
None of this filters anything out. Driffle still transcribes what your microphone picks up, the split depends on the voices being distinct enough to separate, and transcription runs in the cloud, not on your Mac. What changes is the label: a line from the next table reads as someone in the room, which is the cue to check it before it becomes your commitment.
Test it before the call that matters
Make a two minute test call from the place you actually work. Ask someone near you, who is not on the call, to say a short line with a clear commitment in it, something like we will cut the price by ten percent. Then open the transcript and the summary and check three things: whether the line appears, whose name is on it, and whether it became an action item.
If it comes back under your name, the fixes are physical. Use a headset with a mic close to your mouth and confirm your notetaker is using that mic, sit with your back to a wall, or book the phone booth for calls where a stray promise would be expensive. The people around you also never agreed to be transcribed, so a call that deserves notes usually deserves a private room.
If you want a notetaker that tells you when a voice on your mic is not yours, you can try Driffle on an Apple Silicon Mac at driffle.ai/download.
Sources
- How transcription works. Read 14 November 2026. Desktop Granola captures audio from your device with no bot; without speaker tags the transcript shows Me and Them for microphone input and system audio; recommends a headset where appropriate to minimise background noise.
- Speaker tags. Read 14 November 2026. Names come from the participant names and active speaker information the meeting app shows; several people speaking through one shared device cannot be told apart; audio from outside the call may be transcribed without a matching speaker name.
- Use video effects and mic modes during video calls on Mac. Read 14 November 2026. Mic Mode is chosen from the Video or Audio menu while a video call is in progress; Voice Isolation gives priority to your voice and minimises background noise; Wide Spectrum keeps surrounding sound.
FAQ
Will an AI notetaker pick up conversations at the next table?
If it captures audio through your Mac's microphone, yes, whenever those voices are loud enough to reach the mic. Background noise is often filtered or ignored, but a clear nearby voice is speech, and transcription is built to keep speech.
Who gets credited with what a stranger near me says?
It depends on the tool. Granola documents that without speaker tags the transcript shows Me and Them, mapped to your microphone and your system audio, so a nearby voice would sit under Me. Driffle labels a single voice on your mic as You, and once it hears a second distinct voice it labels mic voices Speaker 1 (in room), Speaker 2 (in room) and so on instead.
Does a noise cancelling headset stop it?
A headset with a mic close to your mouth helps most, but only if your notetaker actually uses that mic. Check which input it records from. Driffle, for example, uses the Mac's built in mic when you wear AirPods, and uses a wired or USB mic as is when you select one.
Does Apple's Voice Isolation keep other voices out of my notes?
Apple offers Voice Isolation as a mic mode during video calls, and it shapes what your call app sends. Whether your notetaker benefits depends on how it opens the microphone. Driffle opens the mic directly, outside the call app's voice processing, so it does not rely on that setting.
Is it fine to run an AI notetaker in a cafe?
The people around you have not agreed to be transcribed, and recording rules differ by place. For anything sensitive, the practical answer is the same as for the call itself: take it somewhere private.