Research

What Apple Notes transcription cannot do for you

Dylan de Heer, Co-Founder & CPO

Dylan de Heer

What Apple Notes transcription cannot do for you

Apple Notes records audio, transcribes it on your Mac, and will summarise the transcript for you. It costs nothing, it is already installed, and for a lot of people it genuinely is the right answer. If you sit across a table from one person in a quiet room and want the conversation as text afterwards, stop reading and go use it.

This is about the point where it stops. Apple documents very carefully what the feature does and says almost nothing about what it does not do, so the limits tend to be discovered halfway through a meeting that mattered. There are five of them worth knowing before you rely on it, and one is much bigger than the rest.

The short answer

Notes records the microphone and nothing else. On a video call that means it captures you clearly and captures the other person only as whatever leaks out of your speakers, or not at all if you are wearing headphones. Transcription also requires your Mac's whole system language to be set to one of ten supported languages, and the Apple Intelligence summary asks for a network connection. For an in-person conversation in English on recent hardware, none of that matters. For a remote meeting, the first limit alone decides it.

What Apple actually ships, and it is more than people think

Worth being precise, because the feature has grown quickly and a lot of what is written about it is out of date.

In Notes on a Mac you can start a recording inside a note, watch a live transcript build as people speak, then search that transcript and copy it into the body of the note. If Apple Intelligence is switched on, you can ask for a summary of the recording. Apple's own guide puts the requirement plainly: audio transcription works on "Mac computers with M1 or later".

Separately, macOS now records FaceTime audio and phone calls straight into Notes, filed into their own Call Recordings folder. The app describes that as capturing "a FaceTime audio or phone conversation as text and audio, then view a summary of the call". Those recordings do get both sides, because Apple is recording its own call. That is a real capability and it is the exception to most of what follows.

None of that is a toy. It is a competent recorder with a competent transcriber attached, running on the machine, for free.

Limit one: it hears the microphone, and only the microphone


Two ceramic cups on a table, one upright and full and the other tipped on its side and empty

This is the one that decides whether the feature works for meetings, and it is the one Apple never states.

Every app on macOS has to declare the permissions it uses, and the system keeps that list. On a Mac running macOS 26.6.1, the Notes app declares exactly six: contacts, calendar, reminders, photos, adding photos, and the microphone. Its own explanation of the microphone request reads "This will let you add recorded audio to notes." There is no system audio permission on the list and no screen recording permission, which are the two routes macOS provides for capturing sound the Mac is playing.

So Notes cannot take the audio coming out of your Mac. It takes the air in front of it.

For a conversation in a room, that is the correct design and it works well. For a video call it produces a recording of half a meeting. Your voice arrives clean because you are speaking into the microphone. The other people arrive as whatever their voices sound like after coming out of your laptop speakers and travelling back into the laptop microphone, which is degraded at best. Put on headphones, which most people do on calls, and they do not arrive at all. You get a transcript of yourself.

Nothing in Notes warns you about this. The recording runs, the transcript builds from your side of the conversation, and it looks like it is working.

Weeve asks for the microphone and the Mac's system audio, using the dedicated system audio interface that macOS provides for it rather than screen recording, so nothing on your display is captured. The two streams land in one recording. That difference is not a matter of polish. It is the difference between a record of the meeting and a record of your half of it.

Limit two: the language gate is your whole Mac, not the meeting

Apple's requirement here is easy to skim past. Transcription in Notes works "when the device language is set to English, Spanish, Portuguese, Italian, French, German, Japanese, Korean, Simplified Chinese and Traditional Chinese".

Two things are happening in that sentence. The list is ten languages, of which six are European. And the condition is the device language, meaning the language your entire Mac runs in, rather than the language the meeting happens to be in.

If your Mac is set to Dutch, Polish, Swedish or Greek, transcription in Notes is not available to you, regardless of what language you are speaking. Changing it means changing your operating system's language.

Weeve transcribes in 25 European languages and its speech model detects the spoken language rather than reading a system setting. In Amsterdam, where a single meeting frequently runs in two languages, that is not a specification detail. It is whether the tool works at all.

Limit three: the summary wants a connection

Recording and transcription in Notes run on the Mac. The summary is a different step with different requirements.

Summarising needs Apple Intelligence, which needs Apple silicon, macOS Sequoia 15.1 or later, around 7 GB of free storage, and your device language and Siri language both set to a supported language. Apple lists sixteen languages for Apple Intelligence, a longer list than the transcription one, and describes the feature as available in beta.

The part that matters more is what the app says when it cannot produce a summary. The message reads: "Connect to Wi-Fi or cellular to summarize this transcript." Summarising a transcript in Notes is not something your Mac does by itself while offline.

Apple has built its cloud processing with real care, and the point here is not that it is careless. The point is narrower and it is the one some readers are actually asking about: the summary step involves a network, so "it all stays on my machine" is not a sentence you can say about the whole workflow. If you are recording client sessions and your reason for choosing on-device software is that you have promised the conversation stays on your machine, that promise needs to cover the summary too.

Weeve runs its summary with a local model, Qwen or Gemma 4 depending on what your hardware can carry. The Mac does the work whether or not there is a network.

Limit four: no speaker separation is documented

Apple's guide to recording and transcribing audio in Notes describes viewing the transcript, searching it and copying it. It describes no speaker labelling, and Apple does not document speaker separation for audio you record in a note.

Absence of documentation is not proof, so treat this as the weakest item in this list and test it yourself if it matters to you. What is verifiable is the mechanism underneath: a single microphone stream carries no channel information to separate people on, so anything Notes did here would have to be inferred from voice alone.

Weeve ships speaker identification, and it is on the free Starter plan rather than behind the paid tier. For a two-person conversation you can usually work out who said what from context. For a five-person meeting, an unlabelled wall of text is close to useless a week later.

Limit five: recordings live in notes, and there is no set of meetings


A loose scatter of blank unmarked paper cards spread across a bare wooden table

A recording in Notes sits inside whichever note you made it in. That is a sensible design for notes and an awkward one for meetings.

Apple gives you Find in Transcript, which searches within the transcript you are looking at. So you can find a phrase in a recording you have already opened. Finding which of last quarter's forty conversations contained a phrase is a different job, and it is not the job Notes is built for. The Call Recordings folder collects calls, but audio you record inside a note goes wherever that note goes.

Weeve organises recordings into projects, so a client or a case accumulates as a set you can go back through. This is the difference that shows up slowly. Nobody notices it in month one and everybody notices it in month six.

When Apple Notes is the right choice, and it often is

Being straight about this is the whole point of the page.

Use Notes if your recordings are in-person conversations, your Mac runs in one of the ten supported languages, you rarely need to know precisely who said which sentence, and you look things up by remembering roughly when they happened. Use it if you record occasionally rather than daily. Use it if you record FaceTime audio or phone calls specifically, because Apple records those properly, both sides, and nothing else here matches that for its own calling apps.

Above all, use it if the alternative is not recording at all. A free recorder you already have beats a better one you have not installed.

Where Weeve is worse

The comparison runs both ways, and the honest version has to include this.

Weeve is Mac only. It needs Apple Silicon and macOS 14 or later, and there is no Windows version and no mobile app. Apple Notes syncs to your iPhone and your iPad, and Weeve does not exist there.

Weeve is another application to install, another set of macOS permissions to grant, and another account. Notes is already open.

Weeve is free for 10 recordings a month on the Starter plan and costs money beyond that. Apple's version has no ceiling and no bill.

Weeve has no team seats, no shared workspace and no admin console, so a firm that wants notes pooled centrally will not get that here. Its interface is English only for now, even though it transcribes in 25 languages. And Weeve states its own accuracy limits plainly: transcription degrades on heavy accents, on overlapping speakers, and in noisy rooms. Apple's transcription faces exactly the same physics.

Common questions

Can Apple Notes record a Zoom or Teams meeting?

It can record your side of one. Notes has microphone access and no system audio access, so the other participants only reach it as sound coming out of your speakers, which is poor if you are on speakers and nothing at all if you are on headphones. For remote meetings you need something that captures the Mac's audio output as well as its input.

Does Apple Notes transcription work offline?

Recording and transcription run on the Mac. Summarising does not reliably: the app's own message when it cannot summarise asks you to connect to Wi-Fi or cellular first.

Which languages does Apple Notes transcribe?

Transcription requires the device language to be English, Spanish, Portuguese, Italian, French, German, Japanese, Korean, Simplified Chinese or Traditional Chinese. It is the Mac's system language that has to match, not the language being spoken.

What hardware do I need?

Transcription in Notes needs a Mac with M1 or later. The Apple Intelligence summary needs Apple silicon, macOS Sequoia 15.1 or later, roughly 7 GB free, and a supported device and Siri language. An Intel Mac gets neither.

Does Apple Notes label who is speaking?

Apple does not document speaker separation for audio recorded in a note, and its guide describes the transcript without it. If knowing who said what is important to your work, test it before you depend on it.

Is Apple Notes private enough for client sessions?

The recording and the transcript stay on your Mac, which is a strong starting position. The summary step asks for a connection, and Notes will occasionally offer to send a recording to Apple to improve transcription accuracy. That offer is an opt-in with a clear warning attached, so it is not something that happens quietly, but it exists and somebody in a hurry can tap yes on behalf of a client who was never asked.

Apple built a good recorder into the operating system. It built it for notes, and a meeting is not a note.

If your meetings are remote, multilingual, or the kind you have to be able to find again, Weeve does the same job with both sides of the call and keeps every step on the Mac.