Comparison

Narada vs Apple Notes and Voice Memos — why the free tools stop at your own voice

Apple gives every Mac a recorder, a transcriber and a summariser, for free. There is one thing none of them does, and it happens to be the thing a meeting needs.

Short answer
  • No Apple recording surface on macOS offers a system-audio input. QuickTime’s audio recorder lists exactly one input control — “Microphone: Choose a microphone (if you have more than one available)" — and macOS screen recording offers a microphone picker with a None entry and nothing else.
  • Apple documents Notes audio transcription as available “on Mac computers with M1 or later when the device language is set to English, Spanish, Portuguese, Italian, French, German, Japanese, Korean, Simplified Chinese, and Traditional Chinese” — ten languages, and neither Russian nor Ukrainian is one of them.
  • Voice Memos needs “macOS 15 or later and a Mac with Apple silicon” to transcribe, per Apple’s own guide, and its documented inputs are “the built-in microphone, a supported headset, or an external mic”.
  • Apple Intelligence summarises a recording. Apple’s description is “generate a summary of the transcript” — no action items, no decisions, no owners, no speaker labels.
  • Narada records your microphone and your Mac’s output as two separate tracks and transcribes both on the Mac with Parakeet TDT v3 in 25 European languages. The recorder, the transcription and the files are free; only the cloud AI tier costs money.

If you want a recording of your own voice and a summary of it, Apple already shipped that and you can stop reading. Notes records audio inside a note and transcribes it. Voice Memos does the same as standalone files. Apple Intelligence summarises either one. It costs nothing and it is already installed.

What none of them will do is record the other people on a video call. That is not a preference buried in a settings pane. No Apple recording surface on macOS has a system-audio input, and Apple’s documentation for each one lists a microphone and stops there.

Who each one is for

Apple's built-in tools are for capturing a room you are in, or a thought you are having. A lecture, an interview across a table, a voice note in the car, a meeting where one microphone in the middle is genuinely enough. Free, no account, no download, and the transcript is searchable inside the app. If that describes your recordings, everything below is a solution to a problem you do not have.

Narada is for people whose meetings happen over Zoom, Meet or Teams, where the half of the conversation that matters is coming out of the speakers rather than going into the microphone. It records both, keeps them as separate tracks, transcribes them on the Mac, and files each meeting in a folder you chose. It is macOS 15.6 or later and there is no iPhone app.

The difference that actually matters

macOS treats microphone input and speaker output as two different things with two different privacy rules. Apple exposes the first everywhere and the second nowhere in its own apps.

You can verify this in a couple of minutes. QuickTime Player’s New Audio Recording has one input control: “Microphone: Choose a microphone (if you have more than one available).” The macOS screen-recording documentation is the same shape — “Below Microphone, select the microphone you want to use so that a checkmark appears next to it. To turn off the microphone, select None.” Voice Memos says it in different words: “To create a recording, use the built-in microphone, a supported headset, or an external mic.”

Notes is the one case where I will not claim more than the documentation supports. Apple’s page on recording and transcribing audio in Notes never mentions an audio input source, system audio, or recording sound the Mac is playing. That is an absence of documentation rather than a documented impossibility — but nothing Apple publishes offers it.

The capability does exist in the operating system. Apple’s privacy page describes it plainly: “You can decide which apps and websites are allowed to record your screen and audio”, and an app may be allowed “both your screen and audio, or just your audio.” That last clause is the whole of Narada’s recording design. Apple reserves it for third parties.

One exception is worth knowing, because it is useful and often missed. The Phone app on Mac records both sides of a call, but only calls in the telephony stack: “The Phone app arrives on Mac thanks to Continuity, which lets users relay cellular calls from their nearby iPhone”, and Apple Intelligence then summarises “phone and FaceTime audio calls you make on your Mac.” There is no documented path from there to a Zoom or Meet call, and Apple notes that “call recording isn’t available in all countries and regions.”

Side by side

Apple Notes / Voice Memos / QuickTimeNarada
Records your microphoneYesYes
Records the call’s far endNot offered. QuickTime, Voice Memos and screen recording document a microphone input only; Notes documents no input source at allYes — macOS system-audio permission, on its own track
TranscriptionNotes: M1 or later, ten device languages. Voice Memos: “macOS 15 or later and a Mac with Apple silicon”Parakeet TDT v3, on the Mac, 25 European languages (~600 MB one-time download)
Russian or UkrainianIn none of Apple’s three published language listsYes — 7.2% word error rate for each on the model vendor’s FLEURS benchmark
Speaker separationNone documentedUnnamed Speaker 1 / Speaker 2 labels on the call’s audio; your own lines carry your name because your mic is a separate file
SummaryApple Intelligence: “generate a summary of the transcript”Summary, topics with timings, decisions, open questions and action items
Action items and decisionsNoYes, in a ledger split into I owe / Owed to me / Unclear
Starts from your calendarNoOptional. Default is ask-before-recording; auto-record is opt-in, fires up to 60 s before the scheduled start after a 3-second cancellable countdown
HardwareApple Intelligence needs “Mac models with M1 and later”; Notes transcription M1 or later; Voice Memos Apple siliconmacOS 15.6+, universal binary — Intel Macs record and transcribe. On-device AI is Apple-Silicon-only
Where recordings liveIn the note or the Voice Memos library, iCloud-synced; Phone recordings in “a Call Recordings folder in the Notes app”One folder per meeting where you chose: .m4a, .txt, .md, .json. Unencrypted on purpose
iPhone and iPadYesNo — macOS only
Search across everythingText search inside each appArchive-wide search, chat with citations that open the meeting at the second quoted, read-only Claude Desktop connector (keyword, not semantic)
Screen Recording permissionNot applicableNot required — a CoreAudio process tap, not screen capture
PriceIncluded with macOSFree tier is the complete recorder; Pro AI $6.99/month or $69.99/year

Checked 4 August 2026. Apple’s requirements and language lists change with each macOS release; the sources are listed at the end so you can re-read them yourself.

Where Apple’s built-in tools are the better choice

Pick Notes if your recordings are things you are physically present for and you want them attached to the rest of your thinking. Notes keeps the recording in the note body “with related comments, checklists, and documents”, which is a better home for a lecture than a folder of loose files. Pick Voice Memos if you want a clean library of audio with searchable transcripts and a playhead that jumps to the word you clicked. Both are free and neither asks you to install anything.

Two more reasons that have nothing to do with features. The first is iPhone: Apple’s version follows you onto a phone and an iPad, syncs through iCloud, and works on the train. Narada is a Mac app, and if half your recording happens away from your desk, that is decisive.

The second is the Phone app’s consent model. When you record a relayed cellular or FaceTime call, Apple’s documentation states that “both call participants hear an audio notice that the call is being recorded”, and the guide tells you up front to “make sure the other call participant is willing to be recorded.” No app that records your Mac’s audio output can do that for you. If you would rather the tool announce itself, Apple’s is the one that does.

Where Narada is the better choice

If your meetings are remote, microphone-only is not a small gap. You end up with a recording in which you are crisp and everyone else is a muffled reflection off your own speakers, which transcribes into text nobody can use. Narada takes the output stream directly and transcribes it separately from your microphone, so the far end arrives as text rather than as room noise. No bot joins the call and nothing appears in the participant list, because nothing was invited.

If you work in a language Apple has not covered, this is decisive rather than marginal. Notes transcribes ten languages; Apple Intelligence covers sixteen; the narrower list for Notes transcription summaries omits Danish, Dutch, Italian, Norwegian, Swedish, Turkish and Vietnamese, so you can have Apple Intelligence and still get no summary. Russian and Ukrainian are in none of the three. Narada covers 25 European languages, unevenly — the model vendor’s benchmark runs from 4.0% word error rate in Italian to 36.9% in Greek, with Slovenian, Latvian, Maltese and Lithuanian all above 25%. Check your language before you commit.

And if what you want from a meeting is not prose, Apple’s layer stops short. Apple Intelligence summarises. Narada also writes out decisions, open questions and action items, sorts commitments into what you owe and what you are owed, and nudges the ones nobody has touched in fourteen days. Those extractions are wrong sometimes — the app flags them as drafts for that reason — but a wrong draft you can correct beats a paragraph that sends you back to the transcript.

One more, which is the opposite of a feature claim: Narada compares each finished recording against itself — audio submitted against audio actually in the file — and marks the meeting ⚠ Incomplete when they disagree. It cannot recover what was lost. It can stop you finding out a month later.

Privacy and where your audio goes

Apple’s position here is strong and it deserves quoting rather than paraphrasing.

“We do not use our users' private personal data or their interactions when training our foundation models.”
“The cornerstone of Apple Intelligence is on-device processing, so it is aware of your personal information without collecting your personal information.”

For requests that leave the device, Apple’s Private Cloud Compute documentation says: “Your data is never stored. It is only used to fulfill your requests,” and “Apple only collects limited information about the request, such as the approximate size of the request and response, which features are used for the request, and how long the request takes to complete.”

The honest caveat is not that Apple sends your data somewhere. It is that Apple Intelligence is a hybrid and Apple does not publish which side handles a given feature — its Notes page footnotes the Apple Intelligence reference and points at a general requirements article rather than stating where a transcription summary runs. I also found no Apple sentence saying Notes or Voice Memos transcription is performed strictly on device. It is gated on Apple silicon, which is suggestive, and that is as far as the documentation goes.

Narada’s version, stated the way the app states it: audio never leaves the Mac in any mode. Transcription and speaker separation run locally through Core ML. In Recorder-only and On-device AI modes nothing is sent for AI at all. In Pro AI and bring-your-own-key modes the transcript text — not the audio — goes to OpenAI, through Narada’s proxy or directly with your key, and only after you grant consent in a sheet that says so. Pro AI also sends a per-install device identifier so the proxy can attribute cost. Files on disk are unencrypted by design, so you can open them in anything.

Apple’s framing of third-party recorders applies to Narada as much as to anyone: “If you allow third-party apps or websites to record your screen and audio, any information they collect is governed by their terms and privacy policies.” Read them. That is the correct instinct.

Pricing, plainly

OptionCost
Apple Notes, Voice Memos, QuickTime, PhoneIncluded with macOS
BlackHole virtual audio deviceFree, GPL-3.0; the download page suggests a $10 donation and offers an “I can’t afford to donate” path
Loopback (Rogue Amoeba)$99 one-time. In trial mode “the quality of audio passing through Loopback’s virtual audio devices will be degraded after twenty minutes”
Audio Hijack (Rogue Amoeba)$69 one-time
Narada free tier$0 — calendar-aware recording, mic and system capture, on-device transcription with speaker separation, plain files, labels, the Claude connector
Narada Pro AI$6.99/month or $69.99/year, 3-day free trial on both

Checked 4 August 2026.

Two notes on that table. The free tier is the whole recorder, not a teaser — the subscription buys cloud AI for summaries, follow-ups and chat, and On-device AI gives you those for nothing on an Apple Silicon Mac with enough RAM, at a one-time 3 GB or 7 GB download and lighter quality than the cloud. And the DIY route is only free with BlackHole, which means installing a driver, restarting, building a Multi-Output Device, and remembering that “if you are using a multi-output device, due to issues with macOS the Built-in Output must be enabled and listed as the top device.” The comfortable version of DIY is $99.

How to try both

Apple’s side takes no setup. Open a note, click the audio button, record something, double-click the recording to read the transcript. Do that during a real call — it is the fastest way to hear how much of the far end your microphone actually picks up, and it either settles the question or it doesn’t.

Narada needs two permissions, Microphone and System Audio Recording; neither is Screen Recording. Connect a calendar if you want meetings to start themselves, or skip it and record from the menu bar. There is nothing to migrate in either direction. Run both for a week and keep whichever produces text you reread.

Whatever you use, tell the other side you are recording. The law varies by country and in some places turns on whether everyone consented rather than just you. Narada keeps a reminder on screen for the whole recording, and that is the limit of what software can do — the responsibility is yours.

Sources

  • Apple Support, Record and transcribe audio in Notes on Mac — M1 requirement, ten-language list, no documented input source.
  • Apple Support, View a transcription of a recording in Voice Memos on Mac and Create a recording in Voice Memos on Mac — the macOS 15 / Apple silicon gate, the documented input devices.
  • Apple Support, Record audio in QuickTime Player on Mac — the single microphone input control. https://support.apple.com/guide/quicktime-player/record-audio-qtp3a7cbdc0d/mac
  • Apple Support, Take a screenshot or screen recording on Mac — microphone-only audio option. https://support.apple.com/guide/mac-help/take-a-screenshot-or-screen-recording-mh26782/mac
  • Apple Support, Control access to screen and system audio recording on Mac — the audio-only permission and the third-party terms clause — https://support.apple.com/guide/mac-help/control-access-to-screen-and-system-audio-recording-mchld6aa7d23/mac
  • Apple Support, Use Apple Intelligence in Notes on Mac and Use Writing Tools with Apple Intelligence on Mac — what the summary layer produces.
  • Apple Support, Use Apple Intelligence in Phone on Mac and While on a call in the Phone app on Mac — scope, availability, the audible notice, the Call Recordings folder.
  • Apple Support, Apple Intelligence and privacy on Mac; Apple Legal, Apple Intelligence & Privacy; Apple, Privacy Features — the quoted privacy statements.
  • Apple, Apple Intelligence product page and macOS Feature Availability — the hardware line and the locale lists.
  • Apple Newsroom, macOS Tahoe 26 — the Continuity framing of the Phone app on Mac.
  • Existential Audio, BlackHole; Rogue Amoeba, Loopback purchase page and manual, Audio Hijack purchase page.
  • FluidAudio documentation and benchmarks — the 25-language Parakeet TDT v3 list and its per-language FLEURS word error rates. https://raw.githubusercontent.com/FluidInference/FluidAudio/main/Documentation/Benchmarks.md

All checked on 4 August 2026.

Questions people ask about this

Can Apple Notes record a Zoom call?

It can record through your microphone while a Zoom call is happening, which captures you clearly and the other side only as whatever leaks out of your speakers. Apple’s Notes documentation never mentions an audio input source or system audio, and no other Apple recorder offers one either.

Does Voice Memos transcription work on an Intel Mac?

No. Apple’s requirement is “macOS 15 or later and a Mac with Apple silicon”. Apple also notes that “audio transcription isn’t available in all countries or regions” without publishing the language list.

Does Apple Intelligence pull action items out of a meeting recording?

No. In Notes it generates a summary of the transcript, and on a Voice Memos transcript you get the generic Writing Tools menu — Proofread, Rewrite, tone changes, Summary, Key Points, List, Table. Key Points is a bulleted summary, not an owner-and-deadline extraction.

Can Apple transcribe Russian or Ukrainian?

Not in any of the three language lists Apple publishes for this. Notes transcription covers ten languages, Apple Intelligence overall sixteen, and the Notes transcription-summary locale list is narrower still. Russian and Ukrainian appear in none of them.

Try it on your next call

Narada records your Mac's audio and your microphone, transcribes on-device, and hands you the decisions and follow-ups. Free to use, no account, no bot in the meeting.