ReelTexts

VOICE MEMOS, TRANSCRIBED

How to Transcribe a Voice Memo (iPhone, Mac, and the M4A Trap)

If your iPhone runs iOS 18 or your Mac runs macOS Sequoia, transcribing a voice memo takes two taps. Record it — or open an old one — in Voice Memos, switch from the waveform to the transcript view, and the words are already there. Select them, copy them, paste them anywhere. No app to install, no account, nothing to pay.

The rest of this page is about the parts that break: the import rule that quietly rejects MP3 and WAV, the missing timecodes, and the settings that decide whether the transcript comes back readable or garbled.

What every route has in common

Record on the iPhone, read the transcript on the Mac — Voice Memos syncs everything through iCloud, so the memo made in the car shows up on your desk already transcribed. The transcription itself is free and built in. No export step either: the text lives next to the waveform, and on the Mac it selects like any document text — Command-A, Command-C, paste.

One caveat comes straight from Apple's own support documentation: on the Mac side, transcription requires macOS Sequoia on an Apple-silicon Mac, and the feature is not offered in every country or region. On the iPhone it arrived with iOS 18. Older devices get none of this.

The M4A trap

Here is where the "no extra software" promise quietly collapses.

Voice Memos on the Mac accepts imported audio — but only M4A files. Drag an MP3 in and nothing happens. Drag a WAV in and nothing happens. No error, no explanation; the import just refuses. The app records in AAC audio wrapped as M4A, and that is the only container it will take from the outside.

The same Mac, a different built-in app, opposite behavior: Notes also transcribes recordings in Sequoia, and Notes happily accepts MP3 and WAV attachments. So "import doesn't work" is wrong at the level of the operating system and true at the level of one app. If Voice Memos refuses your file, Notes may not.

Two conversion paths need nothing installed: QuickTime Player's File → Export As → Audio Only, or right-click → Services → Encode Selected Audio Files, whose Spoken Podcast preset also outputs AAC. Videos go through the same service under a different name — Encode Selected Video Files — with an Audio Only choice. Formats that survive neither route need a real converter.

The sync gap between short and long memos

A memo under thirty seconds shows up on the Mac already transcribed, usually before you have put the phone down. A memo over two minutes does not. The recording itself syncs fast — the waveform appears within seconds — but the transcript lags behind it by a margin that widens with length, and there is no progress indicator. The Mac sits there with a playable waveform and an empty transcript pane, and the only signal that anything is still happening is that the text arrives in one sudden block, fully formed, somewhere between thirty seconds and several minutes later. Nothing in the app tells you this. The first time it happens you assume the transcript failed and start looking for a retry button that does not exist. There is no setting that changes this — the only thing that helps is walking away and coming back.

No timestamps, by design

The transcript is one continuous block of plain text. No timecodes, no speaker labels. So it cannot become SRT or VTT — no synced captions without timing it yourself, and most caption workflows refuse bare text. The notable exception is YouTube, which accepts pasted plain text and estimates the timing on its own.

Details only one walkthrough mentions

These never make it into the standard tutorials, and two of them change how you use the app.

Turn on location sharing before you record. An unnamed memo then names itself after your address, down to the unit number, instead of sitting in the list as "New Recording 7" — which matters more the longer your list gets.

Clean the audio before you transcribe. Enhance Recording reduces the background hiss and muffled sound, and Skip Silence compresses the dead air. Both are built into the same edit panel. The transcription engine is only as good as the signal you hand it, and a memo recorded across a table is a different input than one held to your ear.

Deleted memos are recoverable. Voice Memos keeps a Recently Deleted folder; a recording trimmed in panic yesterday is still sitting there.

On the Mac, the transcript is also a navigation tool. Click any word and the audio jumps to that moment; during playback the current word bolds itself, so you can follow along and skip the waveform entirely. And if the Mac runs Apple Intelligence, selecting the transcript opens Writing Tools — Summarize, Make Key Points, Make List — which turns an hour of talking into a page of bullets without leaving the app.

What to expect from the output

Expect imperfection. The built-in transcription is dramatically better than what third-party apps managed a few years ago, but it still stumbles on names, product terms, and numbers — anything a machine cannot infer from context. Budget a proofreading pass for exactly those words and skim the rest.

How the browser tool handled a known script — measured

Numbers need a source, so we made one. A single script was read aloud by a synthesizer and recorded at four lengths — 16, 42, 77, and 129 seconds — then each file was dropped into the converter on this site, in one browser session with the model already cached. The files were 16 kHz mono WAVs, generated and transcribed on one machine — nothing was uploaded at any point, the same no-upload rule this page's converter enforces (25 MiB and 5 minutes, both checked before decoding starts). Timing runs from the click on Transcribe to the last word on screen. Because the script is known word for word, every output can be checked: an error is any word the transcript got wrong, and when the model writes numerals where the script spelled them out ("8452" against eight four five two), that counts as an error too — counted against the tool, not for it.

RecordingWords readWords outWrong wordsError rateTranscription time
16 s4845714.6%16.3 s
42 s122884133.6%27.9 s
77 s22931210345.0%104 s
129 s35522114841.7%147.8 s

The short file is the encouraging one: seven wrong words out of forty-eight. The rest is the honest picture. On the 77-second file the transcript came back with 312 words against 229 read — the model repeated itself and padded the middle. On the 129-second file it went the other way: 221 words out, and two whole sentences from the middle of the script simply never arrived. Proper nouns took the worst of it: Okafor came back as "Okafur", and "Eowyn Stallworth" arrived as "Eau and Stahlworth". None of this hides — the text shows exactly where it broke — but it is why the proofreading pass exists, and why these numbers are printed next to the samples that produced them.

If the memo never touched an Apple device — an Android export, a downloaded interview, anything MP3 or WAV — skip the format dance: drop the file into the WAV to text converter and it transcribes in the browser, same plain text, same proofreading pass. Coming from an iPhone voice memo that will not import, the M4A to text converter reads the file Voice Memos itself records.

FAQ

Does transcribing a voice memo cost anything?
No. On iOS 18 and macOS Sequoia the built-in transcription is free and runs on the device. There is no quota and no watermark.
Why won't Voice Memos import my MP3 or WAV?
On the Mac, Voice Memos only accepts M4A files from outside. Convert with QuickTime Player (File → Export As → Audio Only) or the Encode Selected Audio Files service, or skip conversion and use a browser tool that reads MP3 and WAV directly.
Does the built-in transcript have timestamps?
No. It is one block of plain text, with no timecodes and no speaker labels. For synced captions you need timed text — YouTube accepts plain text and estimates the timing itself.
Why is the transcript missing on my Mac for a long memo?
The recording syncs fast; the transcript trails it by a margin that grows with length, and there is no progress indicator. It arrives in one block, up to several minutes later. Waiting is the only fix.
Can I transcribe a voice memo without an Apple device?
Yes — a browser tool handles it. The file stays on your machine, there is no upload step, and you get the same plain text to proofread and copy.