Apple's Voice Memos app transcribes recordings by itself — on iPhone 12 or later with iOS 18, and on a Mac only with macOS Sequoia and Apple silicon. Outside those requirements, and outside the roughly ten supported languages, the built-in transcript simply does not appear, and you convert the memo by uploading the file instead.
This guide covers both routes, with the device requirements stated plainly rather than glossed as "modern phones do this now". It also includes a real voice memo you can play, its unedited transcript, and a measured answer to a question nobody else asks: does compressing a memo before uploading change what comes back?
Voice memo transcription at a glance
| Built-in on iPhone | iPhone 12 or later, iOS 18 |
| Built-in on Mac | macOS Sequoia and Apple silicon — Intel Macs excluded |
| Built-in languages | About ten, and not in every country or region |
| If your device misses out | Share the file, upload it — no hardware requirement |
| Voice memo file format | .m4a on iPhone; .m4a, .mp3, .amr or .3gp on Android |
| Conversion needed first | None — and re-compressing actively hurts, see the test |
| Free use | First 5 minutes with no account; 3 files/day at 50 MB with a free account |
| Speed | 0.13× real time measured in the cloud vs 0.3–5× published for local Whisper |
- Hear a real voice memo and its transcript
- Does Voice Memos transcribe? The requirements
- How to get a transcript in the Voice Memos app
- Converting a memo the app will not transcribe
- Never re-compress a memo first — we measured why
- How long it takes: cloud vs your own laptop
- Transcribing memos offline for free
- Free web tools compared
- Turning a memo into actual notes
- Where memo transcription goes wrong
- Test methodology and results
- FAQ
- Sources
Hear a real voice memo and its transcript
Press play. This is a genuine voice note — someone thinking out loud, not a scripted read — and the transcript below the player is the unedited output of our pipeline on this exact file. Click any line to jump to it.
This clip is here because it is *messy in the way real memos are messy*. The speaker opens with "Okay, so…", interrupts himself mid-argument with "I mean, I'm sure the vast majority of people don't use the feature", and repeats a phrase while he finds the end of it: "What a ridiculous idea. What a ridiculous thing to include in the camera app."
That is the honest picture of what transcription gives you. It is faithful, punctuated and searchable — and it is not notes yet. Turning it into notes is a second step.
Does Voice Memos transcribe? The requirements nobody prints
Yes — but "modern smartphones handle this automatically" is not true, and the gap catches a lot of people. These are the actual requirements:
| Device | Requirement | If you fall outside |
|---|---|---|
| iPhone / iPad | iPhone 12 or later with iOS 18 | No transcript on iPhone 11 or earlier, whatever iOS is installed |
| Mac | macOS Sequoia and Apple silicon | No transcript on any Intel Mac, at any macOS version |
| Language | About ten, incl. English, Spanish, French, German, Portuguese, Italian, Japanese, Korean, Chinese | No transcript in an unsupported language |
| Region | Apple states it is "not available in all countries or regions" | Feature may be missing even on qualifying hardware |
| Android | No single built-in equivalent | Depends entirely on your recorder app |
The Mac requirement is the one that surprises people most, and it comes straight from Apple's own support page: transcription needs macOS Sequoia and a Mac with Apple silicon. An Intel Mac running the newest macOS still shows no transcript. There is no setting to enable and nothing to troubleshoot — the feature is simply absent.
If any row above rules you out, skip to the upload route. It has no hardware requirement, works on Android and Windows, and supports far more languages.
How to get a transcript in the Voice Memos app
On a qualifying device this is the fastest path, because the recording is already where it needs to be:
1. Open Voice Memos and select the recording.
2. Tap the transcript icon beneath the waveform — on iPhone, swipe up from the waveform area to reveal it.
3. Read the transcript, and copy the text out to wherever you need it.
Two limits are worth knowing before you rely on it. The built-in transcript is text only — there are no timestamps you can navigate by, no speaker labels, and no export to a document or subtitle file. And it is all or nothing: there is no language override, so a memo recorded in an unsupported language, or one that switches language halfway, gives you nothing rather than a partial result.
For reading back a quick reminder, that is plenty. For an interview, a lecture or anything you need to cite by time, it is not.
Converting a memo the app will not transcribe
The upload route removes every requirement above: no minimum device, no macOS version, no Apple silicon, and 99 languages instead of ten.
1. Get the file. In Voice Memos, tap the recording, then the share button, and save it to Files. You get an `.m4a`. On Android, export or share from your recorder app — you may get `.m4a`, `.mp3`, `.amr` or `.3gp`.
2. Upload the original. Drop it into the form at the top of this page or the full uploader. Do not convert it first — see the next section for the measured reason.
3. Set the language if you know it. Auto-detect is fine for a clear memo in a common language; a fixed language is better for accents, noise, or a short recording.
4. Review and export. Correct names and numbers, then take it as TXT, Markdown, JSON, or — on a paid plan — DOCX, PDF, SRT or VTT.
On AMR and 3GP: Android recorders have used these codecs for years, and several online transcribers reject them outright, telling you to convert to MP3 first. We accept `.amr`, `.amr-wb`, `.3gp` and `.3g2` directly, along with 23 other extensions, so there is no conversion step.
Never re-compress a memo first — we measured why
The common advice to "convert your memo to MP3 before uploading" is not just unnecessary. We tested what it actually costs.
We took the 51-second voice note from the demo above and produced three versions of the same recording: the 320 kbps original, a 96 kbps copy and a 48 kbps copy. All three went through the same engine with the same settings.
| Segment at | 320 kbps (original) | 96 kbps | 48 kbps |
|---|---|---|---|
| 9 of 12 segments came back identical across all three | |||
| 0:30 | …you leave it. | …you leave it. | …you leave it somewhere. |
| 0:33 | So I'm like, can't move it… | So I'm going to can't move it… | Can't move it. Can't use it… |
| 0:38 | That's insane. | It's insane. | That's insane. |
Three things follow, and only the first is obvious.
Compression does not degrade a transcript evenly. Across a 6.7× range of bitrate, three quarters of the transcript was bit-identical. The engine is not quietly getting worse everywhere as quality drops.
The damage is concentrated where the speaker was already unclear. All three disagreements fall inside one eight-second stretch — 0:30 to 0:38 — which is exactly where the speaker abandons a sentence and restarts it. Nothing else in the memo moved.
Each bitrate mis-reads that stretch differently. 96 kbps produced "So I'm going to can't move it", which is not a sentence; 48 kbps dropped the false start entirely. Re-encoding does not add a predictable error — it re-rolls the dice on the ambiguous passage.
So: upload the file your phone recorded. And when you review a transcript, spend your attention on the parts where you know you mumbled — that is where every version disagreed, and where your own memory is the only reliable source.
How long it takes: cloud vs your own laptop
Both routes work. They differ by roughly an order of magnitude, and the AI-generated advice that points you at a local tool for privacy rarely mentions the cost.
| Route | Speed | 1-minute memo | 1-hour memo |
|---|---|---|---|
| Our cloud pipeline | 0.13× real time (measured) | ~8 s | ~8 min |
| Audacity + OpenVINO, base model | ~0.3× real time (published) | ~18 s | ~18 min |
| Audacity + OpenVINO, large-v3 | ~3–5× real time (published) | ~3–5 min | ~3–5 h |
The 0.13× figure is engine time measured on the demo clip: 6.65 seconds for 51.12 seconds of audio. The Audacity multipliers are the ones its own project publishes for the plugin's model sizes. The 1-hour column is arithmetic on those rates, not a measurement — real files vary, and the cloud figure excludes upload time, which for an hour-long memo over mobile data can exceed the transcription itself.
The honest read: local transcription is genuinely free and genuinely private, and on the largest model it can take longer than your afternoon. Choose it when the content demands it, not by default.
Transcribing memos offline for free
If the memo is confidential — a client conversation, a medical note, a legal matter — the argument for local processing is not price but that the audio never leaves the machine.
Audacity is the free route. Its OpenVINO Whisper effect runs OpenAI's Whisper model entirely on your own computer, on Windows, Mac and Linux, with no minute caps, no subscription and no watermark. It offers five model sizes from `base` up to `large-v3`, plus a variant with speaker diarization, and Whisper's 99 languages. The cost is setup — install the plugin, download a model — and the processing time in the table above.
TranscribeNext for Mac is the same idea with less assembly: drag a memo in, it transcribes on-device on Apple Silicon with speaker labels and export formats, and only uses the cloud if you tell it to.
Either way, the point is the same: for a sensitive recording, do not upload it anywhere — including here.
Free web tools compared
Published figures from each vendor, checked July 2026. They change; confirm before relying on them.
| Free tier | File cap | AMR / 3GP | Playable proof | |
|---|---|---|---|---|
| TranscribeNext | First 5 min, no account; 3 files/day with a free account | 50 MB free · 2 GB PRO · 5 GB BUSINESS | Yes, direct | Yes — on this page |
| HappyScribe | First 10 minutes free | 45+ formats | Not published | No |
| AudioScribe | Up to 5 min, no signup | 100 MB | States AMR is not supported | No |
| TalkNotes | 1 minute, no account | 5 MB | Not published | No |
| NoteGPT | Free tier; batch of 20 files | 5 GB per file | Not published | No |
| Audacity (offline) | Unlimited, free | None | Depends on your build's decoders | No |
Two notes. HappyScribe's 10 free minutes are the most generous of the browser tools, and it publishes an accuracy figure — "up to 96% on clear audio" — which, like every vendor number, is measured on the vendor's own audio and is not comparable across tools. That is why this page publishes a clip you can check and a test you can repeat instead.
And our own row is not unlimited: the 5-minute free allowance is a preview limit, and DOCX/PDF/SRT/VTT export sits on the paid plans. Free accounts get TXT, Markdown and JSON.
Turning a voice memo into actual notes
A transcript is not notes, and the demo above shows exactly why: restarts, filler, a repeated phrase. Nobody wants to read that back.
The reliable order is transcribe, then structure — two steps, not one:
1. Transcribe the memo so the words exist as searchable text.
2. Summarise it into decisions, action items and open questions, either with an AI summary or by editing it down yourself.
Doing it in that order matters. A summary generated straight from audio hides which step failed when the result is wrong — you cannot tell whether it misheard you or misunderstood you. With the transcript in hand you can check one, then the other.
This is also the honest answer to "turn voice memo into notes" as a single feature: the tools that offer it are running these two steps back to back. Knowing that lets you fix the transcript before the summary is built on top of it.
Where memo transcription goes wrong
Voice memos fail differently from meetings, because you are usually alone, close to the microphone, and thinking rather than presenting:
None of these are fixed by choosing a different tool. They are fixed by re-reading the passages where you know you were unclear — which is a five-second habit, not a workflow.
Test methodology and results
Test date: 27 July 2026
Source clip: “Timelapse Excess”, from David Blue’s Voice Notes library — a genuine self-recorded voice note, published to the public domain (archive.org, Public Domain Mark; the author’s own repository carries an Unlicense public-domain dedication). 51.1 s, mono, originally 48 kHz at 320 kbps.
Served file: re-encoded with ffmpeg to mono 24 kHz at 48 kbps (307 KB). The original was 2.95 MB, most of it a 4008×4008 embedded cover image. The player above serves this file, and the transcript shown is the transcript of it.
Compression test: the same recording encoded at 320 kbps (original), 96 kbps and 48 kbps, each transcribed with identical settings — Whisper large, language set to English, diarization off, default profile.
Result: 9 of 12 segments identical across all three bitrates. The 3 that differed all fall between 0:30 and 0:38, the stretch where the speaker abandons and restarts a sentence. Each bitrate produced a different reading of that stretch.
Speed: 6.65 s of engine time for 51.12 s of audio (≈0.13× real time). Excludes upload.
Comparison figures: Audacity’s published OpenVINO Whisper multipliers, ~0.3× real time (base) to ~3–5× (large-v3). The hourly column in the speed table is arithmetic on those published rates, not our measurement.
Transcript shown: unedited engine output. Line start/end times are measured; the word-level highlight is interpolated inside each measured line, because our current backend returns segment-level timings rather than per-word ones.
One clip on one day. Published so the claims can be repeated rather than believed.
Last tested: 27 July 2026
FAQ
Does Voice Memos transcribe recordings? Yes, on supported hardware. Apple's Voice Memos app transcribes recordings on iPhone 12 or later running iOS 18, and on a Mac running macOS Sequoia with Apple silicon. Apple's own support page states the Mac requirement plainly: transcription needs macOS Sequoia and a Mac with Apple silicon, and it is not available in all countries or regions. Transcription is offered in about ten languages, including English, Spanish, Portuguese, Italian, French, German, Japanese, Korean and Chinese.
Can you transcribe a voice memo on an older iPhone or an Intel Mac? Not with Apple's built-in transcription. An iPhone 11 or earlier and any Intel Mac are outside the requirements, no matter which iOS or macOS version is installed. On those devices, share the memo as a file and upload it to a transcription tool in a browser, which has no hardware requirement at all.
How do I get a transcript of a voice memo? On a supported device, open the recording in Voice Memos and tap the transcript icon beneath the waveform, then copy the text. On any other device, tap the share button on the recording, save it to Files as an .m4a, and upload that file to a transcription tool. The browser route also gives you timestamps and export formats that the built-in transcript does not offer.
How do I convert a voice memo to text on Android? Android has no single built-in equivalent, so the method depends on your recorder app. Export or share the recording as a file — usually .m4a, .mp3, .amr or .3gp — and upload it to a transcription tool. AMR and 3GP are worth checking before you pick a tool: several online transcribers reject them and tell you to convert to MP3 first, which adds a re-encode you do not need.
Should I compress a voice memo before uploading it? No — upload the original. We ran the same 51-second voice note through our pipeline at three bitrates: the 320 kbps original, a 96 kbps copy and a 48 kbps copy. Nine of the twelve transcript segments came back identical across all three. The three that differed were all inside one eight-second stretch where the speaker restarts his sentence, and each bitrate read that mumble differently. Compression does not degrade a transcript evenly; it re-rolls the dice on the passages that were already unclear.
How long does it take to transcribe a voice memo? Cloud transcription is a fraction of the recording's length; local transcription can be several times its length. Our pipeline took 6.65 seconds of engine time for a 51-second memo, about 0.13× real time. Audacity's OpenVINO Whisper plugin publishes roughly 0.3× real time for its base model and 3–5× for large-v3, so an hour-long memo is minutes in the cloud and potentially hours locally on the largest model.
Is there an app that transcribes voice memos? On a supported iPhone the Voice Memos app itself is the app — no installation needed. Beyond that, an app is worth installing in two cases: when you want every recording transcribed automatically rather than one at a time, and when the recordings are confidential and you want them transcribed on-device so nothing is uploaded. For occasional conversions a browser is enough.
How do I turn a voice memo into notes? Transcribe it first, then structure the text. A raw transcript of someone thinking aloud is not notes — it contains restarts, filler and repetition, as the demo on this page shows. Get the transcript, then use an AI summary to pull out the decisions and action items, or edit it down yourself. Transcribing and summarising are two separate steps, and doing them in that order is what makes the result usable.
Can I transcribe voice memos offline for free? Yes. Audacity's OpenVINO Whisper plugin runs OpenAI's Whisper model entirely on your own computer, free on Windows, Mac and Linux, with no minute caps or subscription, and the audio never leaves the machine. The trade-off is time and setup: you install a plugin, download a model, and processing takes roughly 0.3× to 5× the recording's length depending on the model you choose.
What audio formats do voice memos use? Apple Voice Memos saves as .m4a. Android recorders vary: .m4a, .mp3, .amr and .3gp are all common. None of them need converting before transcription — TranscribeNext accepts 27 audio and video extensions including .amr, .3gp and .3g2 directly, so you can upload whatever your phone produced.
Sources
Related guides: Voicemail to text · How to convert M4A to text · How to transcribe audio in Word · Best Mac apps to transcribe audio
Transcribe your voice memo now
If your iPhone qualifies, the transcript is already in the Voice Memos app — open the recording and swipe up. If it does not, share the memo as a file and drop it into the form at the top of this page: the first 5 minutes are free with no account. For memos you would rather not upload, TranscribeNext for Mac transcribes them on-device.
Voice Memos saves as M4A, so M4A transcription has the format specifics. To move around a long memo afterwards, see timestamped transcription.