Transcribe Voice Memos to Markdown on Mac and iPhone

You have a Voice Memo. Maybe it’s a thought you recorded walking to the car, a meeting you captured on your iPhone before you had a real notetaker open, or an interview you did on the fly. Apple’s Voice Memos app held onto the audio faithfully — and then left you exactly where recording it did: with an audio file and nothing readable. If you want that memo as text you can search, edit, paste into a doc, or drop into your notes vault, you’re stuck doing it by hand, or pasting the audio into some cloud transcription site you’d rather not trust with it.

The job most people actually want is narrow and specific: transcribe a Voice Memo into Markdown — clean text with headings and bullets, ready to live in Obsidian, Notion, Bear, or just a .md file — without the audio leaving the device. This post is about how to do that on a Mac or iPhone with everything running on-device, and where it fits in a workflow.

Why Voice Memos stops at audio

Apple’s Voice Memos is a recorder, and a good one. It captures clean audio, syncs across your devices through iCloud, and never gets in your way while you’re talking. What it doesn’t do is turn that audio into structured text. Recent versions surface a basic transcript view, but it’s a raw transcript — a wall of text, not formatted notes — and it isn’t built to export Markdown or hand the result to your notes app in a usable shape.

So the gap is real: the audio is captured and synced, but the text — the searchable, editable, paste-able artifact you actually wanted — isn’t there. Bridging that gap is usually where people start looking for a way to transcribe Voice Memos to Markdown specifically, rather than to a plain blob of text they then have to reformat by hand.

Why Markdown is the right target

Markdown is the lingua franca of modern notes. Obsidian and Bear are Markdown-native. Notion, Craft, and Logseq import or paste it cleanly. It’s plain text, so it’s future-proof, diff-able, and searchable with grep. A transcript exported as Markdown — with a title, a date, headings, and bullet points for the key moments — drops straight into your existing system instead of becoming another orphaned file you have to process.

The difference between “a transcript” and “Markdown notes” is structure. A raw transcript is one long paragraph of everything said. Markdown notes have a shape: a heading, a short summary up top, action items as a list, and the full transcript below if you want it. That structure is what makes the result usable a week later instead of something you have to re-read in full.

Doing it on-device: the privacy reason this matters

There’s no shortage of websites that will transcribe an audio file. The problem is what you hand them. A Voice Memo is often the most personal thing on your phone — a private thought, a conversation, a medical note-to-self, an idea you don’t want indexed by anyone. Uploading it to a cloud transcription service to get Markdown out the other side means that audio now lives on someone else’s server, under someone else’s retention policy.

On-device transcription removes that tradeoff entirely. On macOS 26 and iOS 26, Apple’s SpeechAnalyzer framework transcribes audio on the Neural Engine — locally, with no network round trip. The audio never leaves the device, and you still get accurate text. The same on-device principle that drives on-device meeting transcription applies just as well to a single-speaker Voice Memo: there’s no reason a private recording should travel to a cloud just to become text.

How to transcribe a Voice Memo to Markdown with Dictanta

Dictanta imports an existing Voice Memo, transcribes it on-device, and exports clean Markdown. The flow is short:

1. Open Dictanta and import the memo. Dictanta reads from your Voice Memos library (and any .m4a/.wav/.mp3 audio file). Pick the recording you want. Because Voice Memos syncs through iCloud, a memo recorded on your iPhone is available to import on your Mac.
2. Let it transcribe on-device. SpeechAnalyzer produces the transcript locally — roughly 55% faster than Whisper v3 Turbo on the same chip per Apple’s benchmarks, and with no audio leaving the machine.
3. Get the structured notes. Apple’s on-device Foundation Models generate a short summary and pull out the key points and any action items — locally, no cloud LLM. So you get a titled, summarized note, not just a raw transcript dump.
4. Export as Markdown. Dictanta exports the result as a .md file — heading, summary, bullets, and the full transcript underneath — ready to drop into Obsidian or Bear, or paste into Notion.

The whole path — import, transcribe, summarize, export — runs on the device. The memo and its text never touch a server.

What the Markdown actually looks like

A Voice Memo of a quick project debrief comes out as something you can use immediately:

# Project debrief — 2026-06-15

## Summary
Walked through the Q3 launch blockers. Design is ready; the API
migration is the long pole. Agreed to cut the analytics rework
from v1 to hit the date.

## Key points
- Design sign-off is done, no open items
- API migration is the critical path — owner: me
- Analytics rework moved to v1.1 to protect the launch date

## Transcript
So the main thing on the launch is the API migration, everything
else is basically ready to go...

That’s the difference between a transcript and notes. The summary and key points are generated on-device from the transcript; the full transcript stays at the bottom for when you need the exact words. Paste the whole thing into a daily note, or save the .md straight into your vault.

Voice Memos to Obsidian or Notion

The two most common destinations:

Obsidian. Obsidian’s notes are Markdown files. Export the .md from Dictanta and save it into your vault folder — it shows up as a note with working headings, lists, and (if you add them) links. No conversion step, because there’s nothing to convert.

Notion. Notion pastes Markdown well: copy the exported Markdown and paste it into a Notion page, and the headings and bullets come across as native blocks. For a recurring workflow, keep a “Voice notes” database and paste each memo’s notes in as a new entry.

Bear, Craft, Logseq. All Markdown-native or Markdown-friendly; the exported file or pasted text lands cleanly. The point of targeting Markdown rather than a proprietary format is that it works everywhere your notes already live.

When this is the workflow you want — and when it isn’t

Transcribing Voice Memos to Markdown is the right move when:

• You already have the recording. The audio exists in Voice Memos and you need it as text. This is import-and-transcribe, not record-from-scratch.
• It’s single-speaker. A memo, a thought, a self-recorded note. One voice into the mic is exactly what Voice Memos captures and what transcribes most cleanly.
• The destination is your notes app. Markdown is the bridge into Obsidian, Notion, Bear, and the rest.

It’s not the right workflow when the source is a live two-sided meeting. Voice Memos records your microphone — it captures you and whatever your mic picks up in the room, not the far-end audio of a Zoom or Teams call coming out of your speakers. For meetings you want system-audio capture, which records both sides with no bot in the call. If you find yourself reaching for Voice Memos to capture calls, that’s the signal you want meeting recording instead — covered in the Zoom and Otter alternative guides. Voice Memo import is for the recordings you already have; meeting capture is for the calls you’re about to have.

A note on accuracy

On-device transcription quality depends on the recording. A Voice Memo recorded close to the mic in a quiet room transcribes cleanly. One recorded in a noisy café or with the phone in a pocket will have errors, the same as any transcription would. The editor lets you fix mistakes before exporting, so the Markdown you save is clean — but the better the source audio, the less editing you’ll do. If you’re recording a memo specifically to transcribe it later, hold the phone close and find a quiet spot.

Cross-device: record on iPhone, process on Mac

This is where the Apple-ecosystem angle pays off. Voice Memos syncs your recordings through iCloud, so a memo you dictated on your iPhone on the walk home is already on your Mac by the time you sit down. Import it into Dictanta on the Mac — the bigger screen and keyboard make editing the transcript and tidying the Markdown faster — and the whole thing still runs on-device. Or do it all on the iPhone if that’s where you are. The recording and its text move with you across devices because they live in your iCloud and on Apple silicon, not in a third-party app’s cloud.

Isn’t the built-in Voice Memos transcript enough?

Recent versions of Voice Memos can show a transcript, and for a quick “what did I say” glance it’s fine. The reason it doesn’t end the workflow is the same reason this post exists: a glance isn’t an artifact. The built-in transcript is a raw, unstructured view inside the Voice Memos app — there’s no summary, no key-points list, and no clean Markdown export shaped for a notes vault. You’re still looking at a wall of text trapped in an app, not a .md file with a heading and bullets sitting in Obsidian.

The two differences that matter:

• Structure. The built-in view is the transcript and nothing else. Dictanta runs Foundation Models over it on-device to produce a titled summary and a key-points list, so what you export is notes, not a dump.
• Destination. Voice Memos doesn’t export Markdown to your notes app. Dictanta does — a .md file ready for Obsidian or Bear, or Markdown you paste straight into Notion.

If all you ever need is to skim what a memo said, the built-in transcript covers it. If you need the memo to become a note you keep, that’s the gap.

Clearing a backlog

If you’ve accumulated months of Voice Memos meaning to “deal with later,” the practical move is to batch a session: open Dictanta on the Mac, import the memos one after another, and export each as Markdown into a single inbox folder in your vault. Because everything runs on-device, there’s no upload queue and no per-file cloud cost — the only limit is how fast SpeechAnalyzer chews through the audio, which on Apple silicon is faster than real time. An afternoon clears a year of orphaned recordings into searchable, filed notes.

Bottom line

Voice Memos is great at holding onto audio and lousy at giving you text. The job — transcribe a Voice Memo into clean Markdown you can drop into Obsidian, Notion, or Bear — has a simple on-device answer on macOS 26 and iOS 26: import the memo, let SpeechAnalyzer transcribe it locally, let Foundation Models structure it into a summary and bullets, and export Markdown. Nothing about the recording or its text touches a cloud service.

If you have a backlog of Voice Memos you’ve been meaning to turn into notes, that’s exactly the gap Dictanta fills — free for your first three recordings with no length cap, which is enough to run a few real memos through it and watch them come out as Markdown that lands straight in your notes. Paid tiers are $9.99/mo, $79.99/yr, or $149.99 lifetime.