$ organize my voice memos
Skill · Claude Code
Voice Memo
Organizer.
The recipe a parent dictated. The idea you mumbled on a walk. The voicemail you can't bring yourself to delete. Already on your phone. This is how you find them.
329 untitled recordings. 58 hours of audio. 8 years of "New Recording 47." One skill that renames, transcribes, tags, and indexes the whole folder — locally on your Mac, then syncs the new titles back to your iPhone.
Free · No API key · Runs locally · Nothing uploaded · Requires a Claude Code, Codex, or Claude subscription
What it does
Rename.
Transcribe.
Tag.
Store.
Four operations on the whole Voice Memos folder. The transcription runs on your laptop — the audio never leaves it.
- 01
Rename.
Every recording gets a real title that describes what's actually inside. The new titles sync back to your iPhone through iCloud.
- 02
Transcribe.
whisper.cpp or Parakeet runs locally on your Mac at about 60× realtime. 58 hours of audio finishes in roughly an hour.
- 03
Tag.
Claude pulls themes, key quotes verbatim, and a one-line summary out of each transcript.
- 04
Store.
Everything lands in a searchable master index you can grep through in Obsidian, VS Code, or any text editor.
Walkthrough
How I gave 329 untitled voice memos real titles — on my iPhone.
Walkthrough video
Coming soon.
Want it the minute it's up? Drop your email below — that's the same list that gets the install guide.
Free path
Install it now.
If you already use Claude Code, this is a five-minute job. The skill is one file. Curl it into your skills folder and ask Claude to organize.
- 01
Install Claude Code.
- 02
Curl the skill file into
~/.claude/skills/.mkdir -p ~/.claude/skills/voice-memo-organizer curl -fL -o ~/.claude/skills/voice-memo-organizer/SKILL.md \ https://raw.githubusercontent.com/cathrynlavery/voice-memo-organizer/8c36182da26a27e4a19e4ada95cfe2b2f2342b82/SKILL.md curl -fL -o ~/.claude/skills/voice-memo-organizer/TROUBLESHOOTING.md \ https://raw.githubusercontent.com/cathrynlavery/voice-memo-organizer/8c36182da26a27e4a19e4ada95cfe2b2f2342b82/TROUBLESHOOTING.mdPinned to the reviewed commit — not a moving branch.
- 03
Open Claude Code and say
organize my voice memos.Claude will walk you through Full Disk Access, finding the hidden Voice Memos folder, installing
ffmpegand whisper.cpp or Parakeet, transcribing every recording, and building the index.
Open source. Read it before you run it.
Email path
Want me to walk you through it?
If you've never opened a terminal — or never installed Claude Code, never enabled Full Disk Access — drop your email. I'll send the full install guide with screenshots, plus the link to the skill. And future skills as I publish them.
One email with the guide. Unsubscribe anytime.
- → Full Disk Access in one screenshot
- → Which transcription engine to pick (Parakeet vs whisper.cpp)
- → What to do if your memos live on your iPhone
- → The Voice Memos UI path that syncs new titles back to iPhone
What you get back
One searchable archive. Originals untouched.
~/Documents/Voice-Memos-Organized/
├── voice-memos-master-index.md
├── transcripts/
│ ├── 2017-03-14-recipe-from-mum.md
│ ├── 2019-08-02-walk-idea-bestself-buyback.md
│ ├── 2022-11-09-voicemail-grandma.md
│ └── …
├── summaries/
├── recordings-metadata.csv
└── models/
~/Documents/Voice-Memos-Raw/
The useful file is voice-memos-master-index.md:
every memo, in chronological order, with a real title, summary,
themes, and the best quotes pulled out verbatim. Search it in
Obsidian, VS Code, or grep.
The originals stay where Apple put them. Nothing is deleted. Run the skill again next month — it picks up only the new recordings.
A few notes
-
macOS only.
The skill reads Apple's local Voice Memos database. There's no Windows or Linux equivalent.
-
Full Disk Access required.
Apple stores Voice Memos in a protected folder. Your terminal needs Full Disk Access in System Settings.
-
Memos on your iPhone?
Open Voice Memos on your Mac first and let iCloud sync. The skill reads from your Mac's local copy.
-
Big archives take time.
Apple Silicon transcribes at about 60× realtime, so 58 hours of audio finishes in roughly an hour. It's a coffee job, not a project.