CV

cathrynlavery/voice-memo-organizer

Developer tools
54 stars 품질 40 트렌드 40

Claude Code skill that organizes hundreds of Apple Voice Memos into a searchable archive with transcriptions, summaries, and key quotes. No API keys needed.

개요

A Claude Code skill that turns hundreds of untitled Apple Voice Memos into a searchable, organized archive — with descriptive titles, summaries, themes, and key quotes. Transcription runs 100% locally on your Mac — no API keys needed. The skill supports two engines: (recommended, faster + better punctuation, ~2.3 GB model) or (lighter, ~150 MB model, better fit for Macs with less RAM). Summarization is handled by Claude. Your Apple Voice Memos are buried in a hidden macOS folder, named things like "New Recording 247." This skill: 1. all your Voice Memos (they're at a non-obvious path that requires Full Disk Access) 2. metadata from Apple's SQLite database (dates, durations, display titles) 3. every recording locally — choose Parakeet MLX (recommended) or whisper.cpp (lighter); both run ~60× realtime on Apple Silicon in batch mode 4. each transcript with a descriptive title, themes, key quotes, and type classification 5. a searchable master index document 6.

README

Voice Memo Organizer

A Claude Code skill that turns hundreds of untitled Apple Voice Memos into a searchable, organized archive — with descriptive titles, summaries, themes, and key quotes.

Transcription runs 100% locally on your Mac — no API keys needed. The skill supports two engines: Parakeet (recommended, faster + better punctuation, ~2.3 GB model) or whisper.cpp (lighter, ~150 MB model, better fit for Macs with less RAM). Summarization is handled by Claude.

What It Does

Your Apple Voice Memos are buried in a hidden macOS folder, named things like “New Recording 247.” This skill:

  1. Finds all your Voice Memos (they’re at a non-obvious path that requires Full Disk Access)
  2. Extracts metadata from Apple’s SQLite database (dates, durations, display titles)
  3. Transcribes every recording locally — choose Parakeet MLX (recommended) or whisper.cpp (lighter); both run ~60× realtime on Apple Silicon in batch mode
  4. Summarizes each transcript with a descriptive title, themes, key quotes, and type classification
  5. Builds a searchable master index document
  6. (Optional) Renames memos through the Mac Voice Memos app UI so the new titles sync to your iPhone

Before → After

Before: 329 recordings named “New Recording 1” through “New Recording 329”

After: A searchable markdown index with entries like:

  • “90-Day Focus Sprint Framework” (productivity, goal-setting)
  • “Bedtime Story About the Brave Little Fox” (personal, family)
  • “Legal Call Re: Buying Back Company from PE” (business, legal)

Each with summaries, themes, and key quotes extracted. With Step 7, Claude drives the Mac Voice Memos UI to rename each memo, and the same titles appear in the Voice Memos app on your iPhone.

Start Here: Beginner Walkthrough

If you have never installed a Claude Code skill before, follow these steps in order.

1. Open Terminal

  1. Press Command + Space.
  2. Type Terminal.
  3. Press Return.

All commands below are pasted into Terminal unless a step says otherwise.

2. Make sure your Voice Memos are on your Mac

  1. Open the Voice Memos app on your Mac.
  2. Wait until your iPhone recordings appear there.
  3. Leave Voice Memos closed when Claude starts processing, unless Claude asks you to open it.

3. Install the required apps

You need Claude Code and Homebrew before installing this skill.

Check whether you already have them:

claude --version
brew --version

If claude is missing, install Claude Code from Anthropic’s setup guide: https://docs.anthropic.com/en/docs/claude-code

If brew is missing, install Homebrew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

After installing Homebrew, close Terminal, reopen it, and run:

brew --version

4. Give Terminal permission to read Voice Memos

Voice Memos live in a protected Apple folder, so this step matters.

  1. Open System Settings.
  2. Go to Privacy & Security.
  3. Open Full Disk Access.
  4. Turn on Full Disk Access for your terminal app, usually Terminal.
  5. Quit Terminal and reopen it.

5. Install this skill

Paste this into Terminal:

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/main/SKILL.md

curl -fL -o ~/.claude/skills/voice-memo-organizer/TROUBLESHOOTING.md \
  https://raw.githubusercontent.com/cathrynlavery/voice-memo-organizer/main/TROUBLESHOOTING.md

6. Run the organizer

Paste this into Terminal:

claude "organize my voice memos"

Claude will guide the rest. It will:

  1. Find your Voice Memos database and audio files.
  2. Ask which transcription engine to use. Pick Parakeet unless you have a low-memory Mac or need a language outside Parakeet’s coverage.
  3. Install transcription tools if needed.
  4. Copy your recordings into ~/Documents/Voice-Memos-Raw/.
  5. Transcribe them locally on your Mac.
  6. Create transcripts, summaries, and a searchable master index.

The first Parakeet run downloads a model of about 2.3 GB, so the first run can take longer than later runs.

7. Optional: Rename the memos in Apple’s Voice Memos app

Only do this if you want the titles inside Apple’s Voice Memos app to change too.

Important:

  • Claude must rename through the Mac Voice Memos app UI.
  • Do not use direct SQLite edits for renaming. They can look changed on the Mac but fail to sync to iPhone.
  • You may need to enable Accessibility permission for the app running Claude: System Settings > Privacy & Security > Accessibility.
  • Claude should test one rename first, ask you to confirm it appears on iPhone, then continue in small batches.

Use this prompt:

claude "rename my useful generic Voice Memos using the voice-memo-organizer skill. Use the Mac Voice Memos UI only, back up the database before each batch, verify CloudKit metadata after each batch, and skip anything under 10 seconds or without a usable transcript."

8. Find your finished files

When the run is done, open:

open ~/Documents/Voice-Memos-Organized

The main file to read is:

voice-memos-master-index.md

Quick Install Reference

Use this shorter version if you are already comfortable with Terminal and Claude Code skills.

Prerequisites

Setup

  1. Enable Full Disk Access for your terminal: System Settings → Privacy & Security → Full Disk Access → enable your terminal app (Terminal, iTerm2, etc.)

  2. Install the skill:

    # Create the skills directory
    mkdir -p ~/.claude/skills/voice-memo-organizer
    
    # Download the skill and troubleshooting reference
    curl -fL -o ~/.claude/skills/voice-memo-organizer/SKILL.md \
      https://raw.githubusercontent.com/cathrynlavery/voice-memo-organizer/main/SKILL.md
    curl -fL -o ~/.claude/skills/voice-memo-organizer/TROUBLESHOOTING.md \
      https://raw.githubusercontent.com/cathrynlavery/voice-memo-organizer/main/TROUBLESHOOTING.md
    
  3. Tell Claude Code to use it:

    claude "organize my voice memos"
    

    Claude will walk through each step: finding your memos, installing parakeet-mlx and ffmpeg via Homebrew, transcribing everything, summarizing, and building the index.

What You’ll Get

~/Documents/Voice-Memos-Organized/
├── voice-memos-master-index.md    ← searchable master document
├── transcripts/                   ← full text of every recording
├── summaries/                     ← JSON summaries with titles/themes/quotes
└── recordings-metadata.csv        ← database export with dates & durations

~/Documents/Voice-Memos-Raw/       ← copies of your original audio files

Model files live in ~/.cache/huggingface/hub/ (Parakeet, ~2.3 GB) or ~/Documents/Voice-Memos-Organized/models/ (whisper, ~150 MB) — one-time downloads.

Choosing a Transcription Engine

Parakeet MLX (recommended) whisper.cpp (lighter)
Model size on disk ~2.3 GB ~150 MB (base)
RAM during transcription ~3-4 GB ~500 MB
Speed (batch, Apple Silicon) ~60× realtime ~60× realtime
LibriSpeech test-clean WER 1.69% ~5%
Punctuation/capitalization proper basic
Language coverage English + 24 European per-model (en or multilingual)
Setup pipx install parakeet-mlx brew install whisper-cpp + model download

Default to Parakeet. Choose whisper if your Mac has less than ~16 GB RAM, you want a much smaller disk footprint, or you need a language Parakeet v3 doesn’t cover (Mandarin, Japanese, Arabic, etc.).

Performance

Measured on an M3 Ultra in May 2026: ~60× realtime in batch mode with Parakeet. A library of 329 memos (67 hours) takes roughly 70 minutes end-to-end, including the one-time model download. Whisper.cpp base.en hits similar throughput.

Tips

  • The master index works great in Obsidian, VS Code, or any text editor with Cmd+F
  • Parakeet’s v3 model handles English plus 24 European languages; whisper supports many more via its multilingual ggml-base.bin
  • Works with memos organized into folders in the Voice Memos app
  • Step 7 (optional) must use Mac Voice Memos UI automation for sync-safe renames. Direct SQLite title edits can appear locally on Mac but will not reliably sync to iPhone.

Credits

Built by Cathryn Lavery — founder, builder, and voice memo hoarder.

More tools and frameworks at founder.codes.

View this README on GitHub

추천 도구

다른 키워드를 입력하거나 필터를 제거해 보세요.

설치

npx skillfish add cathrynlavery/voice-memo-organizer