Achhina's Digital Garden

Voice Memo to Journal Automation

3 min read

Objective

Automate the flow from voice memos → transcription → markdown notes in the Obsidian vault, so spoken notes can be captured on the go and later processed into journal entries.

Context

Previously used an Aiko-based Apple Shortcut (“Start Transcribing”) that recorded audio, ran it through Aiko’s on-device Whisper model, and saved the transcript as a markdown file. That shortcut broke on this Mac because Aiko is no longer installed, and two of its six actions show as “Unknown Action.”

Apple’s native on-device transcription has caught up — their new SpeechAnalyzer/SpeechTranscriber APIs (macOS 26 Tahoe / iOS 18+) are 2.2x faster than Whisper Large V3 Turbo with comparable accuracy (per MacStories and 9to5Mac benchmarks). No reason to depend on Aiko anymore.

Current State

MemoToJournal shortcut — installed and syncing via iCloud.

Flow:

  1. Accept audio from Share Sheet (Voice Memos) or record new audio if run standalone
  2. Transcribe using Apple’s native on-device “Transcribe Audio” action
  3. Format two date variables — DateString (yyyy-MM-dd-HHmm) for filename, ISODate (yyyy-MM-dd) for frontmatter
  4. Build markdown with YAML frontmatter (date, tags: voice-memo + transcript, source: voice-memo)
  5. Save as voice-memo-{DateString}.md to vault Inbox
  6. Show confirmation notification

Known limitations:

  • The “Transcribe Audio” action requires macOS 26 Tahoe or iOS 18+. Current Mac is on macOS 15.7.6 (Sequoia) — the transcription step shows as “Unknown Action” and won’t run here.
  • Works on iPhone (iOS 18+). Mac support pending Tahoe upgrade.
  • Save destination needs to be manually pointed at the vault’s Inbox/ folder on first run (one-time setup).

Prior Art Reviewed

  • “Start Transcribing” (own shortcut): 6 actions — Record audio → Aiko transcribe → Wait → Get clipboard → Set name to aiko-note-{date}.md → Save (missing app). Broken on this Mac.
  • “Add Note” (own shortcut): 4 actions — separate quick-note shortcut, not transcription-related.
  • RoutineHub “Local Transcribe to Obsidian”: Uses Aiko/Whisper locally, saves to vault. Same Aiko dependency.
  • RoutineHub “Transfer Voice Memos to Obsidian”: Uses Just Press Record, companion cleanup shortcut.
  • Pedro Amini’s gist: macOS script that extracts Apple’s native transcript from Voice Memo file’s tsrp MP4 atom. No Whisper needed, hourly cron. Interesting but macOS 15+ only and requires filesystem access to Voice Memos container (blocked by sandbox).

Next Actions

  • Point Save File action to Obsidian vault Inbox/ folder on iPhone (one-time setup)
  • Test end-to-end on iPhone: record memo → share to MemoToJournal → verify .md appears in Inbox
  • Upgrade Mac to macOS 26 Tahoe (post-Thailand trip) to enable Mac-side transcription
  • After Tahoe upgrade, test on Mac and verify Transcribe Audio action resolves
  • Consider adding a “check for existing transcript” step once on Tahoe (Voice Memos auto-generates transcripts on iOS 18+ — could skip re-transcription)

Resources

Notes

The long-term vision is: speak a note on the go → transcript lands in Inbox → in a Cowork session, ask Claude to read the Inbox transcripts and weave them into proper journal entries in Chronicles. This keeps the capture fast and frictionless while the journaling stays high-quality.

Graph View