Automating Voice Memos

Automating Voice Memos (no dictionary, data+UI hybrid)

Relationship to the macOS automation skill

• This is a standalone skill for Voice Memos automation.
• For setup help, see the automating-mac-apps skill for permissions (Full Disk Access, Accessibility) and the ObjC bridge basics.
• Prerequisites: Basic JXA (JavaScript for Automation) knowledge; install via macOS System Preferences > Security & Privacy.

Core Framing

• Catalyst App: Voice Memos is an iOS app adapted for macOS without full macOS APIs, hence no AppleScript dictionary for automation.
• UI-first: Prefer UI scripting/keyboard shortcuts for exports to avoid touching the database/container.
• Data (optional): Use data-layer control only if needed (ObjC + sqlite3). This requires broader permissions.
• Permissions: Accessibility for UI automation; Full Disk Access only if you read/write the container/DB.

Workflow (default)

1) UI-first (no FDA): export via menu/shortcut to a folder you control. 2) Optional data path: resolve storage paths; query CloudRecordings.db (Apple epoch +978307200) only if required. 3) For UI actions (recording/export), drive the app with System Events (shortcuts preferred over clicks). 4) For editing, prefer external tools (ffmpeg) after export; avoid writing directly into the container unless you accept FDA.

Quickstart (UI-only export; no Full Disk Access)

• Open Voice Memos and select a recording.
• Export/share UI: the Share sheet can send audio to Notes/other apps (no direct File > Export).
• Transcript copy (UI-only, no FDA):
• Run: osascript skills/automating-voice-memos/scripts/copy_transcript_to_file.applescript "/path/to/output.txt"
• Defaults: Desktop/voice-memo-transcript.txt if no arg.
• Shows transcript (if available), selects all, copies, and writes to the target file via clipboard.

Permissions

• Accessibility: System Settings > Privacy & Security > Accessibility > Enable for System Events and your automation app (Terminal/Python/Script Editor).
• Full Disk Access: Only if you read/write the Voice Memos container/DB directly. UI-only exports do not require FDA.
• Verify: For UI-only, confirm you can open the Export menu and interact with the save dialog. For data access, verify ${home}/Library/Group Containers/group.com.apple.VoiceMemos.shared/.

Validation Checklist

After implementing Voice Memos automation:

• [ ] Verify Accessibility permissions granted for Terminal/Script Editor
• [ ] Confirm Voice Memos app opens and responds to UI scripting
• [ ] Test transcript copy script executes without errors
• [ ] Validate output file contains expected transcript text
• [ ] For data access: verify Full Disk Access and database path exists

Troubleshooting

• Permission denied: Verify Full Disk Access and Accessibility permissions.
• Database locked: Close Voice Memos app before querying.
• File not found: Check macOS version for correct storage path.
• UI automation failures: Ensure Voice Memos is focused and Accessibility is enabled.

When Not to Use

• For cross-platform audio recording (use ffmpeg or platform-agnostic tools)
• When you need programmatic audio capture (use AVFoundation directly)
• For iOS Voice Memos automation (no API available)
• When Full Disk Access cannot be granted

What to load

• Start with basics & prerequisites: automating-voice-memos/references/voice-memos-basics.md (setup and core concepts).
• UI automation: automating-voice-memos/references/voice-memos-ui.md (shortcuts, AX scripting).
• Data access (if needed): automating-voice-memos/references/voice-memos-data.md (storage paths, epochs, database schema).
• Recipes: automating-voice-memos/references/voice-memos-recipes.md (exports, transcripts, recording control).

Voice Memos JXA Basics (no dictionary)

• Voice Memos is a Mac Catalyst app; no AppleScript dictionary. All automation is data-layer + UI scripting.
• Permissions:
• Full Disk Access to read/write the Group Container and CloudRecordings.db.
• Accessibility to drive UI/keystrokes.
• Storage paths:
• Sonoma/Sequoia: ~/Library/Group Containers/group.com.apple.VoiceMemos.shared/Recordings
• Older (fallback): ~/Library/Application Support/com.apple.voicememos/Recordings or ~/Library/Containers/com.apple.VoiceMemos/Data/Library/Application Support/Recordings
• Database: CloudRecordings.db (Core Data SQLite) inside the Recordings folder.
• Core Data epoch offset: 978307200 seconds (Apple epoch). JS date = (zdate + 978307200)*1000.
• Prefer data access first; use UI only for actions that must touch the app (record, transcript view).

Voice Memos Data & SQLite Access

Paths

• Primary (Sonoma/Sequoia): ~/Library/Group Containers/group.com.apple.VoiceMemos.shared/Recordings
• Legacy fallback: ~/Library/Application Support/com.apple.voicememos/Recordings
• DB: CloudRecordings.db inside the Recordings folder.

Resolve home with ObjC:

ObjC.import('Foundation');
const home = $.NSFileManager.defaultManager.homeDirectoryForCurrentUser.path.js;

Core Data timestamps

• Apple epoch offset: 978307200 seconds.
• Convert: new Date((zdate + 978307200) * 1000)

Key tables/columns (CloudRecordings.db)

• ZCLOUDRECORDING
• ZPATH: filename (e.g., 20240101-120000.m4a)
• ZCUSTOMLABEL: user-visible title
• ZDATE: creation timestamp (Core Data epoch)
• ZDURATION: seconds
• ZTRASHEDDATE: non-null => recently deleted
• ZFOLDER: FK to ZFOLDER.Z_PK
• ZFOLDER
• Z_PK: primary key
• ZTITLE: folder name
• ZUUID: stable UUID

SQLite query (export-friendly)

const sql = `SELECT ZPATH, ZCUSTOMLABEL, ZDATE FROM ZCLOUDRECORDING
 WHERE ZTRASHEDDATE IS NULL;`;
const cmd = `sqlite3 -separator "|" "${dbPath}" "${sql}"`;
const out = app.doShellScript(cmd);

File copy with NSFileManager

ObjC.import('Foundation');
const fm = $.NSFileManager.defaultManager;
if (fm.fileExistsAtPath(src)) {
  const ok = fm.copyItemAtPathToPathError(src, dest, null);
}

Safe renaming

• Sanitize title: title.replace(/[\/\\:]/g, "-")
• Avoid collisions by appending counter if destination exists.

Voice Memos Recipes

Batch export (data-first)

ObjC.import('Foundation');
const app = Application.currentApplication();
app.includeStandardAdditions = true;
const fm = $.NSFileManager.defaultManager;
const home = fm.homeDirectoryForCurrentUser.path.js;
const root = home + "/Library/Group Containers/group.com.apple.VoiceMemos.shared/Recordings";
const dbPath = root + "/CloudRecordings.db";
const outDir = home + "/Desktop/Exported Voice Memos";
if (!fm.fileExistsAtPath(outDir)) fm.createDirectoryAtPathWithIntermediateDirectoriesAttributesError(outDir, true, null, null);

const sql = `SELECT ZPATH, ZCUSTOMLABEL, ZDATE FROM ZCLOUDRECORDING WHERE ZTRASHEDDATE IS NULL;`;
const cmd = `sqlite3 -separator "|" "${dbPath}" "${sql}"`;
const rows = app.doShellScript(cmd).split('\r');

rows.forEach(line => {
  if (!line.trim()) return;
  const [file, titleRaw, zdateRaw] = line.split('|');
  const date = new Date((parseFloat(zdateRaw) + 978307200) * 1000);
  const dateStr = date.toISOString().split('T');
  const title = (titleRaw || "Untitled").replace(/[\/\\:]/g, "-");
  let dest = `${outDir}/${dateStr} ${title}.m4a`;
  let c = 1;
  while (fm.fileExistsAtPath(dest)) dest = `${outDir}/${dateStr} ${title} ${c++}.m4a`;
  const src = `${root}/${file}`;
  if (fm.fileExistsAtPath(src)) fm.copyItemAtPathToPathError(src, dest, null);
});

Start/stop recording via shortcuts

const se = Application("System Events");
const vm = Application("Voice Memos");
vm.activate();
delay(0.2);
se.keystroke("n", {using: "command down"}); // start
// ... wait as needed ...
se.keystroke(".", {using: "command down"}); // stop

Transcript scrape (outline)

1) Activate Voice Memos, Cmd+F, type recording name, Enter. 2) Open transcript via View menu or transcript button. 3) Use a recursive UI search to find AXTextArea and read .value(). 4) Save text next to the exported audio. Example write:

const app = Application.currentApplication();
app.includeStandardAdditions = true;
const text = transcriptText; // result of AXTextArea.value()
const out = `${outDir}/${safeTitle}.txt`;
app.doShellScript(`printf %s ${app.doShellScript("python3 - <<'PY'\nimport sys,json\nprint(json.dumps(sys.stdin.read()))\nPY <<< " + JSON.stringify(text))} > ${out}`);

External trim/enhance workflow

1) Export recording (data workflow above). 2) Run ffmpeg via doShellScript for silence removal or trimming. 3) Optional: re-import by simulating drag/drop or opening in Finder, then delete the original via DB/UI if needed.

Voice Memos UI Automation (System Events)

• No dictionary: rely on System Events + keyboard shortcuts (prefer shortcuts over clicks).
• Accessibility tree (Catalyst): expect deep AXGroup/AXSplitGroup nesting; buttons may lack titles, use role/description.

Keyboard shortcuts (preferred)

• Start Recording: Cmd+N
• Stop/Done: Cmd+. or Space
• Play/Pause: Space
• Trim: Cmd+T
• Delete: Delete key (keyCode 51)
• Enhance Audio: Shift+Cmd+E
• Skip Silence: Shift+Cmd+X

Example keystroke:

const se = Application("System Events");
se.keystroke("n", {using: "command down"}); // start recording

Recursive finder (simplified pattern)

Use a stack/DFS to find elements by role/description when needed:

function findUIElement(root, criteria) {
  const stack = [root];
  while (stack.length) {
    const el = stack.pop();
    let ok = true;
    for (const key in criteria) {
      try {
        if (el[key]()!== criteria[key]) { ok = false; break; }
      } catch (e) { ok = false; break; }
    }
    if (ok) return el;
    try {
      el.uiElements().forEach(child => stack.push(child));
    } catch (e) {}
  }
  return null;
}

Transcription view (UI)

• Open the app frontmost, search via Cmd+F, Enter, then use View > Transcript (if present) or locate a button with description like "Transcript".
• Scrape text by finding a large AXTextArea after the transcript pane is shown.

UI fragility

• Sidebar hidden changes group indices; avoid hardcoded indices.
• Always ensure Accessibility permission is granted before automating.