Transcribe voice memos on your Mac.

The backlog problem

Voice memos accumulate. A walk. A driveway. The thirty seconds after a meeting ends. Each one felt urgent enough to record, and almost none get replayed, because audio can’t be skimmed. Ten memos titled "New Recording" are ten ideas you can’t search for.

Transcription is what turns that pile back into something useful. Text can be scanned in seconds, searched by the word you remember, pasted into notes, or handed to an AI to sort out.

From phone to searchable text

If iCloud is on, the memos are already on the Mac: Apple covers seeing your recordings on all your Apple devices, and everything you captured on the phone shows up in the Mac app. AirDrop handles the one-offs.

Apple can read one back to you without any of this. On macOS 15 or later with Apple silicon, the same app will show a transcription of a recording, with the caveat that audio transcription is not available in all countries or regions. For remembering what you said on a walk, that is the whole job done.

Anything beyond reading one memo means getting the file out. Share it from Voice Memos (Apple covers that too) and you have an M4A that any transcription app takes. Dropped onto Edict it comes back in seconds, on your Mac, exportable as text for your notes or as SRT or VTT if you want the timestamps to survive.

Clearing the lot

The difference between transcribing a memo and clearing a backlog is drag-and-drop versus a workflow. Edict takes files one after another without ceremony, and if you use an AI agent, its MCP server lets the agent do the boring part: submit each file for on-device transcription, collect the text, and turn three months of memos into one organized note. The audio never leaves the machine while any of that happens.

Memos with more than one voice (an interview, a conversation everyone knew was being recorded) come back with speaker labels, so the transcript reads as a dialogue, not one continuous voice.

Or skip the memo entirely

Half the memos people record were always destined to be text: a reminder, a draft paragraph, a thing to tell a colleague. If that’s the pattern, collapse the pipeline and dictate straight into the destination instead. Recording, syncing, transcribing, and pasting becomes one step: talk, and the text is already where it belongs. Edict does that with the same on-device models, a one-time purchase covering both jobs; the speech to text overview shows where everything else fits.

Common questions