Can Claude listen to your voice memos?

Claude can't open an audio file. It reads text, images and PDFs, and there is no audio input at all, so if you drag an .m4a out of Apple Voice Memos into a conversation with Claude, there is nothing for it to do with it. The same applies to Claude Code. The recording has to become text before any of it is useful, and that is the part My Transcriber handles.

Once the memo is a text file the whole thing opens up, because now it is just a file sitting in a folder that Claude Code can read, search, and act on.

What Claude Code's voice mode does, and where it stops

Claude Code does have voice dictation, and it works well. You run /voice, hold space, talk, and your words land in the prompt input as you speak. For dictating a request instead of typing it, that is the right tool and I use it.

It is built for composing the message you are about to send, though, so Claude Code has to be the app you are actually looking at while you speak. Your audio is streamed to Anthropic's servers to be transcribed rather than processed on your machine, dictation covers twenty languages, and in tap mode the recording stops on its own after two minutes.

Those are all reasonable tradeoffs for dictating a prompt, and they are also why it doesn't cover the case I hit most often, which is talking for ten minutes about a problem while I page through the code, or recording a meeting I'm sitting in.

How I actually use this

I hit record, switch to another app, review something, and talk out loud about it for minutes on end. Then I paste that back into Claude Code. That is the thing a voice memo does that dictation can't, because the recording does not care which window is in front, and it does not care that I have stopped to read a diff for two minutes in the middle.

The other half of it happens away from the desk. I record a voice memo on my Apple Watch, either my own thoughts or a conversation or a meeting, and as soon as it syncs over iCloud it gets transcribed on my Mac. By the time I sit down the transcript is already there as markdown and I can paste it into a Claude Code session.

Or let Claude Code go and find the memo itself

My Transcriber ships an MCP server, so you can skip the pasting. Claude Code can query your voice memo index directly, which means I can just say "I talked about this in one of my latest voice memos, check it out" and it goes and reads them.

The app registers this for you on first launch by writing an .mcp.json into your data folder, along with a bundled skill so Claude Code knows how to drive the index without being told each time. Running claude from that folder picks it up.

Why it runs on your Mac

Whisper runs locally with Metal GPU acceleration, so the audio never leaves the machine, and an hour of recording transcribes in roughly two minutes on Apple Silicon. That matters more for voice memos than it does for a dictated prompt, because voice memos accumulate. A folder of meetings and half-formed thoughts going back a year is a different thing to hand to a cloud service than one sentence about a refactor.

It also means the language handling is Whisper's, which covers 99+ languages with automatic detection rather than the twenty that dictation supports. You choose separately, per transcript, what you actually share with Claude.

Setting it up

1. Install My Transcriber and grant Full Disk Access, which is what lets it read Voice Memos synced from iCloud.
2. Record a voice memo anywhere — Mac, iPhone, or Apple Watch. It gets picked up and transcribed to markdown in your folder, organised by date.
3. Paste the transcript into Claude Code, or point Claude Code at the folder and let it read them through MCP.

Common questions

Related: using voice memos as context for Claude, ChatGPT and Obsidian, and the full documentation.