Most meeting transcription tools work the same way. Audio leaves your computer, a service somewhere else turns it into text, and the transcript comes back. That’s convenient, and for plenty of calls it’s fine. But if you talk to clients about money, health, careers or anything else they’d rather keep between the two of you, it’s fair to ask a simpler question. Does this audio need to leave my Mac at all?
On a modern Mac, often it doesn’t. Apple Silicon is fast enough to run a speech model on the machine itself, which is what people mean by local transcription. This guide explains what that actually involves, the different ways to get it, how to set it up without opening Terminal, and where the limits are.
What “local” means, and what it doesn’t
Local transcription means the speech recognition runs on your computer. The recording is captured on your Mac, the model that turns speech into text sits on your Mac, and the transcript is written to a folder on your Mac. No account on someone else’s server is needed to do the transcribing.
A few things are worth being clear about, because “local” gets used loosely.
- The model has to arrive somehow. Local apps download a speech model once, usually a few hundred megabytes. After that, transcription doesn’t need a connection.
- The app may still use the internet for other things. Checking for updates or activating a licence are common examples. Local transcription is about where your audio is processed, not a promise that the app never goes online.
- Your files are only as private as the folder they’re in. If the folder syncs to iCloud, Dropbox or a backup service, those files can still be uploaded by that service. More on this below, because it catches people out.
- Consent is still yours to handle. Recording without a bot changes the mechanics, not your obligations. Tell people you’re recording and follow whatever rules apply to your work.
Three ways to get local meeting transcription on a Mac
1. Open-source projects on GitHub
Search for local meeting transcription on Mac and much of what you’ll find is open-source code. Several of these projects are genuinely good, and if you’re comfortable with Homebrew, Python environments and granting permissions to a script, they’re worth a look. The trade-off is setup and upkeep. You install dependencies, you choose models, and when a macOS update breaks something, fixing it is on you.
2. Transcribing files you already have
If your meeting app already saves a local recording, you can run that file through a local transcription app afterwards. This works well for Zoom local recordings, voice memos and webinar downloads. It’s a two-step process, though, and it depends on the meeting app letting you record in the first place, which isn’t always the case if you’re not the host.
3. A native Mac app that records and transcribes
The third option is an ordinary Mac app that captures the call itself and transcribes it on the machine. You install it like any other app, grant two permissions, and press record. This is the route most non-developers will want, and it’s the one the rest of this guide walks through, using TapScribe, the app we make, as the example. Other native apps exist, and the same questions apply to all of them.
How recording both sides of a call works
A Mac call has two audio sources. Your voice goes in through the microphone. Everyone else comes out through your speakers or headphones. A local recorder captures both, which is why it doesn’t need to join the meeting as a bot.
In TapScribe these appear as Me (your microphone) and Them (audio playing through your Mac). If three people are on the other end of the call, all three are in Them. These labels describe where the sound came from, not verified identities, so it’s worth checking an important line against the audio when attribution matters.
Because it records what your Mac plays, this approach works with Zoom, Teams, Google Meet, WhatsApp calls in the desktop app, or anything else with sound, and nobody else needs to install anything.
Setting it up in about five minutes
You’ll need an Apple Silicon Mac running macOS 15 or later. TapScribe transcribes English only.
- Download and install. Get the installer from the Download page, open the .dmg and drag TapScribe into Applications.
- Download the speech model. First-run setup offers a one-time download of roughly 465 MB. Keep the app open and the Mac connected until it’s ready.
- Allow the microphone. This captures your side.
- Allow Mac audio. macOS lists this under Screen & System Audio Recording. TapScribe uses it for sound only and doesn’t record pictures of your screen.
- Choose your recording folder. Do this before your first real call (see the next section).
- Make a test recording. Play a short video, say a sentence, stop, and listen back. Check that both Me and Them have audio.
Headphones help. If the other side’s voices come out of your speakers, your microphone picks them up again and the transcript gets messier.
The full walkthrough, with screenshots, is in How to record and transcribe a meeting on Mac without a meeting bot.
The sync folder catch
This is the part most “local” claims skip. New TapScribe installations save recordings in ~/Documents/TapScribe. If iCloud Drive’s Desktop & Documents Folders option is turned on, anything saved in Documents can upload to iCloud automatically. Transcribing locally doesn’t switch that off.
If a conversation needs to stay off cloud storage, pick a recording folder outside anything iCloud, Dropbox, Google Drive or OneDrive watches, and check your backup settings too. The Storage and privacy page explains the two folders TapScribe uses and how to change them.
After the call
Once the recording stops, transcription runs on the Mac. A long meeting takes a little while, and you can keep working.
What you get is a transcript you can read alongside the audio. Click a line and you hear that moment, which makes it quick to check a name, a number or anything you’ll quote back to a client. You can export the transcript as Markdown, so it drops straight into Obsidian, a client folder or whatever notes system you already use. There’s a separate guide on saving meeting transcripts to Obsidian.
You can also bring in audio and video files recorded elsewhere, including WhatsApp voice notes, and transcribe those the same way.
Honest limits
Local transcription isn’t the right choice for everyone, so here’s what you give up.
- English only, for now.
- Apple Silicon and macOS 15 or later. Older Intel Macs aren’t supported.
- No AI summaries. TapScribe gives you the recording and an accurate transcript, not a generated recap or action items. If you want those, you’d paste the text into a tool of your choice, and that tool’s privacy terms then apply.
- No shared team workspace. Cloud notetakers are built for teams sharing calls. A local app is built for one person keeping their own record.
- Speaker labels are by source. Me and Them separate your voice from the call, but several people on the other end share one label.
Is local transcription right for you?
It’s a good fit if you’re an independent professional, coach, consultant or adviser who records your own client calls, works on a Mac, and would rather the audio stayed on the machine. It’s a poor fit if you need transcripts in other languages, shared team notes, or automatic summaries.
If that sounds like you, TapScribe has a 14-day trial with no card needed. Try it with a test call first, check your storage folder, and see whether the transcript is good enough for your work.