How to transcribe a Zoom meeting

You want the words from a Zoom call, not another recording of it. There are three honest routes to a transcript, and which one fits comes down to a single question: how much do you mind your meeting audio leaving your Mac. Zoom can transcribe the call itself, but only from a cloud recording on a paid plan. You can upload a recording you already have to a cloud service, at the cost of that audio sitting on someone else's servers. Or you can transcribe it on-device, where the audio never leaves the machine. Here is what each one actually gives you, and where each one falls down.

Route 1: Zoom's own transcription

Zoom's built-in transcript is real, but it is gated. The saved audio transcript comes from a cloud recording, and cloud recording needs a paid plan (Pro and up, starting around $16 a month) with an admin switching audio transcript on before the meeting. When it is on, Zoom drops a timestamped VTT file into your recordings, usually within an hour of the call ending, in English only for now.

One catch trips people up constantly: Zoom never transcribes a local recording. If you hit Record and saved the file to your own disk instead of the cloud, there is no transcript waiting for you, by design. A local file still needs one of the other two routes below.

There is also Zoom AI Companion, which writes a meeting summary with topics and action items. It does not need a recording, but it is cloud-based, gated to paid Workplace plans, and it is a summary, not a verbatim transcript. If you need the exact words rather than a tidy recap, it is the wrong tool. And on free Zoom nothing is saved at all: you get live captions during the call, and they vanish when it ends.

Route 2: Upload the recording to a cloud transcriber

If you already have the recording as a file, the common route is to upload it to a cloud transcription service, Otter, Sonix, Rev, Descript and the rest. You pick the language, it runs, and a transcript comes back in minutes, often with speaker labels and accuracy claims up around 99 percent. It works whether or not you were the host, because all it needs is the file.

The trade is that your entire meeting audio leaves your Mac and sits on that company's servers. For a public webinar that is nothing. For a client call, a medical conversation, a legal matter, or anything under a confidentiality clause, you have just handed a full recording to a third party. Read what they retain and for how long before you upload; for regulated work, this route is often simply off the table.

Route 3: Transcribe it on your Mac, on-device

The third route runs the speech-to-text on your own machine. A local model, OpenAI's Whisper or Apple's newer Parakeet, reads the audio and writes out the text without sending anything anywhere. The audio never leaves the Mac, so there is no server, no retention policy, and no upload to think about.

It is not free of trade-offs. A local model usually wants a short proofread for proper nouns and names, where a big cloud service might edge it on raw accuracy. And you have to get the audio into it in the first place, which for a Zoom call means capturing the sound your Mac is playing. Open-source Whisper is free but takes setup; a packaged app does the capture and the transcription in one step.

The three routes side by side

Here are the routes on the axes that actually decide it.

Why speaker labels are the unreliable part

Every route promises speaker labels, and that is the part I would trust least. Working out who said which line, called diarization, is genuinely hard, and it gets harder fast when two people talk over each other. Systems that score well on clean, one-at-a-time audio degrade the moment speech overlaps.

It gets worse with the way most Mac captures work. When you record the call as system audio, everyone is mixed down into a single stream, both sides sharing one track. There is no clean per-speaker channel to pull apart, so the labels a tool prints are educated guesses, and on a fast back-and-forth they are often wrong. This is why I would rather ship an honest verbatim transcript with no labels than a labeled one you cannot trust. A wrong name against a quote is worse than no name at all.

Recording and transcribing a Zoom call on your Mac

The on-device route is the one I actually use, so I built the tool for it. Parameety is a menu-bar app that captures the call's audio through ScreenCaptureKit, the sound your Mac is already playing, optionally mixed with your mic. No bot joins the call, nothing shows up in the participant list, and it records audio, not video.

When you stop, it transcribes the audio locally with Parakeet and writes a Markdown transcript to ~/Documents/Parameety. The audio is discarded once the text is written, so only the transcript is kept, nothing lingers on disk and nothing is uploaded. It is a verbatim transcript, with no speaker labels and no AI summary, for the honest reasons above. It runs on Apple Silicon, macOS 15 or later.

Capturing the call in the first place is a separate job from transcribing it. If that is where you are, see how to record a meeting on a Mac for the full picture, and how to record a Zoom meeting on a Mac for the Zoom specifics. Record it, then let it transcribe.

FAQ

How do I get a transcript of a Zoom meeting?

There are three routes. Zoom can make one itself, but only from a cloud recording on a paid plan, and an admin has to switch audio transcript on before the meeting starts. If you already have the recording as a file, you can upload it to a cloud transcriber like Otter or Sonix, though your meeting audio then sits on their servers. Or you can transcribe it on your own Mac with a local model, so the audio never leaves the machine. Zoom never auto-transcribes a local recording, so that last route or a cloud upload is what turns a local file into text.

Can I transcribe a Zoom meeting for free?

Not through Zoom itself in a way that saves. Free Zoom shows live captions during the call, but nothing is kept afterward, and the saved audio transcript is tied to cloud recording on a paid plan (Pro and up). For free, your real options are a free cloud transcriber, which has monthly limits and uploads your audio, or a local open-source model like Whisper running on your Mac, which is free and keeps the audio on the machine but takes some setup.

Do I need to be the host to transcribe a Zoom call?

Not if you capture the audio yourself. Zoom's own transcript needs you to be the host recording to the cloud. But if you record the call's audio on your Mac, through system audio capture, and transcribe that, being the host is irrelevant. No bot joins the call and no recording notice fires; you are just transcribing the sound your own machine is already playing. Check the meeting's consent rules before you record either way.

Can I transcribe a Zoom meeting without uploading the audio to the cloud?

Yes, by transcribing on-device. Both Zoom's own transcript and cloud tools like Otter send your meeting audio to a server. A local model such as Whisper or Parakeet runs the speech-to-text on your Mac instead, so the audio never leaves the machine. That is the route I built Parameety around: it captures the call's audio, transcribes it locally to a Markdown file, and discards the audio afterward, keeping only the text.