Guide

How to transcribe your lectures

A recorded lecture is only useful if you can search it. Here is how to get a text version of a semester's worth without paying by the minute.

30 August 2026

Recording a lecture is easy. Finding the thing the lecturer said about the coursework, eleven weeks later, in a folder of ninety-minute audio files, is not. A transcript turns that folder into something you can search, and that is most of the value.

Here is a workflow that survives a whole semester.

First, ask #

Do this properly, because it is both the polite thing and the thing that protects you.

Most institutions have a policy on recording lectures, and many have an explicit process for it — often tied to accessibility support, and often already covering you if you have a registered condition. Some lecturers are entirely relaxed about it. Some are not, particularly in seminars where students speak, or when the material is unpublished research.

Ask once, at the start of term, in an email. You will usually get a yes, and now you have it in writing. The situations that go badly are the ones where nobody asked.

Two rules that apply regardless of the answer: do not publish or share the recordings, and treat anything a fellow student says in a seminar as theirs, not yours.

Recording so the transcript is usable #

The single biggest factor in transcript quality is not the app. It is where the microphone was.

• Sit near the front if you can. A phone in the third row pointed at a lecturer with a lapel mic on a PA system does fine. A phone at the back of a raked lecture theatre records the room, not the speaker.
• Point the phone at the lecturer, propped against your bag or a book rather than flat on the desk.
• If the lecture is streamed or recorded by the institution, use that file instead. It is usually a direct feed from the room's microphone, which is dramatically cleaner than anything you can capture from a seat.
• Stop the recording when the lecture ends. Twenty minutes of chairs scraping and people packing up is the main cause of transcripts that dissolve into repeated text.

Voice Memos on iPhone is fine for this. So is any recorder app. The format does not matter much.

Transcribing without a subscription #

This is where the economics matter for a student.

Cloud transcription is sold per minute of audio. Four lectures a week, ninety minutes each, over a twelve-week term is around seventy-two hours of audio. At typical per-minute rates that is a serious number — considerably more than most students want to spend on note-taking.

Local transcription has no meter. The model runs on your own laptop or phone, so the hundredth hour costs exactly what the first one did, which is nothing. On an Apple Silicon Mac a ninety-minute lecture takes a few minutes to transcribe.

With Offline Transcription, drop the recording in, set the language if the lecture is not in English, and let it run. There is no upload, which also means it works in a building with no signal — and campus Wi-Fi in a basement lecture theatre is exactly the case where an upload-based tool fails.

Turning the transcript into notes #

A raw transcript is not notes. It is a searchable index of the lecture, which is a different and still very useful thing.

Search it first, read it never. You are not going to read ninety minutes of text. You are going to search it for "assessment", "exam", "due", a technical term, or a name, and find the thirty seconds that matter.

Export with timestamps for anything you might need to hear again. Text alone loses the ability to go back to the audio. If your tool has a segment view, use it: find the line, play that line, confirm what was actually said. This matters most for the things you would otherwise get wrong.

Check these before you rely on them: names of researchers and papers, technical terms specific to the field, numbers and dates, and anything the lecturer said about assessment. Automatic transcription is confident and sometimes wrong, and unfamiliar proper nouns are where it is wrong most often.

Keep the audio until the module ends. The transcript is derived; the recording is the source.

A file-naming habit worth having #

Ten weeks in, New Recording 34 is worthless. Sixty seconds of setup at the start of term saves the whole thing:

2026-09-14 PHIL201 Lecture 3 — Kant.m4a

Date first so files sort chronologically, module code so you can filter, topic so you can find it. Transcribe with the same name and the transcript sits next to its audio.

The short version #

Ask permission at the start of term. Sit close, point the phone at the lecturer, stop recording when it ends. Transcribe locally so seventy hours of audio a term costs nothing. Then treat the transcript as a search index rather than a document, check names and anything about assessment against the audio, and name your files so you can still find things in week eleven.