Konthora

Knowledge Center

How to Transcribe a Lecture

A lecture recording is worth more as text than almost any other kind of audio. Students skim notes instead of rewinding at double speed, you can reuse a recording as written material, and captions make the recording usable by the people in the room who cannot rely on sound alone.

Captions and access in the room

A recording with no captions excludes the students who are deaf or hard of hearing, and it excludes everyone else whenever the room is too noisy to hear clearly on a laptop speaker. Captions are what turns a recording into an accessible one.

Caption quality matters more than caption presence. Word-level timestamps are worth choosing here, because captions that drift out of sync with the audio are arguably worse than no captions at all, because viewers trust them and stop watching the screen.

Ship both when you can. A transcript is more accessible for someone skimming or using a screen reader; captions are more accessible for someone watching the video. Neither replaces the other, and together they cost nothing extra.

Turning a lecture into study material

The best argument for a lecture transcript is that it inverts how students use the recording. Audio forces linear consumption at real time or nobody uses it. Text can be skimmed, searched, and read at 3am before an exam, and that is when a lecture is actually worth having.

Edit rather than dump. Restructure by lecture segment, fix the technical vocabulary, and pull out definitions, worked examples and anything you flagged as important at the time. A raw transcript is a document to edit; the edited version is a study resource.

Distribute the notes before the exam, not after. The value is retrieval and revision, and both are worthless if the material arrives too late to be useful.

Technical vocabulary and notation

The transcription model is general-purpose, so discipline-specific terms are where you will do the most correction. Course names, abbreviations, notation spoken aloud, and anything with a non-obvious pronunciation are the usual casualties.

A practical trick is to write the key terms down as you record the lecture. A short list of spellings to check afterwards is much more effective than reading the entire transcript hunting for errors you do not know to look for.

Worked examples deserve particular attention. A mis-transcribed number or a swapped variable turns a correct method into a wrong one, and a student following the transcript will get the wrong answer while believing they have understood it.

Lectures longer than ten minutes

Media is capped at 10 minutes and 100 MB per file, so a single lecture hour needs around six parts. Use your lecture outline to place the splits.

Overlap each join by about a second. Models occasionally drop the opening words of a file, and a one-second overlap means a sentence spanning a boundary survives in at least one part.

Label each part as you go. With six files to reassemble, an untitled set of parts is genuinely hard to put back in order a week later, and that is exactly when the notes are being written.

Step by step

1

Split the lecture by topic, not by clock

Align the split to your lecture outline — introduction, first concept, worked example, second concept. Each part then has a title you can use in the notes, and reassembly is obvious.

2

Upload the part you want to distribute

For a recorded lecture with a camera, upload the MP4 and the audio track is extracted automatically. A phone recording of the audio alone works too. No account is needed.

3

Choose sentence timestamps for study notes

Sentence mode gives each line a start and end time, which is what you want for notes students can scan and search. Switch to word mode only for captioned video where exact alignment matters.

4

Transcribe and check the technical terms

Domain vocabulary is where a general-purpose model struggles. Expect errors in course names, notation read aloud, and discipline-specific terms, and correct those specifically.

5

Export TXT for notes, SRT or VTT for captioned video

Notes are usually text. Captioned course video needs SRT or VTT. Export each part before the 60-minute retention window closes.

Open the transcription tool

Frequently Asked Questions

Can Konthora handle a full 90-minute lecture?
Not in one file. The limit is 10 minutes and 100 MB per upload, so a 90-minute lecture needs about nine parts. Split along your lecture outline and overlap the joins by a second, then transcribe each part in order.
Which timestamp mode should I choose for a lecture?
Sentence mode for study notes, because students scan and search text rather than follow timings. Word-level only when you are producing captioned video, where captions that drift out of sync are worse than no captions at all.
How do I fix errors in technical terminology?
The model is general-purpose, so expect errors in course names, notation spoken aloud and discipline terms. Keep a list of the terms that matter as you record, and check only those. Worked examples need the closest attention, since a wrong number or swapped variable produces a correct-looking but incorrect method.
Do I need both captions and a transcript?
For accessibility, ideally yes. Captions serve someone watching the video; a transcript serves someone skimming or using a screen reader. Both come from the same transcription run, so exporting both costs nothing extra.