Before you start: record well and ask permission
A transcript can only be as good as the recording. In a lecture hall, what hurts most isn't the professor's accent but the echo, the scraping chairs and a phone left at the back of the room.
- Put the phone close to the speaker, in the front rows or on the nearest desk, screen down. One metre away is noticeable; five is a lot.
- One file per lecture. If your recorder splits files, name them with the date and topic, because "Voice memo 14" will confuse you later.
- Turn on airplane mode or Do Not Disturb so a notification doesn't interrupt or pollute the audio.
- Ask the lecturer for permission and check your university's rules: what is accepted for personal study varies from one institution to another, and sharing the recording or the text with others is a different matter. The lecture has an author, and classmates who ask questions are in the audio too.
How to transcribe a lecture, step by step
- Move the recording to your computer if you can. A 90-minute lecture is a lot for a phone: a laptop with Chrome or Edge is much faster.
- Open the Transcribito tool and drop the file. Audio from your recorder (M4A, MP3, WAV) works, and so does the lecture video (MP4, MOV, WEBM).
- Leave the language on automatic detection, unless the lecture mixes languages: then choose the main one.
- Choose "More accurate" if the room has echo or the microphone was far away. It's slower, but it makes fewer mistakes with noise and technical vocabulary.
- Click "Transcribe" and keep the tab open. You can read the first parts while the rest is processed.
- Download the TXT and start working on it. If you'd rather follow the lecture with the video, download the SRT too.
How long does a lecture take?
It depends on your device. In our tests, on a Mac with Chrome and fast mode, 14 minutes of audio took about 2 minutes 40 seconds. Without WebGPU, or on a phone, the time can approach the length of the lecture or exceed it, and accurate mode is slower than fast mode.
If the lecture was online
Download the recording from the platform (Teams, Zoom, Moodle, Panopto…) and use it as it is. You don't need to extract the audio: the browser does that for you.
From transcript to notes you can study
A verbatim two-hour transcript isn't notes: it's raw material. What turns it into study material is what you do next.
Split it by topic using timestamps
Read it through once without correcting, and add a heading each time the lecturer changes topic. Timestamps let you jump back to the audio at that exact point by clicking the sentence.
Check terms, names and figures
This is what fails most. Whisper handles everyday language, but it may misspell an author, an acronym or a unit, and it doesn't transcribe formulas: what the lecturer writes on the board or shows on slides, you complete from those.
Cut what you don't need
Delete jokes, housekeeping announcements and repeated "okay, so" filler.
Be careful if you summarise with AI
Transcribing in your browser sends nothing anywhere, but if you then paste the text into an online AI assistant, that text does leave your device. Only do it if your institution's rules and the lecturer allow it.
Privacy: you're not the only voice in the recording
A recorded lecture contains the lecturer's voice and those of classmates who asked questions. With Transcribito the file is read and processed in your browser and isn't sent to any transcription service.
The only thing downloaded is the Whisper model, the program that recognises speech, from Hugging Face and jsDelivr servers: it's like installing an app, and it doesn't involve sending your audio. Those services, like any website, see your IP address when they serve you the files. The protection holds as long as the device and browser are yours: a browser extension or a shared computer can still access what you do on it.
Advantages and limits of transcribing locally
The good
- No account and no daily file limit: you can transcribe every lecture of the term.
- The lecture audio doesn't go through a third party.
- Timestamps and a player to check every sentence.
What it doesn't do
- It doesn't tell the lecturer from the students. Questions from the room end up mixed into the text.
- It doesn't transcribe live. You record first, then transcribe.
- It depends on your device. A file longer than 3 hours has to be split.
- It doesn't turn the lecture into a summary. That part is up to you.
Related guides and uses
Frequently asked questions
How long does a two-hour lecture take to transcribe?
From a few minutes, on a recent computer with WebGPU and fast mode, to about as long as the lecture or more, on a phone or in accurate mode. Keep the tab open and the device plugged in.
Is it legal to record a lecture for studying?
It depends on your university's rules, the country and what you do with the recording. Ask the lecturer for permission, don't share or publish the audio or the text without their agreement, and check your institution's policy. This page is not legal advice.
Does it tell the lecturer's voice from the students'?
No. Whisper transcribes what is said but doesn't separate speakers. If it matters who asked, mark it yourself when you review the text, using the timestamps.
Can I transcribe a lecture in Spanish or another language?
Yes. The language is detected automatically by default; if the lecture mixes languages, choose the main one. In English, Spanish, French, German and other languages the result is good with clear audio.
What if the audio has a lot of echo or noise?
Use "More accurate" mode and, if you can, sit closer to the speaker next time. No program can fix bad audio: the best improvement is recording better next time.