Skip to content

Transcribe a lecture into searchable notes

Drop a lecture recording in and get a searchable, timestamped transcript — free, with no upload and no per-minute charge, which matters when you have a semester of recordings rather than one file.

The point is not to avoid attending. It is to be able to search fifty hours of audio for the ten minutes where the assessment criteria were explained.

Starting the engine…

Drop an audio or video file

or , paste with Ctrl + V, or

  • MP3
  • MP4
  • WAV
  • M4A
  • MOV
  • WEBM
  • OGG
  • FLAC
  • AAC

The file is read by this page and never uploaded — no account, no size limit but your own RAM.

Preset: balanced English model, timestamped output.

How it works

  1. Open this page — the tool is already set up for “Lectures”.
  2. Drop the file in, browse for it, or paste it with Ctrl+V.
  3. The model runs on your own device; the transcript appears as it goes.
  4. Correct anything misheard, then download text, subtitles or notes.

A transcript is an index

Audio is unsearchable, which is why lecture recordings so often go unwatched. A transcript turns a two-hour recording into something you can scan in five minutes and search in seconds, with every paragraph anchored to a timestamp you can click to hear the original. That is a different activity from reading a textbook — it is finding the bit you need.

The Markdown export is the useful one here: a headed document with timestamped paragraphs that pastes into Obsidian, Notion or OneNote and stays linkable. Build one note per lecture and the whole module becomes searchable at once.

Lecture halls are difficult rooms

A phone at the back of a large hall records mostly reverberation. Recognition degrades sharply with distance, and a lecturer who moves around, turns to the board or takes questions from the room will produce patchy results. Sitting nearer the front does more for the transcript than any model setting.

Where the institution provides a recording — a lecture-capture system with a lapel microphone — the audio is dramatically better and the transcript will be close to clean. If both exist, always use the official recording.

Technical vocabulary

Subject-specific terms are the predictable errors: a general model has no idea that your course says "eigenvalue" rather than "Ivan value". These cluster, because the same terms recur. Fix the first instance, then use the search box to find every other occurrence and correct them in a couple of minutes.

Recorded lectures in a second language are handled by the Multilingual model, which can also translate the speech into English if you are studying in a language you are still learning — imperfectly, but often enough to follow the argument.

Things that save a re-run

Read the transcript against the audio before you rely on it. Every speech model — this one and the paid cloud ones — mishears names, numbers and crosstalk, and it does so confidently. The player and editor here exist so that check takes minutes rather than an afternoon.

Frequently asked questions

Is it free for long recordings?

Yes — there is no per-minute charge and no cap, because your own computer does the work. A semester of two-hour lectures costs nothing but time.

How good is it on a phone recording from the back of a hall?

Mediocre, honestly. Distance and room reverberation remove speech detail before it reaches the file. It will still be useful as a rough index into the audio, but expect real errors. An official lecture-capture recording transcribes far better.

Can I transcribe a lecture in another language?

Yes. Switch to the Multilingual model, which detects the language automatically or lets you pick one, and can also translate the speech into English text.

Am I allowed to record lectures?

That depends on your institution and sometimes on the individual lecturer — many require permission, and some prohibit it. Check your own rules; this tool has no view on it and only works with a file you already have.