An AI note taker for students, on your own Mac

Lectures, seminars, study groups and online classes, turned into text you can search before the exam. Recorded with permission, transcribed on your own laptop, and never uploaded unless you choose.

Last checked:

A student’s week is mostly listening: lectures, seminars, lab briefings, supervision meetings, group projects, the odd online class at eight in the morning. The notes you take are only as good as what you managed to write down while also trying to understand it.

An AI note taker changes what you have to do in the room. You listen; the transcript and a first draft of the notes come afterwards. This page is about how that fits into a student’s life, where the usual tools fall short, and how Hearkeeper, a Mac app that is not released yet, is built for it.

If you only want the lecture part, the page on our lecture recorder covers the app side, and the guide on how to transcribe a lecture into notes goes step by step from recording to revision notes.

What students actually record

Not everything in a semester is a lecture, and each kind of session has its own needs.

  • Lectures. One speaker, long sessions, a lot of detail. A transcript lets you look up the exact definition instead of reconstructing it from half a sentence in your notebook.
  • Seminars and tutorials. Several voices, often a discussion. Speaker separation helps you tell the tutor’s point from a classmate’s question.
  • Study groups and group projects. Who agreed to write which section, and by when, is easy to forget. A summary with the to-dos saves an argument a week later.
  • Supervision and office hours. Feedback on a thesis chapter is worth hearing twice.
  • Online classes. The sound comes out of your computer, not into your microphone, so the app has to record the computer’s audio directly.

Hearkeeper records your microphone and your computer’s audio on the Mac, so it covers a lecture hall, a group room and a Zoom, Google Meet or Microsoft Teams class. No bot joins the call.

Permission first

This is the part a lot of apps leave out, and it matters more for students than for most people.

Many universities have rules about recording teaching, and many lecturers have their own. Some record and publish their lectures already, some are happy for students to record for personal study, and some say no, for example when a guest speaks or when students share personal views in a discussion. Your classmates’ voices are in the recording too.

  • Ask the lecturer before the first session, and say the recording is for your own study.
  • Check the course page and your university’s policy. Search its website for “lecture recording”.
  • If you need recordings as a disability accommodation, go through your university’s disability or accessibility office. They usually have a set route for it.
  • In a study group, ask the group. It takes ten seconds at the start: “Mind if I record this so we have the to-dos?”
  • Keep recordings to yourself unless everyone agreed to share them, and delete them when you no longer need them.

Hearkeeper is built to make the asking visible. Every recording starts with a consent step, where you read a short notice aloud or post it in the chat of an online class, and a consent receipt is saved with the recording. While it records, a recording pill stays on screen, shows in screen shares and cannot be hidden. There is no hidden mode. For the general rules on recording other people, see recording meetings with consent.

Works in the lecture hall, without Wi-Fi

Campus Wi-Fi tends to give up exactly when two hundred people open their laptops at once. Many AI note takers need a connection to transcribe and summarise, because that work happens on a server.

Hearkeeper does it on your Mac:

  • Live captions appear about 2–4 seconds behind the speaker, so you can glance at a word you missed.
  • A final pass after you stop, with Whisper large-v3 turbo, replaces the live text with a more accurate transcript and separates the speakers.
  • A summary, decisions and to-dos are written by a language model on your Mac, with no API key and no subscription to an AI service needed.

None of it depends on the network, so it keeps working in a basement lab, on a train or on a flight home at the end of term. Nothing is uploaded unless you connect a team server or your own API key. The page on local AI meeting notes explains what that means in practice.

Revising from a semester of transcripts

Transcripts are most useful weeks later, when the exam is close and you cannot remember which lecture covered what.

  • Full-text search finds the exact term across every recording: a formula name, an author, “the 2008 case”.
  • Semantic search finds the idea when you have forgotten the wording.
  • Ask a question across your lectures, and each part of the answer points to the second it was said. Click it and you hear the lecturer’s own explanation, not the model’s version of it.
  • Your own notes sit next to the transcript. Type during the session as usual; the transcript fills in what you did not have time to write.

Treat the summary as an outline to study from, not as the studying. Rewriting a topic in your own words is still one of the better ways to learn it, and a small local model can be wrong. That is why every line links back to the recording.

Notes export to Markdown, plain text that opens in any notes app or editor, so your revision notes are not locked into ours.

Icelandic and other languages

Transcription covers 99 languages. Icelandic is first-class: Hearkeeper is made in Reykjavík, and the speech pipeline is tuned for Icelandic. Sessions that switch between Icelandic and English, common when the slides are in English and the lecture is not, are detected automatically, with each part kept in its own language.

That also helps exchange students and anyone studying in a language that is not their first. See Icelandic transcription for more.

What you need

  • A Mac with macOS 14.2 or later, Apple silicon or Intel. On an Apple silicon Mac we measured the final pass at about ten times faster than real time; an older Intel Mac will be slower.
  • Disk space for the speech and language models, which download once. The app shows the size before it downloads anything.
  • In a large hall, a seat within clear hearing distance of the lecturer, or a small external microphone.

What it is not

  • Not a way around the lecture. A transcript of a class you did not attend is a poor substitute for being there.
  • Not for recording people who have not been told. The consent step and the pill are there on purpose.

When is it available?

Hearkeeper is coming to the Mac first, with other platforms later. The price has not been announced. It is made in Reykjavík by Drangalia ehf.

If you want lectures and study sessions you can search and question on your own laptop, join the waitlist and we will email you when there is a build for you.

Try Hearkeeper first

Private meeting and lecture notes, made on your own Mac. Join the waitlist and we’ll email you when it’s ready.

One email when it’s ready, nothing else. Leave any time. Privacy