Record a lecture.
Study it tonight.

How converting lecture audio to flashcards actually works

What happens to a 50-minute lecture recording

A 50-minute lecture is roughly 7,500 spoken words. Laxu AI uploads the audio, transcribes it with Whisper, then reads the transcript for the things worth remembering: definitions, formulas, worked examples, and the reasoning that connects them. Filler, admin and repetition are discarded. What comes back is 22 to 40 flashcards, a set of structured notes broken into chapters, and a practice quiz — usually in under two minutes for a recording of that length. You do not choose which parts matter in advance. The transcript is processed whole, so a definition given in the last five minutes is weighted the same as one given in the first five. Files up to two hours are supported, and MP3, M4A and WAV all work without conversion. Longer recordings are processed in ten-minute pieces, so a two-hour seminar does not have to finish uploading before work begins on it.

How many flashcards an hour of audio produces

Expect 22 to 40 cards per upload. The range is not arbitrary — it scales with how much distinct material the recording actually contains, not with its duration. A dense 30-minute organic chemistry lecture routinely produces more cards than a rambling 90-minute seminar, because the card count follows the number of testable facts rather than the clock. This matters when you are deciding what to record: a lecture with 12 named reactions will yield more usable cards than one hour of discussion. Every card carries a question, an answer, and where the concept was covered, so you can jump back to the moment in the recording rather than re-listening to the whole thing to check a single point. If a recording genuinely contains only twenty minutes of material, you get cards for twenty minutes of material rather than padding.

Where transcription gets things wrong — and what to do about it

Speech-to-text is accurate on ordinary prose and unreliable on the exact things a lecture is made of: proper names, chemical compounds, mathematical notation, and any term specific to your module. A lecturer saying “Le Chatelier” can come back as “Le Chaudelier”. Because flashcards are generated from the transcript, an error there propagates into the cards. The practical answer is not to distrust the whole set but to check the handful of cards containing names and formulas against your slides before you start memorising them. Recording closer to the speaker helps more than any setting — microphone distance affects accuracy far more than file format or bitrate does. A phone on the desk in front of you will beat a better microphone at the back of the room every time.

Recording live versus uploading a file afterwards

Both routes reach the same place, but they fail differently. Recording live inside the app means the audio starts uploading in pieces while the lecture is still running, so transcription is already underway when you stop — your notes appear a minute or two after the class ends rather than several minutes after you get home. Uploading a file afterwards is better when the recording came from somewhere else: a podcast, a recorded seminar, a voice memo, or audio pulled from a video. The trade-off is that a live recording depends on your phone staying with you for the hour, whereas an uploaded file can be processed on a laptop with a better connection. If the lecture is long and the room has poor signal, recording and uploading afterwards is the more reliable of the two.

Best ways to turn lecture audio into flashcards, ranked

1. Record in a study app that generates cards directly. One step, no transcript to move by hand. Laxu AI does this on web, iOS and Android; free accounts get one upload to test it, and Pro is $4.99/month for unlimited.
2. Transcribe first, then paste into a card generator. Otter.ai and similar tools transcribe and summarise well but do not build flashcards or quizzes, so you are copying text between two products every time.
3. Use a general note app with an audio feature. Quizlet added lecture recording in July 2026, but free accounts get one recording, length is capped by tier, and the recorder does not work on mobile web.
4. Type cards manually from your own notes. Slowest by a wide margin, and still the best option when the material is mathematical enough that transcription would mangle it anyway.

Common questions