Notes you can check.
Lectavi sits through the lecture with you and writes down what was said, with the second it was said. Every line traces back to the recording, so you can press a timestamp and hear the sentence for yourself.
Windows only for now. The app is not code signed yet, so Windows will show a SmartScreen warning on first run. Choose More info, then Run anyway. We would rather say that here than surprise you with it.
A sentence the lecturer said. It carries the second it was said.
A sentence Lectavi wrote. It never carries a time, because nobody said it.
The difference is not styling. A said line cannot be stored without a start and an end, so a claim with no moment behind it cannot pretend to be evidence.
Press Play the lecture. The notes are written as it runs, not afterwards.
A demonstration, not a recording of a real class. The sentences are written to show the two channels and the timestamps behind them.
Every other tool hands you a summary and asks you to trust it.
Here are the same two minutes of that lecture, written twice. The difference is not the quality of the prose. It is whether you can find out where any of it came from.
The lecture covered the chain rule for composite functions. The professor explained that you differentiate the outer function while holding the inner one constant, then multiply by the derivative of the inner function. An example using a trigonometric function was worked through, and the topic was noted as exam relevant.
Every sentence here might be right. There is no way to find out, and "noted as exam relevant" is either something he said or something the model decided.
This is on the midterm. I am not going to tell you which question.
Flagged as an exam signal, and you can hear him say it.
The claim about the midterm is in the first typeface, so he said it. The judgement that it matters is in the second, so Lectavi said it. Press the plate and the recording proves which is which.
Four ways in, and one of them is the phone in your pocket.
A lecture is not always a room. Lectavi takes the microphone in front of you, the sound coming out of your own machine, both at once, or a phone propped against your laptop.
In person
The laptop microphone, in the room, with the lecturer thirty feet away.
Video
The sound your machine is making. An online class, or a recording you are playing back.
Hybrid
Both at once, mixed, for the seminar where half the room is on a call.
Phone
Open a paired page on the same network and the phone becomes the microphone. Nothing to install.
An mp4 or an mp3 you already have comes in as a lecture too, and so does a YouTube lecture, with the same notes, cards and diagrams on the other side.
It reads the syllabus you were given in week one.
Upload the PDF and Lectavi pulls out the part everybody scrolls past: how the grade is actually divided, and the dates you cannot miss.
Read out of the document itself, in the order they arrive, with no year invented for a date the syllabus never gave one.
Connect Canvas or a calendar and what is due turns up beside the class it belongs to, and drops off the moment you hand it in.
We publish our error rate. Nobody else does.
Here is the measurement, with the conditions it was taken under. A benchmark without its conditions is marketing, which is exactly why this one carries them.
| Condition | Word error | Technical terms | Dates and numbers |
|---|---|---|---|
| Clean audio, no course vocabulary | 0.77% | 2.4% | 100% |
That middle column is the one that matters, and it is the one giving Lectavi your syllabus fixes. It learns the words your course actually uses, so the names and the terms come out spelled properly instead of approximately.
What we do with the audio, stated exactly.
This is the part most tools are vague about, so here it is in the two halves that actually matter: what stays with you, and who else touches it.
The recording file stays on your disk. It is never uploaded and we never keep a copy of it.
Audio is streamed to Deepgram to be transcribed while you record. That is how the words get written down, and we are not going to describe it as anything else.
Deepgram transcribes. OpenRouter writes the notes and the cards. Supabase holds your account. Stripe takes the payment.
All four are listed, with what each one receives, on the subprocessors page.
Three lectures a month, free, for as long as you want.
Enough to record a week of one class and find out whether any of this is true. Both plans get the whole app: the notes, the cards, the diagrams, the practice tests. What changes is how much you record.
- Notes, cards and diagrams
- Every timestamp, every citation
- No card, and nothing expires
- Record as much as you like
- Practice tests and the Learn path
- Canvas, the calendar and syllabus reading
- Cancel from inside the app, any time
$20 a month, billed monthly. Cancel whenever you like.
$120 once a year. Half the monthly rate, and it covers two terms and the summer between them.
Prices in US dollars. The free plan needs no card. You choose a plan inside the app after you sign in, paid plans are billed through Stripe, and you can cancel from Settings.
The awkward questions, answered straight.
Is this cheating?
Taking notes is not cheating, and neither is recording a lecture you are enrolled in. Handing in text a model wrote and calling it yours is, and no tool changes that. Lectavi is built so you can always tell which sentences came from your lecturer, which is the opposite of what a tool for cheating would do. Our position is written out in full on the academic integrity page, and it is the page to show a professor who asks.
Do I need my lecturer’s permission to record?
Often yes, and it depends on your institution and where you live. Many universities require it, some courses grant it in the syllabus, and recording law differs by state and country. Ask, and get the answer in writing. Lectavi shows an indicator whenever it is capturing, and asks you to confirm you have permission for a class before it records one.
What if my professor has a strong accent?
The number we publish is for clean audio, and a difficult room or an unfamiliar accent will be worse than 0.77%. The single thing that helps most is the syllabus: loading your course vocabulary is what takes the technical terms from a guess to a spelling. And because every line keeps its timestamp, a sentence that came out wrong is one press away from the audio that settles it.
What happens with no internet?
The recording still happens and the file is still yours, because capture is local. Transcription needs the connection, so the words arrive when you are back online.
Is there a Mac version?
Not yet. Windows only today, and we would rather say so than take your money and explain later.
Record one lecture and press one timestamp.
That is the entire argument, and it takes about fifty minutes plus one click to find out whether it holds.
Download for WindowsFree tier, three sessions a month, no card.