Can NotebookLM take notes from a recorded lecture?

Short answer first: yes, once the lecture exists as a file. NotebookLM, which Google renamed Gemini Notebook in July 2026, takes an audio upload of up to 200 MB, transcribes it on import in any of the 71 languages Google lists, and then answers questions about it with citations, writes study guides and makes flashcards. On a computer it cannot record the lecture, and its flashcards come without a review schedule. A recorder for the phone app, announced on 15 September 2026, is still rolling out. For a recording on a laptop you need another app: Otter's free plan caps a conversation at 30 minutes, Turbo AI publishes no cap, and Notibo records for up to two hours.

Every figure below was read on the vendors' own help, pricing and privacy pages on 25 September 2026. Where a company prints no figure, the table says "not published" and the gap stays visible. We did not fill it from review sites. The passages on Google's new recorder were added on 29 September 2026.

What does a lecture recording need from a study tool?

A recorded lecture is an awkward thing to study from. It runs for an hour or more, it only exists if someone pressed record at the right moment, and it is spoken in whatever language the lecturer uses. Before you choose a tool, it helps to know which of those problems it solves and which it quietly leaves with you. These are the questions that separate the options.

Can it make the recording?

On a computer, NotebookLM starts from a file. If the tool cannot capture the audio itself, you need a second app in the room and a routine for moving the file from one to the other. Whether ChatGPT can fill that gap depends on your plan and your computer, as our check of whether ChatGPT can transcribe a lecture shows. Recording a lecturer also needs their permission, and many universities have written rules on it. Our guide to lecture recording policies covers how to ask.

Will a whole lecture fit?

Limits come in two forms. Some apps cap one conversation in minutes, others cap one upload in megabytes. By our own arithmetic, a 90-minute lecture at 128 kilobits a second, a common MP3 setting, comes to about 86 MB. The same lecture as an uncompressed CD-quality WAV works out close to 950 MB, so the format you record in matters as much as the plan you pay for.

Does it understand the language of the lecture?

A tool that handles English well can still fail a seminar in Dutch, or a lecture that switches to Arabic for a quotation. Look for a published list of languages rather than a round number in a banner. A list tells you whether your language is on it. A banner only tells you how many the vendor claims.

Will it bring the material back before the exam?

A summary helps the day after the lecture. What helps in week ten is review spread out over time, which is the idea behind spaced repetition. A tool that makes flashcards but never schedules them leaves the timing to you, and it is easy to review too late or all in one evening.

How does NotebookLM compare with apps that record?

The table puts NotebookLM next to three apps built to make the recording. Each cell is what the vendor's own page says, in the currency that page showed.

Google's price comes from its US plans page, Otter prints US dollars on its pricing page, and Turbo AI's figure comes from a post on its own blog, because the company has no pricing page. What you pay in another country can differ.

What happens when you upload a lecture to NotebookLM?

Once the file exists, the route is short. You open a notebook, choose Add sources and pick the audio file. Google's help page says the audio is transcribed at the moment of import and the text is saved as a new source, so everything you ask afterwards works on that text rather than on the sound.

• Each source can hold up to 200 MB or 500,000 words. Google's page on adding sources applies that to uploads and lists 22 audio formats, from MP3, M4A and WAV to Opus and WMA.
• 71 languages are listed for audio import. The same page names them one by one, with Danish, Swahili and Urdu among them, and warns that low-quality audio may fail to import.
• Public YouTube lectures need no download. Google accepts public videos with captions, imports only the caption text, and sets no length limit unless the captions pass 500,000 words.
• The phone app is getting a recorder. Google's page on the mobile app still lists PDF, website, YouTube, audio file and copied text as the ways to add a source. Google's post of 15 September 2026 announced an audio recorder in the mobile app for lectures, and its Workspace Updates post of 18 September says users can now use it, in a rollout that can take more than 15 days. Neither post names a language list or a length limit for the recorder.

For most phone recordings the 200 MB ceiling is generous: by the same arithmetic, three hours of compressed audio fits in it. The work sits in the step before. Someone has to make the recording, trim it if the phone kept running into the break, and get it onto the computer or into the app. On a computer, NotebookLM only starts after that.

What does NotebookLM give you back from a lecture?

With the transcript in place, a notebook behaves like a tutor who has read nothing but your material. That is the main reason to use it for revision: the answers stay inside the lecture instead of drifting into whatever the wider internet says about the topic.

• Answers with in-line citations. Google's overview of the product describes chat grounded in your sources with citations, so you can check a claim against the transcript passage it came from.
• Study guides, briefings and mind maps. The same overview lists these among the formats a notebook can turn your sources into.
• Audio Overviews. Google's Audio Overview page describes them as discussions between AI hosts that summarise your sources. The Brief format uses one speaker and runs under two minutes.
• Flashcards and quizzes without a calendar. The flashcards help page lets you choose easy, medium or hard, mark each card Got it or Missed it, retake the missed ones and download the deck as a CSV file. It describes no review schedule.

The CSV download matters more than it looks. If you want the cards to come back at the right intervals, you can move them into a flashcard app that schedules reviews. For a weekly routine built on recordings, see our guide to studying from recorded lectures.

How much can you do on NotebookLM's free plan?

NotebookLM is free with a personal Google account. Google's usage page says the way its limits are counted changed on 2 September 2026, from a fixed count of chats to a compute quota, while the numbers of notebooks and sources stay fixed per plan. Not every Google page has caught up, as the bullets show.

• 100 notebooks and 50 sources per notebook for free. Google's upgrade page raises that to 200 notebooks and 100 sources with Google AI Plus, and 500 and 300 with Google AI Pro.
• Chat and Studio use is measured in compute. Google's usage page says the quota refreshes every five hours until you reach a weekly limit, and that Plus gives twice the standard allowance and Pro four times.
• Google's pages disagree on the numbers. The upgrade page still prints the old daily counts, 50 chats and 3 Audio Overviews a day for free and 500 and 20 on Pro, and the US plans page says Pro gives 5x more Audio Overviews. Read the multipliers as approximate.
• 4.99 or 19.99 US dollars a month in the United States. Google's US plans page lists Google AI Plus and Google AI Pro at those prices and restricts the plans to people aged 18 and over.
• School accounts can use it too. Google makes the product available to Workspace for Education users of every age, once the school's administrator has switched it on.

For a single course the free plan is usually enough, because a notebook per module with one source per lecture stays far below 50. The compute quota is harder to plan around. Google says it weighs the length of your chat and the features you use, so a week of long chats and Audio Overviews will reach it sooner than a few short questions.

Does Google train its models on your lecture uploads?

A lecture recording holds other people's voices and sometimes their questions, so it is fair to ask what happens to it. Google answers for personal accounts and for school accounts separately.

• Personal accounts: no training unless you send feedback. The overview page says your data is "not used to train Gemini Notebook unless you provide feedback", and feedback lets Google review your queries, uploads and the responses.
• Feedback can be kept for three years. Google's feedback page says trained reviewers may read it and that reviewed feedback, with the content attached, is kept for up to 3 years, disconnected from your account.
• Workspace accounts are excluded. For Workspace and Workspace for Education users, the upgrade page says uploads, queries and responses are not reviewed by people and not used to train AI models.
• No storage region for personal accounts. None of the pages we read names one. Only the Google Cloud version says it honours data regionalisation.

In practice: do not press the thumbs up or down on a notebook built from a lecture, and use your university account if the school has NotebookLM switched on. Neither step replaces the lecturer's permission to record in the first place.

Is Otter the better recorder to pair with NotebookLM?

Otter is built around the step NotebookLM leaves out on a computer. Its web, desktop and phone apps transcribe while you record, and even the free Basic plan exports the audio as MP3 or the text as TXT. NotebookLM accepts both, so an Otter recording becomes a notebook source with one export and one upload.

• Basic caps each conversation at 30 minutes. Otter's pricing page gives Basic 300 transcription minutes a month and three imported files for the life of the account.
• Pro allows 90 minutes per meeting. It costs 16.99 US dollars a month, or 8.33 a month billed yearly, for 1,200 minutes. The page also showed Pro at 8.49 US dollars with a "save 50%" label, an offer its footnote limits to India-issued credit cards.
• Six transcription languages. The same page lists English, Spanish, French, German, Japanese and Chinese.
• Otter trains on de-identified data. Otter's privacy and security page says it removes identifying details from user data before training its models.

A 50-minute lecture only fits the free plan in two pieces, so Basic suits short seminars more than full lectures. If you record in English and pay for Pro, Otter and NotebookLM make a workable pair: one captures the lecture, the other lets you question a whole term of transcripts at once.

Does Turbo AI cover both the recording and the studying?

Turbo AI, called TurboLearn until its rebrand, does in one app what Otter and NotebookLM do in two. Its App Store listing says you can record a lecture and get organised notes, then quizzes, flashcards, podcasts and summaries from the same material.

• 20 US dollars a month, or 10 billed yearly. Turbo AI has no pricing page, and a Turbo AI blog post from 9 July 2026 gives those figures, which come to 120 US dollars a year.
• A free tier without numbers. Turbo AI's homepage calls it generous and names note generation, flashcards and quizzes, but gives no minutes or file counts.
• No language list and no cap per recording. We found neither on the site or in the App Store listing.
• US storage and outside AI providers. Turbo AI's privacy policy says data may be stored and processed in the United States and names OpenAI, Grok, Anthropic and Modal Labs as the providers that process content.

Turbo AI suits a student who records on a phone and wants everything in one place. The missing numbers mean you discover the limits by reaching them, so run a full lecture through the free tier before you pay.

Which should you choose?

• If your course already posts recordings or public YouTube lectures, start with NotebookLM. The file exists, the free plan holds 50 sources per notebook, and all you add is a way to schedule the flashcards it makes.
• If you record in person in English and your classes run under 30 minutes, Otter's free plan covers the capture, and its MP3 or TXT export goes straight into a notebook.
• If your lectures run 90 minutes and you want one app on your phone, Turbo AI records and makes the study material in one place, at 20 US dollars a month by its own blog, with limits you will have to test.
• If the lecture is in a language outside Otter's six, check Google's list of 71 for the upload step and choose a recorder that publishes its own list.
• If your university gives you a Google Workspace for Education account, use NotebookLM through it, since Google keeps those uploads out of training.

Whichever you choose, take one real lecture from recording to first flashcard review before you pay for a year.

Where Notibo fits

Notibo covers the recording and what comes after it, on a computer. With the lecturer's permission it records from the microphone. In Chrome or Edge it can also take the sound of a browser tab, and on Windows the whole system sound, which helps when the lecture is streamed. One recording can run for two hours; it is saved in smaller pieces as it runs and joined at the end.

From that recording you get a transcript in the language spoken, from 89 languages with automatic detection, plus structured notes with a summary, main points and action items, and flashcards that return on a spaced-repetition schedule. You can correct the notes and export them as PDF or Word. If you like NotebookLM's chat, that export can go into a notebook as a source.

Pro is free for 14 days, with 240 minutes of audio, and no card is asked for. Afterwards the free plan gives 30 minutes of audio and five documents a month. Pro costs 9.99 euro a month or 88.99 euro a year, for 1,200 minutes a month. Files are stored in the EU on Supabase in Frankfurt, and transcription and the notes run on US processors under standard contractual clauses. The audio is deleted once the notes are saved, and Notibo does not train on what you upload. The lecture recording page shows the steps.