NotebookLM’s newest trick makes every tool it already had much more powerful

I've used NotebookLM for a long time, but lately I have drifted toward an open-source alternative and more or less stopped opening it. Then an update pulled me right back. It isn't a flashy new feature, but rather, it's a new supported source format: NotebookLM now lets you upload EPUB files.

That sounds small, but it really isn't. Because it makes every feature NotebookLM already had work the way it always should have. Audio Overviews, video overviews, slide decks, mind maps, reports, flashcards, quizzes, infographics, data tables — every feature in NotebookLM's Studio inherits this upgrade.

EPUB support changes the whole game

Every NotebookLM tool just got better

I used NotebookLM with my own notes, other people's notes, and research papers — but never with books. That's because, until now, the only way to get an entire book into NotebookLM was to upload it as a PDF. And a book-length PDF is a problem for two reasons.

First, it's a massive upload. Second, and more importantly, a PDF throws away the book's structure. To a PDF, the words "Chapter 12" are just text sitting on a page. The embedding model and the LLM behind NotebookLM see that line as nothing more than another scrap of body text — not as the start of a new chapter.

The same goes for everything else that gives a book its shape: the model can't tell a footnote from body text, a heading from a paragraph, or a page number printed at the top of the page from an actual sentence in the prose.

EPUB fixes this, because EPUB keeps the formatting. Now that NotebookLM can read EPUB files, it knows where each chapter actually begins and ends. It knows what's a footnote and what's the main text. It knows the number floating at the top of the page is a page number, not part of what the author wrote.

That means you can finally drag and drop a whole book straight into a notebook without carving it up into excerpts, chapters, or notes first. And because the entire book lands as a single file, it takes up far less room in your notebook.

It also kills an old trade-off. Before, I had to choose: build a general notebook, where the model understands the broad context but might wander, or build a narrow notebook, so it stays locked onto exactly what you want. With full books in one file, I no longer have to pick. I get both in the same notebook.

Books can become databases now

Data tables, meet your new library

A book is full of data, but it's buried in prose. NotebookLM's data table feature is already very good at pulling structured data out of prose and laying it into a table — and now I can point it at entire books.

I can drop a history book into a notebook and ask for a data table of every event, or every figure, it covers. Once that data is structured, I can feed it straight into my other tools (or use it as is) because it's no longer trapped in paragraphs.

The same trick works on almost any book. I use it on my coursework: I can load a textbook and have it extract the key concepts — and because it's reading EPUB now instead of PDF, I can get far more precise. I can ask it to pull the specific agents introduced in Chapter 11, and it actually works, because it finally knows where Chapter 11 is.

The timeline report feature benefits from exactly the same thing.

Make a full reading supplement

Quiz, flashcards, study guide — what else do you want?

The study guide report is where multiple books really pay off. I can now hand NotebookLM several books at once, say four textbooks I need to study, and prompt it to build a study guide for specific chapters across them. I just tell it which chapters I care about, and it builds the guide around those.

But because I uploaded the entire books, it understands those chapters in the context of the whole, so it gives me a much stronger result: it starts me on the fundamentals and works up to the harder material. And it does that across all the books at once.

The same approach works for studying, say, a philosopher's body of work. If I have four books from one philosopher, I can ask for a study guide drawn from whatever combination of that material I choose.

Quizzes and flashcards work the same way they always did, except now I can widen or narrow them however I need. In a single notebook, I can have a quiz that drills one chapter, and a quiz that tests that chapter alongside its connections to the rest of the book.

Audio Overview gets a new trick

Let two books fight it out

Audio Overview has always been one of the most interesting parts of NotebookLM for me — I've leaned on it heavily, and it was actually one of the reasons I'd switched to the open-source alternative in the first place. The EPUB update gives it a new angle, and the one I keep coming back to is debate.

I can upload two opposing takes — say, one whole book by Nietzsche and another by Camus — and prompt Audio Overview in its debate format, asking each host to argue the position of one of the books and go head-to-head. Because the model has both books in full, each side has a real case to make.

This works in text, too, though I haven't yet nailed a prompt I love for it. You can take two books — even two novels — and compare and contrast their characters, or two of their ideas, because the notebook now holds both the whole and the specifics at once.

It's still not perfect

DRM and the mobile app

This is a real step toward NotebookLM becoming a tool I can use exactly the way I want, but there are two caveats.

The first is DRM. Most ebooks we buy are locked behind it, and NotebookLM can't read a DRM-protected file. The workaround is to lean on the many sources of free, legal, DRM-free ebooks and upload those instead.

The second is the mobile experience. As great a reading companion as this makes NotebookLM, some Studio features simply aren't available in the phone and tablet apps — they're limited to the web version.

Mind maps are one example. I'm not always reading at my desk, and it would be nice to have these everywhere. What makes it stranger is that I can open the NotebookLM website on my iPad and see the full-featured version just fine — but the NotebookLM app on that same iPad won't expose those features. I don't really get why.

Regardless of the few features left out of the app, there's a lot available on the mobile app. And if not, I can always sync the outputs to my Google Drive and then to my Obsidian vault, which I can read everywhere.