Skip to content

How to turn a PDF or slide deck into notes and flashcards

By · Notibo

Published · 8 min read · updated

Key takeaways

  • Read the deck once for structure, then write one question per idea. The questions are your notes, and the cards come straight from them.
  • Highlighting and rereading were rated low utility in the 2013 Dunlosky review. Writing questions and answering them from memory was rated high.
  • A scanned PDF with no text layer gives an AI tool nothing to read. Try to select a sentence; if you cannot, run OCR first or get the original slides.
  • Upload one lecture's slides at a time. A 300 page reader hits page and file size limits and produces vague cards; a chapter produces specific ones.
  • Review the cards on day 1, 3, 7, 14 and 30. Cards you never review are a document, not a study method.

Short answer first: read the file once for structure, write one question per idea, answer the questions from memory, and keep the questions and answers as your cards. That is the whole method, whether you do it by hand in an evening or let a tool draft the first pass. By hand it takes about an hour for a 40 slide deck. With an AI tool it takes a few minutes, but only if the PDF contains real text, and you still need ten minutes to edit what comes out.

Why not just read the PDF?

Because reading is the study activity with the worst return on time. It feels productive, the slides become familiar, and the familiarity gets mistaken for knowledge. In the 2013 review by Dunlosky and colleagues, rereading, highlighting and summarising were all rated low utility, while testing yourself and spreading study over time were rated high. The UNC Learning Center puts it more bluntly in its Studying 101 guide: rereading leads to quick forgetting, and quizzing yourself is what works.

A slide deck makes this worse than a textbook does. Slides are prompts for a speaker, not explanations. Twenty words and a diagram made sense with the lecturer talking over them; on their own, three weeks later, they are a list of headings. Turning them into questions forces you to reconstruct the explanation, and that reconstruction is the learning.

How do you turn a PDF into notes by hand?

The manual method has four steps and takes about an hour for one lecture's slides. It is worth doing by hand at least a few times, because it teaches you what a good card looks like before you let a tool make them.

Step 1: Read for structure, not content. Go through the whole deck in five minutes without trying to learn anything. Note the section titles and where the lecturer spends the most slides. A topic that gets twelve slides is a topic that gets exam questions.

Step 2: Extract one line per idea. Go back through, and for each slide that carries an idea write one line in your own words. Skip title slides, agenda slides, and any slide that repeats an earlier diagram. For a 40 slide deck you should end up with 15 to 25 lines. This is the point where a reading becomes notes; effective note taking covers the same move for lectures.

Step 3: Turn each line into a question. "Mitochondria produce ATP through oxidative phosphorylation" becomes "How do mitochondria produce ATP?" The line you wrote is the answer. Where a slide shows a process or a comparison, write the question so it asks for the process ("What are the steps of X?") rather than a single fact. The Cornell method does this with a cue column beside the notes, which is the same idea on paper.

Step 4: Answer from memory, then check. Cover the answers and go through the questions once. Mark the ones you missed. That first pass is not revision, it is the check that your questions are answerable and your answers are complete. It takes ten minutes and it is the step most people skip.

What you have at the end is a set of question and answer pairs. That is a deck of flashcards, whatever you write it in. How to make flashcards from your notes goes into what separates a good card from a bad one.

How do AI tools turn a PDF into notes and flashcards?

They do the same four steps, badly at step one and well at steps two and three. You upload the PDF, PowerPoint or Word file, the tool extracts the text layer, sends it to a language model with instructions to write structured notes and then question cards, and returns both within a minute or two.

What the tool is good at

Extraction and phrasing. It will not miss a slide, it does not get tired on slide 35, and it writes clean questions. For a 40 slide deck you typically get a page of notes with headings and 20 to 40 cards. The time saving is real: an hour becomes ten minutes.

What you still have to do

Three things. First, delete the cards that ask for trivia: the lecturer's name, the date on the title slide, the definition that appeared on one slide and never again. Second, check the cards against what the exam actually asks for. If your exams ask you to compare and explain, cards that ask for a single term are not enough, and you need to rewrite a few into "why" and "how" questions. Third, split anything that has two answers into two cards. That editing pass is where you judge what matters, which is the part of the manual method worth keeping.

Manual or generated?

Aspect By hand With an AI tool
Time for a 40 slide deck About an hour Ten to fifteen minutes including editing
Cards produced 15 to 25, all yours 20 to 40, some to delete
Learning during creation High, the writing is a study session Low, the editing is the study
Needs a text layer No Yes
Best for The hardest two or three topics The bulk of a course

What goes wrong with scanned PDFs?

The most common failure is a PDF with no text layer. A scanned reader, a photographed handout, or slides exported as images look like text to you and like pictures to the tool. It has nothing to extract, and you get empty notes, a handful of cards built from the file name, or an error.

How to tell

Open the PDF and try to select a sentence with the mouse. If the whole page highlights as one block, or nothing highlights, there is no text layer. Adobe's own OCR page uses the same test: if you cannot select or highlight the text, you are looking at a scan.

How to fix it

Run OCR before you upload. Acrobat, most PDF readers and several free tools will add a text layer to a scanned PDF. The recognised text will have errors in equations, tables and handwriting, so check the cards from those pages with extra care. Better still, ask for the source file. A lecturer who exported slides as images almost always still has the PowerPoint, and a PowerPoint file needs no OCR at all.

What are the page and file limits?

Every tool caps uploads, usually by page count, by file size, or both. The caps are listed in each tool's help pages and differ between free and paid plans, so check before you upload a whole reader.

Limits aside, one lecture at a time gives better cards than a whole course in one file. A model asked to summarise 300 pages writes general cards; asked to summarise 40 slides it writes specific ones. Split a course reader by chapter, and name the files so the cards can be traced back to the source when one of them turns out to be wrong.

How should you review the cards?

Cards you never review are a document. The schedule that makes them a study method is the one from spaced repetition: a first review the next day, then after three days, a week, two weeks and a month.

Review When What to do
1 The next day Answer every card from memory. Missed cards get a second attempt in the same session.
2 Day 3 All cards again. A card missed twice gets rewritten; the question is usually the problem.
3 Day 7 All cards, shuffled with the cards from other lectures that week.
4 Day 14 Only the cards you have not got right twice in a row, plus a quick pass of the rest.
5 Day 30 Everything once. Cards you get right here are safe until the exam.

Two rules keep the schedule honest. Answer before you flip, every time, because looking at the answer first turns the card back into rereading. And send a missed card back to the start rather than nudging it along. The evidence for this kind of retrieval practice is strong: in Karpicke and Blunt's 2011 study, students who answered questions from memory after reading a text outperformed students who spent the same time building concept maps, including on questions that asked them to draw inferences. Active recall explains why the effort of retrieving is what does the work.

Where does Notibo fit?

Notibo is one of the tools that does the generated version: you upload a PDF, PowerPoint or Word file, and it returns structured notes and a flashcard deck that it schedules for review with spaced repetition. The same editing rules apply. Delete the trivia, rewrite the single term cards into questions that ask why, and answer before you flip.

Frequently asked questions

Can I turn a PDF into flashcards automatically?
Yes. Tools such as Notibo, Turbo AI and Coconote read the text layer of a PDF, PowerPoint or Word file and write notes and question cards from it. The result depends on the file containing real text rather than scanned images, and on you deleting or rewriting the cards that do not match what your exam will ask.
Why does my PDF upload give empty or garbled notes?
Almost always because the PDF is a scan with no text layer. The tool sees pictures of pages, not words. Open the file and try to select a sentence. If you cannot, run OCR in Acrobat or a similar tool first, or ask the lecturer for the original PowerPoint file.
How many flashcards should one lecture's slides produce?
Ten to thirty for a 40 slide deck. One card per idea, not per slide. Title slides, agenda slides and repeated diagrams give you nothing. If a tool hands you 80 cards for one lecture, delete the ones that ask for trivia and keep the ones that ask why or how.
Is it better to make the cards myself or let a tool do it?
Making them yourself is a study session in its own right, so do it by hand for the two or three hardest topics of a course. For the rest, generate the cards and spend the saved hour answering them. Editing a generated card still forces you to judge what matters, and takes a quarter of the time.
How do I review flashcards made from a PDF?
Like any other cards. Answer from memory before you flip, then review again after one day, three days, a week, two weeks and a month. A card you miss goes back to the one day interval.
Why not just read the PDF?
Because reading is the study activity with the worst return on time. It feels productive, the slides become familiar, and the familiarity gets mistaken for knowledge. In the 2013 review by Dunlosky and colleagues, rereading, highlighting and summarising were all rated low utility, while testing yourself and spreading study over time were rated high. The UNC Learning Center puts it more bluntly in its Studying 101 guide: rereading leads to quick forgetting, and quizzing yourself is what works. A slide deck makes this worse than a textbook does. Slides are prompts for a speaker, not explanations.
How do you turn a PDF into notes by hand?
The manual method has four steps and takes about an hour for one lecture's slides. It is worth doing by hand at least a few times, because it teaches you what a good card looks like before you let a tool make them. Step 1: Read for structure, not content. Go through the whole deck in five minutes without trying to learn anything. Note the section titles and where the lecturer spends the most slides. A topic that gets twelve slides is a topic that gets exam questions. Step 2: Extract one line per idea. Go back through, and for each slide that carries an idea write one line in your own words. Skip title slides, agenda slides, and any slide that repeats an earlier diagram.
How do AI tools turn a PDF into notes and flashcards?
They do the same four steps, badly at step one and well at steps two and three. You upload the PDF, PowerPoint or Word file, the tool extracts the text layer, sends it to a language model with instructions to write structured notes and then question cards, and returns both within a minute or two.

Sources

  1. Improving Students' Learning With Effective Learning Techniques (Dunlosky et al., Psychological Science in the Public Interest, 2013) (highlighting, rereading and summarising rated low utility; practice testing and spacing rated high)
  2. Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping (Karpicke and Blunt, Science, 2011) (answering questions from memory beat building concept maps from a text, including on inference questions)
  3. Studying 101: Study Smarter Not Harder (UNC Learning Center) (rereading leads to quick forgetting; quiz yourself and explain the material in your own words)
  4. OCR PDF: convert scanned PDFs to searchable text (Adobe Acrobat) (a PDF whose text cannot be selected is a scan, and OCR is what adds the text layer)

About the author

Farhad Ba-Ali

Farhad Ba-Ali builds Notibo in Odense, Denmark. He writes about study methods, memory and note-taking from the questions students send in, and from what the product's own data shows.

Ready to take better notes?

Let Notibo listen along and write your notes and summaries. Free to get started.

Create free account

Back to the blog