If you want vocabulary cards from the words you meet while reading a PDF, box each word, keep the sentence it sits in, and send the pair to Anki. You can do that by hand for free, and it works on scanned PDFs too as long as you can type or copy the line. LeafyApp does the same thing with one shortcut when the text cannot be selected.
Two different things people mean
Searching for Anki cards from a PDF turns up two kinds of tool. The first kind takes a whole document, usually a textbook chapter or lecture slides, and uses AI to generate question and answer cards from it. That is a fine job if you are revising for an exam from notes. It is not this page.
The second job is smaller. You are reading a PDF in another language, or a dense paper in your own, and a handful of words stop you. You want those words, with the sentence each one appeared in, as cards. The PDF is just where you were reading. For that job you do not need a generator. You need a quick way to grab the word and its sentence, and a way to get them into Anki.
The free route, no install
Anki is free on the desktop, and the manual's section on editing covers everything you need. Copy the sentence from the PDF, open Anki, and click Add. Set the note type to Cloze and paste the sentence into the Text field. Drag over the word you are learning and click the […] button. Anki wraps it for you, so "Canberra was founded in 1913" becomes Canberra was founded in {{c1::1913}}. For a vocabulary card the sentence becomes She gave a {{c1::terse}} reply. and the card asks you to fill the gap.
Anki's manual also lets you add a hint after two colons, such as {{c1::terse::brief, almost rude}}, which is a good place for the definition. It takes about a minute a card, which is the problem, but for ten words a week it is perfectly reasonable and costs nothing. If you would rather not use Anki's own cloze type, a Basic card with the word on the front and the sentence on the back works too.
Why scanned PDFs break the free route
Copying a sentence only works if the PDF has a text layer. A scan is a stack of pictures, so you cannot select anything, and typing out a line in a language you are still learning is slow and full of typos. Preview can fix the file itself: in macOS 14 and later, File, Export, Embed Text writes the recognised text into a new copy. The steps, and what Live Text does and does not do, are on the page about how to OCR a PDF on a Mac.
That is the right move when you want the whole file searchable. It is a lot of ceremony when you wanted four words from page 112, and it does nothing for a PDF that is locked, or open in a viewer that is not Preview.
Keeping the sentence with one shortcut
LeafyApp works on whatever is on your screen, so the PDF can be open in Preview, Skim, a browser or anything else. Press ⌥A and drag a box around the word. macOS text recognition reads that patch of screen, so a scanned page and an image-only page behave the same as a normal one. The explanation you get is for the sense that sentence uses, and the sentence is saved with the word, along with the source and the date.
Nothing is converted and the PDF is not touched. The other half of the work, finding the sentence and typing it out, is gone.
Getting the words into Anki
LeafyApp talks to Anki through the free AnkiConnect add-on, which it can install for you under Settings, Anki, Set up. You need Anki 23.10 or later. After that there is an Add to Anki button on each word with a deck picker, and a whole folder can be given its own deck, so a PDF you are working through can be one folder and one deck. Send every word, if you switch it on, sends everything from then on without asking. You pick the card layout in Settings and see a preview before anything lands in the deck. The setup steps are in the Anki setup guide.
If you would rather not connect Anki, the same library exports to Anki, CSV, Excel, JSON or PDF. To see how the deck side works from scratch, read how to make your own Anki deck.
What this does not do
It does not read a whole PDF and write you question cards. Each card exists because you pressed the shortcut on a word, so the deck holds only what you met. It is not a PDF reader either, and it does not save the page picture. The picture is read on your Mac and thrown away. Only the word and its sentence go out so the AI can explain it, with nothing attached that says who you are, and the library stays on your Mac. The privacy page has the detail.
The Anki side needs Anki 23.10 or later. The app itself is Mac only (Apple Silicon, macOS 14 and up).
Questions people ask
How do I make Anki cards from a scanned PDF? A scan has no text to copy, so either run OCR on the file first (see OCR a PDF on a Mac) or box the word on screen with ⌥A in LeafyApp, which reads that patch of screen and saves the word with its sentence for Anki.
Can I turn a screenshot into an Anki card? Anki cards hold text, so the screenshot has to be read first. LeafyApp does that for the patch you box: the text is recognised, the picture is discarded, and the word and its sentence go to your library and then to Anki.
How do I make an Anki vocabulary deck from a book? Collect the words while you read rather than afterwards, with the sentence each one came from. In LeafyApp a book or its PDF can be one folder, and a folder can be given its own Anki deck.
Is there an OCR to Anki shortcut? Not as a single built-in Anki feature. The manual route is OCR, copy the sentence, then add it with the Cloze type. LeafyApp joins the steps: box the text, read it, save it, send it with Add to Anki.
Read the PDF. Keep the sentence.
Free while LeafyApp is in beta. One download, macOS 14 and up.
Anki is a trademark of its respective owner. LeafyApp is an independent app, not affiliated with or endorsed by Anki, and the name is used here only to describe compatibility. Refold and Preview are trademarks of their respective owners and are named only for reference.