Guides · Flashcards

Flashcards from photos of handwritten notes

Yes, if the photo is readable. Shoot one page per photo, flat and in daylight, then ask for one fact per card in plain wording. Check every card built from a word the tool had to guess, give each diagram its own card, and review on a widening schedule rather than all at once.

By AgoFact · · 3 min read

Can AI read my handwriting?

Usually, when the photo gives it a fair chance. A model stumbles on the same things a classmate would: faint pencil, a page shot at an angle, symbols drawn in a hurry, and abbreviations only you know.

472 photos

and images uploaded by students on AgoFact to build from, second only to PDFs

n = 62 students · to 5 October 2026

How do I take a photo a model can read?

  1. One page per photo

    Two pages in one shot halves the resolution of each. Shoot them separately.

  2. Flat and square-on

    Lay the page flat and hold the phone directly above it. A tilted page bends the lines the model has to follow.

  3. Daylight, no flash

    Flash leaves a glare spot that wipes out whatever sits under it, and a shadow across the page does the same.

  4. Fill the frame

    Crop the desk out. The page should reach the edges of the photo.

  5. One course per batch

    Keep each set of photos to one subject, so the cards stay in one deck and the model is not guessing which course a term belongs to.

What gets misread, and how do I catch it?

  • Symbols. μ read as u, a subscript read as a normal number, Σ dropped. Check every card with a formula against the photo.
  • Arrows. An arrow can mean causes, leads to or is converted into. A model may drop it or pick the wrong verb.
  • Your own shorthand. Write a short key at the top of your first page ("w/ = with, ↑ = increases") and photograph it too.
  • Crossed-out lines. Struck-through text can come back as content. Delete any card built from a line you crossed out.

The quickest check is to make the tool tell you where it guessed:

Prompt

These are photos of my handwritten notes for [course]. Make flashcards with one fact per card, using my wording where you can read it. Put [?] after any word you had to guess, and list any part of a page you could not read at all. Make a separate card for each labelled diagram.

What makes a good flashcard?

  • One fact per card. A card that asks for three things gets marked right when you remember two.
  • Fill-in-the-blank for definitions. "The ___ transports oxygen in red blood cells" is quicker to make and to review than a paragraph.
  • Split long lists. Seven causes on one card is seven cards, or one ordered list you rehearse as a sequence.
  • A picture for anything spatial. Anatomy, circuits and maps stick better as labelled images than as sentences.
  • Short wording. Cut every word that does not change the answer.

These follow Piotr Wozniak's rules for formulating knowledge, written for spaced-repetition software and still the clearest guide to writing cards.

How often should I review them?

Spread the reviews out instead of cramming them into one sitting. A 2006 review of 317 experiments on spaced study (Cepeda and colleagues) found spaced sessions beat massed ones, and that the best gap between sessions grows the longer you need to remember. In practice: review the next day, then a few days later, then after a week, and widen the gap each time you get a card right.

Ready to try it on your own course?

Make flashcards from my notes →

Questions

Should I type up my notes first?
Only the pages that are hard to read. Typing everything costs hours; a clear photo usually does the job.
Do I need cards for the diagrams?
Yes. Diagrams often hold the relationships an exam asks about, so give each labelled diagram its own card, with the image on one side.
How many cards should one lecture make?
As many as it has separate facts worth recalling, often fifteen to forty. Cut any card you could answer before the lecture.

Sources

  1. Effective learning: Twenty rules of formulating knowledge (1999). Piotr Wozniak, SuperMemo. Read 5 Oct 2026.
  2. Distributed practice in verbal recall tasks: A review and quantitative synthesis (Psychological Bulletin 132(3), 354–380, 2006). Cepeda, Pashler, Vul, Wixted and Rohrer. Read 5 Oct 2026.

How we made this

Drafted with AI from our own numbers and the sources above, then rewritten and checked by the AgoFact team. Every number on this page carries its sample size and date; we re-check them when they change.