How to turn notes into flashcards: one lecture handout, 17 generated cards, and the review pass in between
A complete run: the notes we uploaded, the 17 cards generated in 24 seconds, the ones we rewrote, and a checklist for reviewing AI cards before you study.
Write each line of your notes so it makes one claim, export the page as a PDF, generate, then spend five minutes reviewing before you study anything. That last step is the one people skip, and it is the one that decides whether the deck is worth having.
What follows is a complete run of that process on 7 September 2026, using Memor More 1.3.1 on an iPhone simulator: the notes that went in, the cards that came out, and every change we made afterwards.
The notes we started with
A single page of lecture notes on the action potential, written the way a tidy student would write them: a heading per stage, a bullet per fact, and no abbreviation left undefined.
Resting state
- Resting membrane potential: about -70 mV, with the inside of the neuron negative relative to the outside.
- The sodium-potassium pump maintains the gradient using ATP, moving 3 Na+ out for every 2 K+ in.
Triggering
- Threshold: about -55 mV. Depolarisation that reaches threshold triggers an action potential.
- All-or-none principle: the neuron either fires a full action potential or does not fire at all.
The spike
- Depolarisation: voltage-gated sodium channels open, Na+ flows in, the membrane potential peaks near +30 to +40 mV.
- Repolarisation: sodium channels close, voltage-gated potassium channels open, K+ flows out.
- Hyperpolarisation: the membrane briefly falls below the resting potential before returning to -70 mV.
The full page ran to 195 words across six headings, and was saved as a one-page PDF of about 20 KB.
One detail matters before anything else: we also tried a plain .txt file — the rougher, abbreviated version of the same lecture's notes — and the app refused it with "Unsupported file". The on-screen hint lists TXT and MD as supported formats, but the API's upload check rejects them, which we confirmed separately against the deployed code. Until that is fixed, export notes as PDF or DOCX.
What came back
From File mode, Definitions, one file attached. About 24 seconds later the app reported "17 cards ready" and suggested a deck name of its own: Neuro Lecture 4.
Cards arrive in a preview inside a deck that has not been saved yet, so nothing enters your library until you tap Create. You can regenerate the entire set from that screen if the first attempt misses.
A sample of what it produced:
| Front | Back |
|---|---|
| Threshold (action potential) | About -55 mV; depolarisation that reaches this level triggers an action potential. |
| Sodium-potassium pump | A pump that maintains the ion gradient using ATP, moving 3 Na+ out of the cell for every 2 K+ it moves in. |
| All-or-none principle | The principle that a neuron either fires a full action potential or does not fire at all, with no partial responses. |
| Saltatory conduction | Conduction in myelinated axons where the signal jumps between nodes of Ranvier, increasing conduction speed. |
| Hyperpolarisation | A brief period where the membrane potential falls below the resting potential before returning to -70 mV. |
All 17 cards, unedited, are listed on the AI flashcard generator page.
Two things are worth noticing straight away. Every card traces back to a statement on the page we uploaded; nothing was imported from the model's general knowledge of neurophysiology, including the specific ratios and voltages. And the cards come back shuffled, so the deck does not mirror the order of the document. If your notes build an argument step by step, the deck will not preserve that sequence.
The review pass
Four things needed attention. One we fixed in the app, three are judgement calls worth making before the first study session.
1. Fronts that read like headings. The card Role of Ca2+ in synaptic transmission is a topic label, not a prompt. Nothing about it tells you what you are being asked to retrieve. What does Ca2+ do at the axon terminal? is the same card doing actual work. The same applies to Axon diameter and conduction speed, which becomes How does axon diameter affect conduction speed?.
This happens because a definitions-mode generation is trying to name the concept a paragraph is about. Where your notes give a clean term, it produces a clean term. Where your notes give a topic sentence, it produces a topic.
2. A card we rewrote. We opened Nodes of Ranvier and changed the front to What are the nodes of Ranvier?, then saved. Editing is a two-tap operation on any card, before or after the deck exists, and the change persists immediately. This is the single most useful habit to build: if a front does not read like a question you would actually be asked, rewrite it while the material is fresh.
3. A near-duplicate pair. Repolarisation and Voltage-gated potassium channels describe the same event from two angles. Neither is wrong. Keeping both means answering the same thing twice a day forever, so either cut one or narrow the potassium-channel card to the mechanism alone.
4. Answers that run long. Depolarisation carries three facts: which channels open, which ion moves, and where the potential peaks. That is three retrievals hiding in one card, and it is the classic reason a card feels impossible when it is really just overloaded. Splitting it costs thirty seconds. The reasoning behind this is covered in flashcards for exam revision, which goes into the minimum information principle in more detail.
Nothing had to be deleted for being wrong. That is the useful headline: the failure mode here is not invention, it is shape. Cards come back accurate and unevenly framed.
Writing notes that generate well
Once you have seen what the generator does with a page, the way to write notes for it becomes fairly obvious.
One claim per line. A bullet that contains a definition, an example and an exception will come back as one long card. Three bullets come back as three cards you can actually answer.
Define the term in the same line you use it. "Threshold: about -55 mV" generates a good card. "Threshold — see diagram" generates nothing useful, because the diagram is not in the text.
Expand each abbreviation once. If the page says "NT" throughout and never says "neurotransmitter", you get cards about NT.
Keep the document focused. The app's own guidance is that a few dozen pages works much better than hundreds, and that matches how the server handles a long upload: it samples across the document rather than reading every paragraph. One chapter or one lecture at a time.
Do not lean on layout. Colour, indentation and the fact that something is boxed do not survive text extraction. If a point is important, say so in words.
Keep open questions out of the file. A line like "??? ask in tutorial" is an instruction to yourself, not material to memorise. Strip it before you upload, or you are inviting a card built on something you do not yet know.
Format details that will trip you up
- Plain text is refused. TXT and MD are listed in the app but rejected on upload. Export as PDF or DOCX.
- HEIC photos are refused. iPhone cameras produce HEIC by default; export or convert to JPEG before uploading a photo of a page.
- One file, 20 MB. There is no multi-file upload; combine pages into a single PDF first.
- A scanned PDF still works, because a file with no text layer is sent to the model as page images instead. Legibility does the rest of the work.
- Handwriting is a photo problem, not a format problem. A sharp, flat, well-lit page reads far better than a dim one at an angle.
What the app will not do for you
Being clear about the boundaries saves you looking for features that are not there.
There is no cloze-deletion card type. Cards have a front and a back. If you want a gap-fill prompt, write the sentence with the blank into the front yourself. Separately, the Practice feature does build fill-in-the-blank and multiple-choice exercises out of a finished deck, which covers much of the same ground in a different way.
There is no automatic reverse card. To study a pair in both directions, duplicate the card and swap the two sides by hand.
Generation runs in the full native app on iPhone, iPad and Mac, needs an account and a connection, and costs one of your three weekly free generations. Premium raises that ceiling to 200 a week. Editing and studying are always free.
The card-quality checklist
Run this once over any generated deck before the first review session.
- Does each front read as a question or a bare term, rather than a heading?
- Could two different answers both be correct for this front? If so, make the front more specific.
- Is every number, date and definition the same as in the source you uploaded?
- Does any answer contain more than one fact? Split it.
- Are there two cards asking essentially the same thing? Cut one.
- Is there a card whose answer you would accept as vague if a marker read it back to you?
- Does the deck name say what it is in three months' time, not just what the file was called?
- Have you deleted the cards you already know cold? They cost review time and teach nothing.
Eight questions, about five minutes for a 20-card deck. It is the highest-value part of the whole workflow, because the alternative is drilling a badly-shaped card daily for a month.
Then let the schedule do its work
A reviewed deck is worth much more than a bigger unreviewed one. Once the cards are in reasonable shape, the studying itself is a solved problem: retrieve, rate honestly, and let the intervals spread out. The evidence for that is genuinely strong, and summarised in what the research shows about spaced repetition and in what active recall actually is.
If you would rather not build a deck at all for a topic that thousands of people study, the free public decks are worth checking first, and Memor More for students covers how the daily routine fits around a syllabus.
Start with one lecture. Generate, review it properly, study it for a week, and you will know quite quickly whether your notes are written in a way the generator can use.
Written by
Founder & developer of Memor More. I build iOS and Mac apps and write about the science of memory and learning. @Jerelii on X
