How to Make Flashcards from a PDF (Step-by-Step Guide)

Most students have a Downloads folder full of Lecture_8_v3_FINAL.pdf files that got highlighted once and never reviewed. Highlighting feels productive. Recall on the bus doesn't happen until you turn the PDF into small, testable cards.

This is the workflow I use for engineering lectures at UofT. Total time from PDF drop to first real review: usually under 30 minutes if you scope correctly.

Card quality rules: how to use flashcards effectively. App comparison: best Anki alternatives.

Tools (honest comparison)

Tool	PDF → cards	Typical time per lecture	Ads on free?
Nebulearn	AI draft, you edit	~15–30 min total	No
Knowt	AI import	Similar + ads during review	Yes
Quizlet	Manual or limited AI	Long if typing; search if lucky	Yes
Brainscape	Manual typing	1–3 hours	No (own cards)
Anki	Manual typing	1–3 hours	No

Step 1: Scope to one chunk

Uploading a 200-page textbook gives you 400 sloppy cards and digital guilt. Scope to what the next exam actually covers:

• This week's slides only
• One chapter
• One syllabus unit (e.g. weeks 4–6)

One lecture PDF is usually 40–80 cards after editing. That's a Monday night, not a Sunday crisis.

Step 2: Generate a draft (don't study it yet)

On Nebulearn: Create → Generate → upload the file.

The AI pulls headings, definitions, repeated prof phrases. Output is a draft. Vague cards, duplicates, occasional "discuss slide 12" garbage. Normal.

Free limits: 20 AI generations/week, 300 saved questions. One lecture ≈ one generation.

Knowt / Quizlet: Similar AI on Knowt (ads during review). Quizlet weak for custom PDF unless you type manually or get lucky with a public set.

Anki / Brainscape: Read PDF, type cards. Free, no ads, costs your evening.

Step 3: Edit for 10–20 minutes (non-negotiable)

This step is where most people skip and then blame the app.

Delete:

• Slide title cards ("Chapter 4 overview")
• Figure caption cards that won't be on the exam
• Duplicates

Split:

• Any answer longer than one sentence → multiple cards
• "Explain the Krebs cycle" → one step per card

Fix:

• "Important enzyme" → name the enzyme
• "What does the prof emphasize?" → specific question with specific answer

I skim-edit once and still got burned on a biochem midterm. Now I treat first-pass editing as part of studying, not optional cleanup.

Step 4: Tag and folder

One folder per course. Tags per exam chunk: midterm-2-thermo, week-5-circuits.

Before the midterm, filter to only that tag. After the exam, archive it so it leaves your daily queue.

Full tagging rationale: flashcard guide.

Step 5: Review (this is studying)

Generating cards feels like progress. It isn't. Reviewing is studying.

First week target: 15 minutes/day on the FSRS queue. Cards you miss come back tomorrow. Easy ones wait longer. What is FSRS.

Exam week: reviews only. No new cards. Your pile is big enough.

Common PDF → card mistakes

Mistake	Fix
Whole textbook upload	One week / one unit
Study draft without editing	10–20 min edit first
Cards from unread PDF	Skim lecture once before generating
Giant combo cards	Split to one idea each
Generate once, never review	15 min/day minimum

Paid tiers (when free stops being enough)

Pay when you hit AI caps during midterms, not on day one because a banner said "unlimited."

Tool	Free limit	Paid
Nebulearn	20 AI/wk, 300 Qs	~$40–66/yr
Knowt	Ads, AI caps	Ultra ~$10+/mo
Quizlet	Ads, limited AI	Plus ~$36/yr

Checklist

• One scoped PDF section for the next test
• Draft generated
• 10–20 min edit (split, delete, fix)
• Folder + exam tag set
• First review scheduled this week

Same night the slides drop, you can have a working deck. Won't be perfect. Will be reviewable.