The best notes from a YouTube video are not a copy of what was said. They are a short record of the ideas, written in your own words at planned stops, with timestamps, and turned into questions you can answer later without the video. This guide shows a method that does that, with a worked example from a real lecture.
The method in five steps
- Before you play, write the goal and set up a page with three parts: a narrow cue column, a wide notes column, and space for a summary at the bottom.
- While you watch, write keywords and timestamps in the notes column, not full sentences.
- At each chapter break, pause, look away, and write two or three sentences explaining the section from memory.
- After the video, turn each block of notes into a question in the cue column, then write a three-sentence summary.
- The next day and a week later, cover the notes, answer the cue questions, and rewatch only the timestamps you got wrong.
This is the Cornell note-taking system, adapted for video. The rest of this guide explains each step and why it works.
Why notes from a video are different
A video gives you two things a live lecture does not: a pause button and a transcript. Both make it tempting to capture everything. That is the trap. Copying is easy, and easy feels productive, but notes only help if producing them made you think.
Research on video lessons points the same way. In a 2020 study, 196 students watched a video lesson on how kidneys work. Those who stopped after each part to write an explanation in their own words beat those who rewatched each part, on both remembering the material and applying it to new problems. Writing from memory did more than watching again.
There is a second benefit. In a 2013 study of online lectures, students who answered questions between lecture segments took notes on far more of the slides than students who did not (24% against 7 to 9%), and their minds wandered much less. Stopping to think makes the rest of your note-taking better, too.
Step by step
Before you play: set up the page
Write the goal at the top in one line, for example "how a neural network turns pixels into a digit". Then draw a narrow column on the left for questions, a wide one on the right for notes, and leave a few lines at the bottom for a summary. On a laptop, a two-column table in any notes app works.
Open the video's chapters if it has them. Each chapter becomes a heading in your notes and a planned stop. If there are no chapters, plan a stop every 5 to 10 minutes.
While you watch: keywords and timestamps
In the notes column, write only what you would need to rebuild the idea: key terms, a number that matters, the example they used, and the timestamp. Not sentences. Not the transcript. A timestamp next to each note costs a second and saves minutes later, because it tells you exactly where to rewatch.
Pause when something is dense, but do not pause to transcribe. If you find yourself typing what the speaker just said word for word, stop and write the idea in five words instead.
At each chapter break: explain from memory
This step matters most. At each planned stop, look away from the video and write two or three sentences explaining the section as if to a friend. Then check them against what you saw and fix anything wrong.
Your explanation goes in the notes column under the chapter heading. It will usually be shorter and clearer than anything you would have copied.
After the video: questions and a summary
Go back through your notes and, for each block, write one question in the cue column that the block answers. Prefer how and why questions to what questions. "How does the network pick its final answer?" is more useful than "What is the output layer?"
Then write a summary of the whole video in three sentences at the bottom, without looking at your notes.
Review: answer the cue questions
The cue column is your quiz. Cover the notes column, read each question, and answer it out loud or on paper before checking. Do this the next day and again about a week later, because spreading reviews out beats doing them all at once. When you miss one, the timestamp next to it tells you exactly which minute of the video to rewatch.
A worked example
Here is what Cornell notes look like for 3Blue1Brown's "But what is a neural network?", written this way.
Goal: understand how a simple network recognises a handwritten digit.
| Cue questions | Notes |
|---|---|
| What does a single neuron hold? | A number between 0 and 1, its activation. Input layer: one neuron per pixel, 784 for a 28 by 28 image. Activation = pixel brightness. |
| How does the network pick its final answer? | Output layer: 10 neurons, one per digit. The brightest one wins (5:03). |
| Why have hidden layers at all? | Hope: layers build up parts. Edges, then loops and lines, then digits (6:53). Same idea as sounds, then syllables, then words. |
| How can one neuron detect an edge? | Weighted sum of the previous layer. Positive weights on the pixels you want bright, negative weights around them (9:08). |
| What does the bias do? | Shifts how high the weighted sum must be before the neuron activates. Example: only fire above 10, so add a bias of -10 (10:24). |
| How many parameters are there? | 784 × 16 = 12,544 weights into the first hidden layer alone, not counting biases (11:38). About 13,000 in total. |
| What squashes the sum into 0 to 1? | Sigmoid. Few modern networks still use it; most use ReLU: zero below a threshold, the input itself above it (17:35). |
Summary: a network is a function from 784 pixel values to 10 digit scores. Each neuron takes a weighted sum of the layer before it, adds a bias, and squashes the result. Learning means finding the weights and biases that make the right output light up.
Two things to notice. The notes are short and mostly in our own words, and every block has a question that can be answered without the video. Those seven questions are the whole review session.
Slides, diagrams, code and equations
Slides and diagrams. Take a screenshot only when the picture is the point, and write one line under it saying what it shows. A screenshot you never describe is just another thing to reread.
Code. Do not copy code from the screen into your notes. Type it into an editor and run it, then note what it does and where in the video it was explained. Rebuilding it without the video afterwards is the real test.
Equations. Write the equation, then a line for each symbol saying what it means. If you cannot say what a term does, that is your cue question.
Should you use the transcript or an AI summary?
Use them to find things, not to replace your notes. YouTube's transcript is good for checking a term you misheard or finding the minute where a topic starts. An AI summary is good for deciding whether a long video is worth your time.
Neither makes you produce anything, and producing is what makes notes work. If you start from an AI summary, treat it as a draft: close it, write the explanation yourself, and use the summary only to check what you missed.
Handwriting or typing?
A widely shared 2014 study suggested that typed notes hurt learning because students type word for word. A direct replication in 2021 found that typed notes were indeed longer and more verbatim, but quiz scores did not differ. Pick whichever you will actually review. The method above matters far more than the tool.
How long should notes from a video take?
Expect a lecture to take longer than its running time, because the stops are where the learning happens. Expect it to take noticeably longer than the running time, then a few minutes for each review. If a video takes you no extra time at all, you are probably watching, not studying.
Tools that help
We make Pausly, so weigh this with that in mind.
- Any notes app with tables or columns works for the Cornell layout. Paper works too.
- YouTube chapters and the transcript help you plan stops and check details.
- Pausly can draft study notes from a YouTube video's transcript, in the format you choose, next to the video. Use them as a starting point, then do the explain-from-memory step yourself. In Focus Mode, Pausly also pauses the video at moments from the lesson and asks you a question, which handles the stops for you.
For the bigger picture, including playback speed, spacing and common problems, read our guide on how to study from YouTube videos.
Frequently asked questions
Should I pause to take notes, or watch first and take notes after? Pause at planned stops, every chapter or every 5 to 10 minutes, and write from memory there. Watching the whole video first and writing afterwards works for short videos, but you will forget more of a long one.
Do YouTube transcripts replace note-taking? No. A transcript records what was said. Notes are useful because you have to decide what matters and put it in your own words.
What is the best note-taking method for videos? A Cornell layout with timestamps works well, because the cue column turns your notes into questions you can review. The format matters less than explaining from memory at each stop.
How long should my notes be? Short enough to review in a few minutes. For a 20-minute lecture, one page of keywords, explanations and five to ten questions is plenty.
Can AI take notes from a YouTube video for me? It can draft them from the transcript, which saves time on finding the structure. You still need to explain and test yourself, because that is what makes you remember.