Turn Any 2-Hour YouTube Lecture Into 15 Minutes of Real Learning

You bookmarked that MIT 6.S191 lecture three months ago. You've started it five times. You've never finished it. It's 1 hour 48 minutes long and every time you open it you make it 20 minutes in before your attention melts.

This isn't a personal failure. It's a format mismatch. YouTube lectures are dense — the speaker packs 90 minutes of actual learning into a linear video you can't index, can't search, can't quiz yourself on. You're trying to eat a steak through a straw.

Here's what changes everything: you don't need to watch the video end-to-end. The video's transcript is the content. Once Ritsu has the transcript it can do everything the speaker was building toward — it can extract the core concepts, walk you through the hard parts, quiz you on the ideas, and schedule spaced repetition for the bits worth remembering.

This tutorial shows you the 4-step workflow. Works for MIT OpenCourseWare, 3Blue1Brown, Fireship, Andrej Karpathy, Primer, Veritasium, basically any educational video on YouTube.

Prerequisites

• A free Ritsu account.
• A YouTube URL. Any length — from 5 minutes to 3 hours. Longer videos benefit more from this workflow because the time savings compound.
• 15-30 minutes of focused time, depending on video length.

What you'll build

By the end you'll have:

1. A written summary of the video structured by concept, not by timestamp.
2. A concept-level quiz on the key ideas — not quotes, not trivia.
3. A flashcard deck on spaced repetition for the 5-15 most important takeaways.
4. The option to dive deeper on any concept via Ritsu's command system without re-watching.

You keep the ability to open YouTube and scrub to any moment if you want to re-watch a specific derivation — Ritsu doesn't replace the video, it replaces the 90 minutes of linear passive viewing.

Steps

1. Paste the URL

Open Ritsu. Click New session → YouTube → paste the URL. Ritsu fetches the transcript automatically. If the video doesn't have subtitles, Ritsu runs transcription for you. This usually takes 20-60 seconds depending on length.

A note on sources: this workflow works brilliantly for educational/technical content where the speaker is explaining something structured. It works less well for interview podcasts, reaction videos, or content that's mostly tonal/personality-driven. For the latter, you're watching for entertainment — Ritsu isn't the right tool.

Name the session after the video topic, not the channel. "Transformer attention explained" beats "Andrej Karpathy video 3" for future-you.

2. Ask for the outline

Don't scrub through the video looking for the good parts. Ritsu will tell you what they are:

/explain the structure of this video — what are the major sections and the key concept in each?

You get a clean outline of what the lecture actually covers, in the speaker's order, with the key concept of each section surfaced. For a 90-minute MIT lecture you'll typically get 4-8 sections. You now know the shape.

Here's the crucial move: read the outline. Decide which sections you actually care about. You're not obligated to learn every minute of a video just because the speaker recorded it. Educational content is a menu, not a contract.

For the sections you care about, you move to step 3. For the sections you don't, you skip them and lose nothing.

3. Drill each section with Ritsu commands

For the sections you care about, run this trio for each:

/explain [section topic]
/visualize [section topic]
/why is this important?

The first gives you a cleaner version than the lecturer's delivery. The second asks Ritsu to produce a mental model or diagram — insanely useful for concepts with geometric intuition (attention mechanisms, gradient descent, electromagnetic fields, etc.). The third grounds the concept in a reason to care.

If the video covers math or derivations you want to understand — not just the result but the steps — use:

/derive [formula or result]

Ritsu walks you through the derivation step-by-step. You can ask follow-up questions mid-derivation: "where did that substitution come from?" or "why is this term squared?" The video can't answer you back. Ritsu can.

For code-heavy videos (e.g. programming tutorials, deep-learning implementations):

/code [concept from the video]

Ritsu produces a runnable example of the concept in the language of the video. You can tweak and rerun it mentally without leaving the session.

4. Quiz + flashcard the takeaways

This is the step that separates "I watched a video and kind of understood it" from "I actually learned something." Type:

/quiz — 10 questions from the video's core concepts, mix of definitions and applications

Ritsu builds the quiz from the transcript. Take it. For everything you get wrong, type:

/eli5 why that answer is correct

Then:

/flashcard — generate cards for the 5-10 most important takeaways

Ritsu creates spaced-repetition flashcards you'll see tomorrow, in three days, in a week. This is the part that means you're still going to remember this lecture in a month — which is almost never true for videos you passively watch.

Troubleshooting

"The video is a recorded seminar and the transcript is messy." Ritsu's extraction cleans up filler and transcription artifacts automatically. If a specific passage is garbled in your summary, ask /explain [timestamp] and Ritsu will focus on that section.

"There's an important graphic in the video that's not captured in the transcript." Screenshot the graphic, drop it into the session. Ritsu's image understanding picks up diagrams, chalkboard math, slide content. Then run your commands as normal — they'll include what's in the image.

"The video is a lecture from a non-English speaker and the English subtitles are bad." Paste a link to the original-language subtitles if available. Ritsu handles multilingual content. You'll learn in your preferred language while Ritsu keeps the source intact for reference.

"I'm watching a longer series — MIT 6.S191 is 10 lectures." Don't start a fresh session for each. Create ONE course-level project, add each video URL as a new source, and use /next between them. Ritsu treats the whole series as one body of knowledge so your quizzes span topics — which is exactly how real understanding works.

"The video has code demos but no GitHub repo." Run /code with "from the video's implementation of X" and Ritsu reconstructs runnable examples matching what the speaker demoed. Not a replacement for the original repo, but workable when the author didn't publish one.

Try it yourself

Here's the test: that 2-hour video you bookmarked three weeks ago and haven't finished. Right now, copy the URL. Paste it into Ritsu. Run the four steps above. In under 30 minutes you'll have learned more from that video than you would have by grinding through it at 1.5x.

YouTube is maybe the best library of technical knowledge ever assembled. The only thing that's ever held it back as a learning tool is format. Ritsu fixes format.

Paste your YouTube link below to start. Ritsu will pull the transcript, build the outline, and walk you through the rest.