I Turned a 3-Hour YouTube Course Into a 2-Minute Read Using Gemini Notebook — Here's the Exact Process

I had a YouTube course sitting in my "Watch Later" list for two months. Three hours long, in one sitting, no chapters worth trusting. Instead of watching it, I fed the link into Gemini Notebook and had a clean, organized written summary in about four minutes. Here's exactly how, and where it quietly falls apart.

Section 1: What Gemini Notebook Actually Does With a YouTube Link

Quick naming note first, because it trips people up: if you've used NotebookLM before, you already know this tool. In July 2026, Google folded NotebookLM into its Gemini brand and renamed it Gemini Notebook. Same product, same notebooks, same sources — new name inside the Gemini app ecosystem. If you're coming from our earlier post on using Gemini Audio Overviews for on-the-go learning, this is the same underlying tool, just a different output format.

Here's the part most people get wrong about "converting a YouTube video into text": Gemini Notebook doesn't watch the video the way you would. When you add a YouTube URL as a source, it imports only the video's text transcript — either the creator's own captions or YouTube's auto-generated ones. It does not process the visuals, the audio tone, on-screen graphics, or anything not spoken aloud. If a video is mostly silent screen-recording with text on slides, this tool has almost nothing to work with.

That matters because it sets your expectations correctly. This is genuinely excellent for talking-head videos, lectures, podcasts-on-YouTube, tutorials with verbal explanation, and interviews. It's close to useless for silent screen recordings, music videos, or heavily visual content where the value is in what's shown, not said.

A few technical limits worth knowing before you start:

• The video must be public. Private or unlisted videos won't import.
• It needs captions. Either manually uploaded by the creator or YouTube's auto-generated captions — no captions at all means no source.
• Very recent uploads may not work yet. Videos posted less than 72 hours ago sometimes aren't available to import because captions haven't fully processed.
• There's effectively no length limit. A video can be any duration unless its caption file runs past roughly 500,000 words, which is an extreme edge case.
• If the source video disappears, so does your summary — eventually. If a video is deleted or made private after you've imported it, Gemini Notebook removes it from your notebook within 30 days.

Knowing these limits before you start saves you from importing a video, getting a confusing error, and assuming the tool is broken when it's actually just a caption problem.

Section 2: My Real Test — 3-Hour Course to a 4-Minute Import

I picked the video I'd been avoiding: a long-form tutorial-style course, just over three hours, single continuous recording, verbal explanation with some on-screen text I fully expected to lose.

What I did

1. Opened Gemini Notebook and created a new notebook titled after the course topic.
2. Selected "Add," then pasted the YouTube URL directly into the source field.
3. Gemini Notebook pulled in the transcript as a single source, correctly labeled with the video's title.
4. Opened the auto-generated Source Guide summary first, before asking anything myself.
5. Then asked follow-up questions directly in chat for the sections I actually needed depth on.

What worked well

The auto-summary was genuinely usable, not just a gist. The Source Guide gave me a structured breakdown of the video's main sections in roughly the order they were covered, which for a 3-hour unchaptered video was already worth the four minutes it took to generate.

Follow-up questions were where the real value showed up. Asking something specific — "what did the instructor say about setting up the second configuration step" — returned an accurate answer pulled directly from the transcript, with a citation I could click to jump to that point in the reasoning. That's meaningfully better than skimming a raw auto-caption dump yourself, since raw YouTube captions have no punctuation logic and are painful to scan.

It handled a full 3-hour transcript without choking. No length warnings, no truncation. Given there's effectively no length ceiling short of an extreme word count, this held up exactly as expected.

Where it fell short

Anything shown on screen but not said out loud was invisible to it. At one point the instructor referenced "the settings shown here" while pointing at something on screen — the summary had no idea what "here" was. If a video leans on visual demonstration more than narration, you will lose real information this way.

Auto-generated captions occasionally introduced small errors. A couple of technical terms got transcribed as the nearest-sounding everyday word, which the summary then repeated confidently. Nothing that broke the overall summary, but worth a skeptical eye on anything technical or jargon-heavy.

It's a summary, not a transcript you can quote from with confidence. For getting the gist and deciding whether the full video is worth your time, this is excellent. For pulling an exact quote to cite somewhere, go back to the source video directly.

Net result: a video I was never going to watch in full became something I actually understood and could act on, in about four minutes of setup plus a few minutes of follow-up questions. That's the honest win here — not that it replaces watching, but that it replaces not watching at all.

Section 3: Step-by-Step — How to Do This Yourself

Basic setup

1. Open Gemini Notebook at your notebook workspace, or find "Notebooks" inside the Gemini app. Sign in if needed.
2. Create a new notebook, or open an existing one if this video relates to material you're already collecting.
3. Select "Add" to add a source, and paste the full YouTube video URL directly into the source field.
4. Wait for import. Gemini Notebook pulls in the video's caption track as a text source — this is usually quick, but can fail if the video has no captions or was uploaded very recently.
5. Open the Source Guide in the source viewer to see the auto-generated summary of the entire video before asking anything.
6. Ask specific follow-up questions in chat for anything the auto-summary didn't cover in enough depth — mention the source by name if you have multiple sources in the notebook, so Gemini Notebook knows where to look.

Getting better results

• Favor talking-heavy content. Interviews, lectures, podcasts-on-YouTube, and narrated tutorials are the sweet spot. Save silent screen recordings for actually watching.
• Ask narrow questions, not broad ones. "Summarize this" gives you the gist. "What specific settings did they recommend for X" gives you something you can act on.
• Cross-check anything technical. If a term matters — a config value, a tool name, a statistic — verify it against the actual video before repeating it elsewhere, since auto-captions can misheard technical vocabulary.
• Combine it with other outputs if you want more than a summary. The same imported source can also generate a study guide, FAQ, timeline, or an Audio Overview — useful if you want to listen to the summary later the way we covered in our Audio Overviews for passive learning post.

Who this is genuinely useful for

This earns its place if you regularly have long videos — courses, conference talks, interviews, tutorials — that you'd rather triage than sit through blind. It's especially strong for deciding whether a long video is worth your full attention before committing to it. It's a weak fit if the videos you care about are visual-first, since it only ever sees what was said, never what was shown.

If you're building out a broader research or content workflow around AI tools like this one, it's worth reading our comparison of AI tools for content research and summarization to see how this stacks up against dedicated video-summary tools for different use cases.

What's the longest video you've been avoiding watching? Try this on it and tell me in the comments whether the summary actually held up.