Using YouTube and Google NotebookLM For Speed Studying
TLDR; NotebookLM can create summaries and briefing notes of YouTube videos very quickly which you can use for speed studying.
Using NotebookLM for super fast learning, studying and professional development
NotebookLM is an amazing free tool for increasing the speed of your studying and learning.
I created a video showing how I use it to summarise YouTube content:
Good Use of AI
NotebookLM is a great example of using AI to augment and not replace. A tool that is helping us improve rather than one that is replacing us and making us redundant.
Professional development and studying has always been important for me. And previously when people have asked how to improve their skill sets the answer has been simple - read and study every book you can get your hands on, and practice as much as possible.
But now we live in an age of youtube videos, podcasts, blog posts, ebooks. And all of those can be fed as sources into NotebookLM and we can then use Notebook LM to speed study by creating:
- Briefing Docs which summarise the main points in content
- Study Guides which provide outline topics and Quizes
- FAQs to provide notes in a basic Q&A format
- Timelines which can help when the content has historical overview
- Interactive mindmaps which visually summarise the content and I can navigate through it, and deep dive into what the mindmap nodes represent
NotebookLM also provides audio overviews - which I don’t use. I’ve seen some people and companies use the output from NotebookLM as a podcast but I don’t like that use-case. If you learn well from audio sources then create your own audio from NotebookLM using your own prompts and sources.
Just as a general concept. I think you can get the most out of NotebookLM if you learn to speed read. Speed reading is a fundamental skill which will help you study in general. But is really important for NotebookLM because most of the content and interaction you’ll have with NotebookLM is through text.
Studying Using NotebookLM
If I’m studying a topic then I will pull in as many sources as I can.
In the olden days that meant stacks of physical books - so I used ebay a lot to build topic libraries cheaply and then give away books that didn’t work for me.
For the last 10 years, for me, that has meant using Oreilly Safari, and reading stacks of digital books.
And now, with NotebookLM, in addition, it can mean stacks of free sources from anywhere on the internet.
I read Seth Godin’s Strategy book recently. And I also used NotebookLM to help me study it.
- I read it as an ebook and so I made notes.
- My notes were in a text file and I added those notes into NotebookLM, and converted them into a source so that I could add that into the LLM context.
- I found as many videos of Seth Godin talking about his Strategy book and added them as sources to NotebookLM
- I created briefing documents using default NotebookLM functionality from each of the video sources.
- By selecting sources I could ask questions in the chate interface. e.g. “What is a modern business plan?”
I immersed myself in the topic using the original, and multiple sources. Using both my notes and the AI generated notes.
I found this to be a very useful way of exploring the topic.
I don’t worry too much about AI hallucinations when using the process because I’ve already read the source material and I’m using AI and multiple sources to provide different views onto the material and reinforce it from different view points.
Working with Summaries and Transcripts
The out of the box functionality of NotebookLM all works fine for me.
- I don’t use the audio
- I don’t export the mindmaps
After I’ve read the Briefing Summary, and I’ve made the decision that I want more from the video, then I read the transcript, but the imported transcript is horrible to read.
At this point, I might jump off to YouTube and read the transcript there. This is how I have been running through YouTube videos for the last few years when I’m speed studying or catching up on podcasts.
To read a transcript on YouTube:
- expand the description of the video
- scroll to the bottom of the video and you’ll see a “Show Transcript” button
- this shows the transcript in a tiny window with timestamps to jump to points in the video
- sometimes this is how I actually watch a video, reading and skipping ahead to points I’m interested in
There are also tools which will extract and format YouTube transcripts to make them easier to read:
But now I can experiment with prompts in NotebookLM to achieve what I want.
I found the following prompt works well for a podcast with multiple speakers:
Transform the source raw video transcript into a well-organized, easily readable document.
Ensure the content is formatted for improved grammar while leaving all words in the text in the order they are in the source.
# Steps
1. **Read and Understand**: Begin by reading the entire transcript to understand the main topics and flow of the content.
2. **Segment Content**: Break down the content into clear, logical sections or paragraphs, using headings where appropriate to indicate shifts in topics or themes.
3. **Punctuation and Grammar**: Correct any punctuation errors to enhance readability.
4. **Add Speaker Labels**: If there are multiple speakers, label each section with the speaker's name or role.
# Output Format
Provide the final formatted transcript as a cleanly organized text document, with clear paragraphs and speaker labels (if applicable).
I documented my prompting experiments in my ai-supported-testing-experiments github repo.
https://www.eviltester.com/ebooktop6">Read our free ebook on the Software Testing books you Must Read (and Why)