NotebookLM audio overviews/podcasts have been an absolute boon for my homeschooled kids. They devour audiobooks and podcasts, and they love learning by listening to these first. Then when we come together for class, we discuss what was covered, and can spend time diving into specifics or doing activities based on the content. It’s super nice to have another option for a learning medium here.
To generate them, we’ve scanned the physical book pages, and then with a simple Python script fed the images into GCP’s Document AI to extract the text en-masse, and concatenated the results together into a text-only version of the chapter. Give that text to NotebookLM and run with it.
I've used them. They're very nifty. Google did good here.
One thing I'll note is they only cover the "high level" aspects. No depth. I'd recommend them for someone who is either already very knowledgeable or for someone not at all knowledgeable who is looking for an overview before they plan to do deeper learning/studying through reading.
> or for someone not at all knowledgeable who is looking for an overview before they plan to do deeper learning/studying through reading
Yep. This is what I have used them (sparingly) for — a scaffold to build the deeper learning onto. My brain struggles to retain information when it doesn’t have a high-level understanding of how/why a system works and how individual parts connect and interact, even if it is all eventually revealed later.
Why not simply upload the pdf version of the scanned book or document? Extracting the text out of a scanned document via GCP Document AI API sounds like unnecessary use of resources
NotebookLM podcasts are like a caricature of a real podcast. Every little verbal technique or narrative style that might be used by a normal podcaster in a subtle way is taken to an extreme.
The last one I listened to one host would repeat a keyword or phrase the other host had just said for emphasis — except they did incessantly — with multiple words in every sentence for many sentences in a row.
Although I 100% agree, there is still a place for it. We place generated conversations with our case studies, and have receive good positive feedback so far, especially from the non-technical crowd. See example https://resonancy.io/case-studies/flava-process-digitization
Of course one can invest more in better authenticity but for what it is, I believe it is a good bang for effort..
Also, if you listen to it for a while, and get over the initial cringe, it becomes enjoyable, at least for me. Some visitors even asked if it was Ai generated. lol
Excited and frightened about the future where its more a real. This was a cool comparison I came across recently [2]
Interestingly I saw today the Descripts Avatars are made to sound and look non-realistic on purpose to avoid I guess all kind of issues, but they claim they want to leave something authentic on the table for real content. Which I think is a good move..
Yeah it was incredible in the beginning because it was so novel. Now it's just annoying. Half of the dialogue is repeated and it takes forever to get a point across. Never used NLM, but I wonder if that's something that can be tuned out?
> NotebookLM podcasts are like a caricature of a real podcast. Every little verbal technique or narrative style that might be used by a normal podcaster in a subtle way is taken to an extreme.
I tried it with Japanese, and it sounded about as good as in English. Only at one point did it sound unnatural. Japanese two-person conversation uses a lot of backchannelling (aizuchi), that is, semilinguistic sounds made by the listener to indicate attention and emotional reaction. At one point, the female voice said very distinctly "fumu fumu," which is how such aizuchi might be written in a script or manga. In actual speech, though, it would be a continuous sound without syllables and with a rising and/or falling intonation.
That brief TTS-like moment was the only time I was reminded that the voices were not human.
The podcast is about the impact of AI on higher education in Japan. I prompted NotebookLM briefly in Japanese about the topic, and it collected ten sources in Japanese and English that it used as the basis for the audio overview.
Do people find NotebookLM useful? For my use case of converting papers into podcasts, the explanations are too general (which misses the important parts of the paper) and contain too much fluff.
I suspect that changing the underlying model to Gemini 2.5 Pro would produce better transcripts, but right now there's no way of determining what model is being used.
I use it for loading up source materials and notes for a DnD campaign I run. Then I ask it questions when I need off the cuff answers, instead of researching.
It's also good for when I can't think of anything (like a background NPCs name and backstory)
Generate a deep technical briefing, not a light podcast overview. Focus on technical accuracy, comprehensive analysis, and extended duration, tailored for an expert listener. The listener has a technical background comparable to a research scientist on an AGI safety team at a leading AI lab. Use precise terminology found in the source materials. Aim for significant length and depth. Aspire to the comprehensiveness and duration of podcasts like 80,000 Hours, running for 2 hours or more.
I've found it useful for processing the documentation for our data system. The vendor provides the doc in something around 60 PDF files, and a lot of the information is poorly organized within the PDFs.
I can say, "Hey, NotebookLM, explain the difference between feature X and feature Y to me," or, "How do I configure Z to work the way we want?" And while the answers still kinda suck because the documentation is pretty shitty, it's way faster than digging through the PDFs. And it cites the PDFs so I can (with some trouble) find the actual documentation in the PDF if I need it.
The worst part of it is that it only accepts 50 PDFs at once.
Honestly, though, the best use for it I've seen was when my GM added the PDF rulebooks to our TTRPG to NotebookLM. We were then able to ask NotebookLM rules questions, and it would answer us pretty well. That's what it's really great for.
I don't care about the audio features at all. The first thing I do is close the audio pane.
It’s useful for getting summaries of long YouTube videos - I’m found it semi helpful for improving my Davinci Resolve skills.
That said Google is screwing the pooch as usual by trying to make it another walled garden. Slap an API on NoteboolLM already! The market research has already been done - there’s even an unofficial API https://www.reddit.com/r/notebooklm/comments/1eti9iz/api_for...
Full disclosure, I work for Google opinions are my own etc etc
The LLM built into YouTube is one of the few LLM chatbots bolted onto existing apps that I actually find useful. Not just for summaries but questions like "what is the timestamp in this 2 hour video where they talk about _____".
I thought it was for everyone my bad. Turns out except for some educational videos it's just for premium subscribers with certain location/language combos (you can probably guess which...)
I used it for a a bootcamp class to study for an exam. I recorded about 50 hours of lecture and Q&A, and was able to generate good Anki cards from it. What was awesome was that I could ask “make a list of all of the topics the instructor thought would be questions on the exam” and it did a great job at that.
I've found it very useful for providing accessible introductions to technical papers that are otherwise difficult for me to get started with understanding.
If I encounter a paper that is too difficult for me to digest just by reading, then I take a step back, feed it into NotebookLM, and listen to that summary. I've only done this a few times, but so far it hasn't failed to give me the overview and momentum that I need to take another stab and successfully dig into the paper and digest it on my own.
As others have noted, it can gloss over certain details and miss important points from time to time, but overall it does a fantastic job of giving me an introduction to a complex topic and making it far less indimidating / overwhelming.
You can enter a prompt from the "customize" dialog. Have you tried asking for a more specifics, assume the audience is an expert on the subject, and cut down on the fluff?
I've run them on my own papers and, while sometimes they are accurate, they are sometimes very very wrong and misrepresent things. And I don't mean in nuanced or unimportant ways.
The TTS is amazing, but the audio overviews are frankly useless for me.
I haven't really found it interesting for technical content but do think it's somewhat useful for hashing out more subjective and/or personal things like goals, difficulties, conflict, etc.
If you have a corpus of documents you are working with (say thousands of pages of related standards docs), Notebook can be handy for doing targeted summaries of aspects of the docs with pointers back into the actual docs to the relevant source material. That's something I end up needing a lot (I've never used the podcast feature) and so it feels very differentiated to me...
At Shopify I working as an engineer in financial services and certain changes required approval by our banking partners. I was able to upload our credit policies to NotebookLM and easily ask questions without having to ping our the legal team in Slack. I'm about as bearish as they come as far as AI tools go and NotebookLM was one of the few tools that felt useful to me straight away.
I used NotebookLM for holiday planning. I put in a dozen links with touristy things to do at the destinations and 5 odd Youtube videos. I then asked it to craft an itinerary as a travel agent who is planning holiday for a couple without kids. Included the type of things I would like to do and not do as well. The result was pretty good. The podcast generated was fun as well
Absolutely the same complaint. I wanted to see if it could summarize papers well, but I just could not handle all the conversation and attempts to make it 'exciting'. Especially in areas where I already know the background.
Fridge (after a bar code was scanned): "Ah, there we go."
Gilfoyle: "It's bad enough that it has to talk. Does it need fake vocal ticks like 'uh'."
Dinesh: "Well it just makes it sound more human."
Gilfoyle: "Humans are shit. This thing is addressing problems that don't exist. It's solutionism at its worst. We are dumbing down machines that are inherently superior."
I would like to have a Gilfyole mode for NotebookLM where the machine answers only with cold precision instead of endless "Mmmhmm", "Yeah!", "Amazing!", "That's so cool!".
Give German a try - trust me, you don't have to speak the language but anyone can tell that it's quite different in tone. No valley girls in Deutschland!
I like the NotebookLM podcasting feature, have used it a few times to come up to speed. There's one quirk of the dialogue that I find annoying though, the two speakers finish one another's sentences. At first I thought that was a nice touch, but it happens often enough that it became distracting. I should experiment with the prompt to limit how often it happens.
I really don't understand why they went with this podcast style. Sure, it makes an impression the first few times, great for a showcase or an announcement. The problem though is that it soon becomes pretty annoying, especially because the hosts go back and forth between knowing nothing and knowing everything about the topic. They should at least choose randomly which one does the explaining to whom.
Absolutely agree with you, we ran into the same issue. Our company actually tried using it for our software documentation and user onboarding, hoping it would be a helpful and engaging format. But the podcast-style delivery just didn’t fit our needs. It’s fine for a quick showcase or intro, but for ongoing support or business-oriented material, the format became distracting. If only they offered alternative styles—something more structured and professional—we might have stuck with it.
You should check new features - like asking questions as a listener.
I don't use it a lot, but it's useful when you want to have an engaging audio interface to long (50p+) reports, which you wouldn't normally read because it's not your area of expertise or you don't have time, but you can listen while doing some cardio or chores.
The best feature is by far the ability to interact with the "hosts" to ask for clarifications or to guide them into focusing on a particular aspect; even for things that weren't covered in the source material.
I created a NotebookLM podcast based on a blog post I wrote and played it for my parents. They got very excited thinking that I 'made it' because other people were talking about my work. Then I told them what it really was and they were a little bit disappointed and a little bit amazed.
I'm glad the name of my native language is written correctly. In many cases, people say "Farsi", which is offensive to many Iranians because it's the Arabic version of the word "Parsi" (unlike Persian, Arabic doesn't have "p", "g", "ch", "zh").
It's like someone calling English "Anglaise" because that's how the French say it.
PS: Contrary to common belief, Persian and Arabic are totally different languages, though they have borrowed words from one another (think English and French). Persian is an Indo-European language whereas Arabic is Aramaic (same roots as Hebrew).
> It's like someone calling English "Anglaise" because that's how the French say it.
That is the case for some other languages, though. We call the language German rather than Deutsch because Germani was the Latin name for tribes in the area, for example.
Or native names get modified too -- in English we don't call it Espanish, just Spanish, even though it comes from español.
The names of languages in other languages tend to get modified in tons of different and random ways for lots of reasons. Is there really a reason to take offense at it?
It doesn't bother me that Italians call me an americano instead of an American. It's just a letter change. So why is it so bothersome that it's called Farsi rather than Parsi? Can't the change from "p" to "f" be seen as an interesting historical quirk, due to the fascinating effect of Arabic on European languages in the Middle Ages? At the same time that we got Arabic words like "algebra" and "alcohol"?
Interesting. This is the first time I’m hearing that Farsi is offensive to Iranians. None of my Irani friends have objected so I’m curious if I’m missing something.
Wikipedia says Farsi should be avoided in Western languages, but what about others? Persian is called Farsi in Indian subcontinent due to the deep historical connections we share. We have proverbs saying Farsi is the sign of a learned person etc.
Looking at the Persian IPA table[0] for the letters you wrote, we get `/ʒ/` for `ژ` and `/tʃʰ/` for `چ`
In Arabic[1], there are two close phonemes: `/dʒ/` for `ج` and `/ʃ/` for `ش`
The difference in both phonemes is minimal and are practically affricates[2] of each other (where `d` or `t` can precede a `ʒ` or a `ʃ`), so it seems these sounds are present in both Arabic and Persian.
These variations are also within the dialectal distribution of either languages. For example `ج` is pronounced `/dʒ/` in Algeria and `/ʒ/` in Morocco.
Small nitpicking, Arabic is from a different branch of Semitic languages than Aramaic or Hebrew (which are very similar).
And TIL I learned that Aramaic replaced Hebrew in Judea because the Persian Empire maintained Aramaic as the official administrative language, and Jews brought it back, coming back from the Babylonian captivity.
You have to scroll down a couple pages' worth before you even realize this might be SO long you need to collapse it. So then you've got to scroll back UP a couple pages, find the teensy [-] link...
It's enough to just post the link to the list of languages. The list itself doesn't belong in a comment here, when it's that long.
To generate them, we’ve scanned the physical book pages, and then with a simple Python script fed the images into GCP’s Document AI to extract the text en-masse, and concatenated the results together into a text-only version of the chapter. Give that text to NotebookLM and run with it.
reply