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.
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.
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 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.
I find the podcast style audio it produces super annoying.
Is there an easy way to simply have text read to me unaltered?
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
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 like to feed Hacker News comments to generate a podcast.
It's good to get the big picture about the discussion with 300+ comments.
The best thing to feed the podcast is a dump of all one-liner macros we’ve added to an IRC bot over 15 years (for fetching weather, stocks, and 99% stupid jokes) without any context. Cannot stop laughing listening to it trying to figure it out and bringing up the weirdest ones.
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 uploaded a Python script I wrote last week (system backup script, changed extension to .py.txt) and the Podcast it created was pretty suitable to give none-tech people who might be asking what you've been doing an idea about it.
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.
Generated a Bangla (Bengali) podcast from a complicated property lease and price trend analysis document. I'm floored. Impossible to tell that the pod casters not real. I'm sure over a long term it will sound monotonic and disengaging. But what we have here is simply a breakthrough.
Tangential: Anyone knows a free/cheap service that can turn English text articles into an audio file narrating them? Can NotebookLM do this? I don't want to turn them into podcasts or conversations.
His Chinese voice effect is not as good as Minimax.
You can use Hacker Podcadt to compare
I kind of want the opposite of NotebookLM: take verbose conversational information and distill it down to concentrated content
They don't have an app? strange.
https://support.google.com/notebooklm/answer/15731776
- Afrikaans
- Albanian
- Arabic
- Armenian
- Azerbaijani
- Basque
- Bengali
- Bulgarian
- Burmese (Myanmar)
- Catalan
- Cebuano
- Chinese (Simplified)
- Chinese (Traditional)
- Croatian
- Czech
- Danish
- Dutch
- English
- Estonian
- Filipino
- Finnish
- French (Canada)
- French (European)
- Galician
- Georgian
- German
- Greek
- Gujarati
- Haitian Creole
- Hebrew
- Hindi
- Hungarian
- Icelandic
- Indonesian
- Italian
- Japanese
- Javanese
- Kannada
- Konkani
- Korean
- Latin
- Latvian
- Lithuanian
- Macedonian
- Maithili
- Malay
- Malayalam
- Marathi
- Nepali
- Norwegian (Bokmål)
- Norwegian (Nynorsk)
- Oriya
- Pashto
- Persian
- Polish
- Portuguese (Brazil)
- Portuguese (Portugal)
- Punjabi
- Romanian
- Russian
- Serbian (Cyrillic)
- Sindhi
- Sinhala
- Slovak
- Slovenian
- Spanish (European)
- Spanish (Latin America)
- Spanish (Mexico)
- Swahili
- Swedish
- Tamil
- Telugu
- Thai
- Turkish
- Ukrainian
- Urdu
- Vietnamese[flagged]
[flagged]
[flagged]
[dead]
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.