How NotebookLM is improving🐬

Google’s free NotebookLM made improvements today to what has become one of the most powerful ways to use AI.

NotebookLM helps you draw insights and learn from notes, documents, links, and recordings. It doesn’t do research on the Web, scan its internal databases, or make up content to please you. Instead it’s grounded in — and focused exclusively on— analyzing the sources you provide.

Today, Oct 17, Google improved NotebookLM’s most popular feature. You can now customize its automated AI-generated podcasts to focus on a specific topic, target audience, or document you’ve uploaded. I’ve been testing that and consider it a remarkable new way to generate audio.  

Read on for an updated overview of one of the year’s most surprising AI tools, including suggested ways to use it, a few limitations, and some alternatives.

Updates to Google’s Fast-Growing Experiment

What’s new: Since I first covered NotebookLM, it’s improved substantially.

Use it in 200 countries. You no longer have to be in the U.S to join.

Upload more files. You can now add all sorts of doc, links and multimedia.

Use Gemini’s best model. You can now ask questions about images, charts, or diagrams that appear in any of the sources you upload.

More specific, accurate citations. When NotebookLM identifies something in your notes or documents, you can click on a citation link to see a more precise selection from the material in your own notes.

Join as a team. As of Oct 17, Google is inviting businesses, universities, and other organizations to join a new pilot program for team collaboration. That will be useful in enabling team collaboration. It’s initially free, eventually paid.

New: Create customized audio posts

Customize Audio Summaries. You can now customize audio summaries. Here’s one I generated to test out a custom prompt. To create it, I uploaded a YouTube recording of a workshop I led about AI for educators and provided a few sentences of guidance to shape the focus of the audio clip.

How it works: Notebook LM can turn any files you upload into a 7 to 10-minute AI-generated podcast episode. Each notebook you create can include one of these “audio overviews” featuring a conversation between two AI voices. Here’s an example I made by uploading a Wonder Tools post I wrote about Perplexity.

Human-like voices. The quality of the voices and the AI summarization is so good that when I played a recording recently for a group, people assumed the hosts were human.

Enthusiastic. NotebookLM’s algorithm instructs the AI hosts to always be energetic. People have uploaded resumes and credit card receipts to test the hosts; they sound just as excited no matter what material you give them.

English only. NotebookLM can understand materials in close to 100 languages, and queries in numerous languages. The audio overviews are English-only for now because of voice-model limitations.

Mostly 7-15 minutes long. You can’t customize the audio length. If you provide brief material you may get a 3-minute audio result. Exhaustive texts may end up with 20-minute audio summaries.

How to get started

Sign up at notebooklm.google.com. It’s free with no usage limitations. Start by exploring sample notebooks included in your account.

After creating your own first notebook add Google Docs or Slides; PDFs, text and Markdown files; Copy-pasted text; Web links; YouTube links, or audio files.

Upload up to 50 files into each notebook. Each source can be up to 200mb and contain up to 500,000 words.

Create up to 100 notebooks. I have ones for research, teaching, and hobbies.

Apply AI to a collection of notes or documents

NotebookLM’s AI can generate overviews of the material in each notebook. You can ask specific questions about your materials, or you can start with some default resources.

FAQ = Get a summary of questions and answers related to your material.

Table of Contents = See a roadmap of what’s in your notebook.

Study Guide = Get quiz and essay questions and a glossary of key terms.

Timeline = See an ordering of events and a cast of key people.

Briefing Doc = View a summary of key themes, notable ideas, and quotes.

Chat. You can query your notebook just as you would chat with ChatGPT, Claude or any other AI bot. The benefit in NotebookLM is that it will provide a citation to the section of a document it relied on in answering your query.

Noteboard. Inside each notebook, there’s a bulletin-board style space for saving AI responses. You can use it to reflect on new ideas, annotate passages, or save observations that emerge as you interact with the AI.

Share Wonder Tools

How NotebookLM’s creator suggests using it

I spoke recently with Steven Johnson, who helped develop NotebookLM. He’s a popular author of 14 non-fiction books, who also has an interest in tech and serves as Editorial Director of Google Labs. He described his longtime interest in drawing connections between disparate thoughts and ideas that helped lead to the development of this . He shared with me a helpful approach to making the most of NotebookLM.

Create a master notebook. Put in the documents most important to you. A notebook can accept about 25 million words, enough to store a lifetime’s worth of thinking and writing. You can then query that notebook to surface your past ideas, research or writings that relate to current projects.

Create project-specific notebooks. Dump in your most recent 10 docs, slides or other materials and use the notebook as a reference resource.

Create an audio overview to review the week. Put in transcripts of your recent meetings and read the summary briefing and/or listen to the overview on your commute to review key conversations.

Create idea sparks 💡

Set up NotebookLM notebooks with material that inspires you. For example:

Create a collection of your favorite quotes or reading highlights. I exported my collection of thousands of Readwise highlights to a Google Doc, then imported it into a new NotebookLM. That allows me to apply AI analysis to highlights I’ve made over the past decade to Kindle books, online articles, and podcasts I’ve highlighted using Snipd (more on how I do that). I can now chat with the quotes and ideas that have inspired me over the past decade, incorporating them in new ways into my writing and teaching. Listen to an audio summary I asked NotebookLM to personalize for me.

Create an AI-enabled journal. Export your past musings to make it easy to search for patterns in your past thinking.

Turn class or reporting notes into AI-powered idea seeds. Convert anything you’ve learned into an AI-powered notebook. It’s easy to then generate study guides, FAQs, or query the notebook about various themes you’ve been exploring.

Build on your own notes

Add your own notes. After querying NotebookLM you can save its responses and build on them to make new notes. The summary and synthesized responses reside on your noteboard along with other AI queries and new notes you make.

Clarify, critique, or expand on notes. You can highlight any notes on your noteboard and ask for them to be clarified, critiqued, combined, or summarized. Or you can ask for an outline based on a note, or for it to suggest related ideas.

Dialogue with your notes. When you’re writing something or preparing a presentation, use NotebookLM to assist in exploring your materials. Initiate a dialogue with your own notes. The power to surface and connect valuable ideas otherwise hidden deep within your notes or documents is what I find most valuable about NotebookLM.

Query an individual document or a full collection. NotebookLM generates a summary for each source you add. It also adds subject tags based on the content. You can query one or more documents or the full notebook.

Caveats

NotebookLM is labeled an “experimental” service, so Google could potentially pull the plug anytime. Google has a history of suddenly closing promising projects, as documented in Killed by Google. These include Google+, Buzz, Wave, Reader, Songza, Fridge, Picasa, Picnik, Aardvark...

Available only to people 18 and older.

No mobile app, though the mobile Web version works well for simple queries.

Google doesn’t train its models on your material. But if you’re skeptical about the company, avoid uploading private material.

Alternatives

Claude Projects allows you to upload dozens of documents— up the equivalent of a 500-page book— for each of your projects. Like NotebookLM, it will remember everything in those materials so you can prompt the AI to help you understand, analyze, or apply the information, ideas, style, or tone of the work into anything that you’re working on. Here’s why I now rely on this. Pricing: $20/month

Differences: Claude accepts a much more limited number of file types than NotebookLM, has lower file size limits, and can’t generate audio. It excels at generating text, though, and can successfully answer a broader range of queries than NotebookLM.

Wondercraft is one of the most impressive new tools I’ve tested for generating audio summaries. If that’s the part of NotebookLM that interests you most, check out a video showing how you can direct Wondercraft’s AI voices. You can clone your own voice or use top-quality AI voices to generate audio from any text or link you provide. Pricing: free to start, then $29 or $59/month for more advanced usage.

Google’s Illuminate creates free audio summaries much like NotebookLM. It’s focused on summarizing academic papers and you can’t yet upload your own material. After selecting a source paper through a limited search, you specify your desired audience and tone. I chose a paper about ways of detecting AI-generated news and directed (screenshot) Illuminate to tailor the audio for professional journalists with a semi-professional tone. Here’s the result.

AnythingLLM and Jan are free software tools you can download to run an AI model on your own laptop. You can feed in files to analyze with AI. You can run these completely offline so your files files never reach a cloud server.

Differences: These tools work completely offline and allow you to customize which open-source AI models you use. They don’t, however, let you benefit from the specific capabilities of NotebookLM, like generating audio summaries or automatic briefing documents. These tools also require much more processing power from your computer and take up much more space on your local hard drive.

Additional resources

Read Google’s announcement about the new features.

Watch Steven Johnson  use NotebookLM to create a detailed lesson plan for an urban planning class.

Watch Productivity expert Tiago Forte explain how he uses NotebookLM.