I found an open-source NotebookLM alternative that’s powerful, private – and free

Follow ZDNET: Add us as a preferred source<!–> on Google.

ZDNET’s key takeaways

• Open Notebook is an open-source take on NotebookLM.
• Open Notebook can be installed and used for free.
• You do have to understand Linux and containers to deploy it.

NotebookLM has become an incredibly popular tool for research, organizing, interacting with documents, generating summaries, creating flashcards, and producing tailored reports. With NotebookLM, you can add your own sources and then use AI to make informative connections between those sources, which leads to an even better understanding of the material. NotebookLM only works on uploaded material, so you control the sources.

Also: How to use Claude (and how it’s different from ChatGPT)

The biggest issue with NotebookLM is that it’s a cloud-based service, so you cannot run it locally. Although you can link NotebookLM to locally-installed AI (such as Ollama), that combination’s use is limited to certain functions.

If you want to limit NotebookLM to only use locally-installed AI, you’d be out of luck.

What do you do if you place a high value on privacy, but still need to use a tool like NotebookLM within your home network?

Fortunately, there’s an open-source alternative that can be deployed that is just as powerful and can use your AI of choice… even locally-installed AI.

That tool is called Open Notebook, and it is just as powerful as NotebookLM.

If you’re serious about your privacy and the privacy of your data, Open Notebook is the only way to go.

Think about it this way: using NotebookLM requires that you upload your sources to a third-party cloud service. Those sources could contain sensitive information about you, your family, your friends, or your company. Do you really want to hand that over to a third party?

I don’t. 

That’s why I opted for Open Notebook over NotebookLM.

Let me describe my experience installing this tool.

Also: How to use Copilot Pro to write, edit, and analyze your Word documents

It’s all about containers

Deploying Open Notebook requires an understanding of Linux and containers. Before you shy away, using containers isn’t nearly as hard as you might think. No, it’s not nearly as easy as double-clicking an installer file, but it’s also not as hard as, say, installing and using a relational database. Without going too deep into the steps for installing Open Notebook, the process looks like this:

1. Install Docker
2. Clone the Open Notebook Git repository
3. Configure the docker-compose.yml and docker.env files to use whichever LLMs you prefer
4. Deploy the Open Notebook container
5. Point your browser to http://SERVER:8502
6. Create an account (it’s free and only stored locally)
7. Define which LLMs are to be used for specific tasks
8. Create a notebook and upload your sources

If you’re not familiar with editing .yml files or using Docker, deploying Open Notebook will be a chore, but once you learn how it’s done, you’ll have opened your world up to the limitless world of containers… so it’s a bonus.

The challenge of choosing your LLMs

This is where things can get a bit tricky with Open Notebook. Not every LLM is created equal. For example:

• Quen2.5-coder is good as a coding companion
• Wizard-math is a great logic partner
• Gemma 3 is a localized version of Gemini
• Llama 3 is great for long-form text, maintaining context, and performing complex reasoning.
• AnythingLLM is a good option for running AI agents and handling common AI tasks.
• GPT4ALL is good for document analysis and building knowledge bases.
• GPT-OSS is used for reasoning and agentic tasks

Your best bet is to decide on how you plan on using Open Notebook. Once you have the service up and running, you’ll need to select a model for tasks like chat, embedding, function calling, speech-to-text, transformation, and more. The good news is that Open Notebook has documentation to help you understand model types and make the right choices.

Ease of use

Deploying Open Notebook is the only challenge with this tool. Once it’s up and running, using it is very simple. The process looks like this:

1. Create a new notebook
2. Add sources to the new notebook
3. Select the model you want to use
4. Query the local AI about your sources

Also: I turned a YouTube video into flashcards with NotebookLM – here’s how

I ran some quick tests on this and found it to not only be easy to use, but fast and as accurate as the sources provided. In other words, if your uploaded sources are accurate, the results will be as well. If your sources contain errors and wrong information, so too will the results of your chats.

For example. I downloaded this article as a .docx file and then uploaded it as a source. I then queried the AI with “What are the steps for installing Open Notebook?” The tool dove into my sources, and then it responded based on my writing and only my writing.

Next, I uploaded a different article I wrote on deploying Docker, and then queried with “How do I install Docker on Ubuntu?” Once again, because the source was accurate, the results were as well.