It's no secret that NotebookLM has become a fast favorite tool here, especially because of how well it helps dissect complex research topics. But Google isn't the only AI model provider out there, and you're handing even more of your data to the search giant in return for using Gemini's models. I've been trying to self-host more services to reduce my dependence on the big cloud providers, so naturally I went looking to see if there was an alternative to NotebookLM.

And I found one. Well, a few, really, but Open Notebook is the one that I'm playing around with right now. It supports a wide range of model providers, has multiple content formats, has a chat assistant, and has most of the other killer features from Google's offering. Oh, and yes, it has the podcast creation tool, which is pretty fun.

What is Open Notebook?

Self-host your own AI-powered notebooks

Open Notebook is an open-source, AI-powered note-taking platform that lets you create notebooks for querying with the latest AI models. These notebooks could use URLs, PowerPoint, PDF, or other types of files, or you can feed it YouTube links to get AI transcripts for further analysis.

Open Notebook has many of the tools that initially drew me to NotebookLM, plus it's not limited to only Google's models:

  • Model Provider Support - Choose from multiple AI providers
  • Content Support - Support for multiple content formats
  • Chat Assistant - Talk to your AI assistant for insights
  • AI-Powered Notes - Enhanced note-taking with AI assistance
  • Transformations - Process and enrich your content
  • Search and Ask - Powerful full-text and vector search
  • Podcast Generator - Transform notes into engaging podcasts

The only thing it can't use as a source is a direct connection to Google Docs, Notion, GitHub, or any of the other cloud services that some other NotebookLM replacements can link to. In my books, that's a win, because it keeps Open Notebook limited to the data you've explicitly uploaded to it, and it doesn't have full reign to traipse through your cloud storage.

With a focus on privacy

Open Notebook pulls all your sources into a local database and keeps them there. During the upload, you can set several preset transformations, which are things like summaries, key points, and the like, or you can specify your own transformations in the same way that you would craft a prompt for a chatbot. You can also keep all the AI tools local by running an Ollama instance in your Docker stack, downloading the necessary models, and pointing Open Notebook to Ollama.

Open Notebook is quick to get running

But you'll need to go on an API hunt first

If you've got any Docker exposure, running Open Notebook won't take more than a few minutes. Copy the example on the GitHub page, save it as a docker-compose.yaml file like you'd do with any other container, and go on an API key hunt for the docker.env file containing all the container variables.

Open Notebook has a huge range of supported models, including:

  • OpenAI
  • Anthropic
  • Gemini
  • Vertex.ai
  • Mistral
  • Deepseek
  • Ollama
  • Open Router
  • Groq
  • xAI
  • Elevenlabs
  • Voyage AI
  • Azure OpenAI
  • Firecrawl
  • Jina

You can pick the AI models you want for all the different tools in Open Notebook, and I recommend you play around to see which works the best with your workloads. Google's Gemini models are best for larger context windows, and Elevenlabs are best for podcast creation, but other than that, you can pick the models you vibe best with. This gets you a tool suited to your needs and your budget, and not one that's limited in scope.

Once you have sources added to your notebooks, you can get into transformations, which are really the summaries and other insights that NotebookLM will create for you without being prompted. Or you can use the sidebar chatbot to ask deeper questions about context, and then save the answers as notes, deepening your understanding, and the amount of data the AI models have the next time you have questions.

Podcast generation is my favorite tool

Podcast generation was my favorite tool in NotebookLM, and while it's not quite as polished in Open Notebook, it works. Again, customization is key here, and you can set the roles of the participants, their conversation style, engagement techniques, and dialogue structure to get a podcast you'll want to listen to, and not the one Google thinks you want to hear.

Or I can query the documents I have saved for inspiration, recall, or insight

I'll be the first to say my memory is terrible, but that's why AI tools like Open Notebook make so much sense to me. I can add my notes, things I've come across while browsing, and insights from experts, and combine them all into purpose-built notebooks for further query when I need to revisit a topic. It's the best way to support my working method, but it is a boon to anyone who has to research any topic deeply.

Unfortunately, it's not a complete replacement for NotebookLM

The draw for NotebookLM is how easy it is to use, and that's a simple fact. The setup for Open Notebook is tiresome, with a dozen or so different AI provider APIs to handle, then integrate and troubleshoot when they give you errors. I quickly ran into token limits for my free tiers of every service, so this is more a tool for those who want the best AI model for their needs, balanced against the cost of using the API over time.

Self-hosting my AI notebook keeps my essential data where I want it — on my own server

Despite the amount of setup woes I encountered, I'm going to keep using Open Notebook. I like having my data saved on my own hardware, so that if Google, or anyone else, decides to stop developing the tool I've been using, I don't lose my curated data. That's happened more times than I can remember with Google services, especially, and I don't want to make that mistake again.

blurb