open-notebook

An open source, privacy focused take on Notebook LM that talks to eighteen plus AI providers, including local ones, and lets you script your own podcasts instead of accepting whatever the host generates.

In a world dominated by AI, the README argues, the ability to think and acquire knowledge should not be restricted to a single provider. Open Notebook is the group's answer: an open source implementation of Notebook LM with more flexibility and features, written in TypeScript under the MIT license. Help and feature talk happen on a Discord server, and the project site is open-notebook.ai.

A comparison table in the docs lines the two products up. Open Notebook offers self hosting for privacy and control, provider choice across OpenAI, Anthropic, Ollama, LM Studio, and more, podcast speakers numbering one to four with custom profiles, content transformations, a full REST API, and deployment via Docker, cloud, or local. On each dimension the table positions the open alternative as the more flexible option.

Notebooks can hold PDFs, videos, audio, and web pages, and the AI work can run against eighteen plus providers. That range is the practical payoff of the philosophy above: switch providers when prices change, run sensitive material on a local model, and keep the same notebooks either way.

The README's own reasons to choose it run from privacy first design and cost control through cheaper providers or local models, to better podcasts because you control the full script, unlimited customization, and no vendor lock in since providers can be switched and deployments moved. The repo sits near 37,000 stars, which is loud for an alternative product.

It reads less like a clone of Notebook LM and more like a counterargument to it. How well the counterargument works depends on how much you actually want to self host.

• Source 1 · Project repository
• Source 2 · README