Open Notebook

🚀 Description

lfnovo/open-notebook by Luis Novo — the open-source, self-hostable answer to Google’s NotebookLM. Drop in PDFs, web pages, YouTube videos, audio, or office docs; chat with them, ask grounded questions with citations, and generate a synthesized multi-host podcast from the material. ~28k stars in June 2026.

Unlike Google’s hosted product, you own the data and pick the models — run it on hosted APIs (OpenAI, Anthropic) or completely local via Ollama / LM Studio. 16+ providers supported.

🧩 Features

  • Source ingestion — PDFs, URLs, YouTube, audio, office docs; even thousand-page PDFs.
  • Grounded Q&A — answers cite the specific source passage.
  • Podcast generation — multi-host, configurable tones, editable script; “hyper-customizable because it’s all local and open source.”
  • Transformations — extract key insights, dense summary, analyze-a-paper, reflection questions, simple summary + table of contents.
  • Per-process model selection — separate models for chat, embeddings, TTS, STT, tools, and transformations; mix providers freely.
  • Local-capable — power voice + LLM entirely on-device.

Reasoning for

A self-hosted research companion that doubles as a podcast generator — directly relevant to Travelcast AI (audio/podcast pipeline) and to my own Building a Second Brain / Second Brain Design workflow for interrogating long sources. The local option keeps private documents off third-party servers. Pair the podcast voices with ElevenLabs for more natural output (as the source video does).


Template: tool