Podcastfy is an open-source Python alternative to Google NotebookLM’s podcast feature: it transforms websites, PDFs, images, YouTube videos or a bare topic into engaging, multilingual two-host audio conversations. It has earned about 6,700 stars on GitHub, is Apache 2.0 licensed, and is actively developed in 2026.
NotebookLM’s Audio Overviews showed how compelling AI-generated podcast conversations can be โ but they sit inside a closed product. Podcastfy reimplements the idea as a programmatic pipeline you control: feed it one or more content sources, and it extracts and cleans the material, drafts a natural conversational transcript with an LLM of your choice (100+ models via OpenAI, Anthropic, Google or local models), and renders the dialogue with text-to-speech from OpenAI, ElevenLabs, Google or Microsoft Edge voices. Short episodes of a few minutes or long-form shows of 30+ minutes are supported, in multiple languages, with customizable voices, styles and formats. It runs as a Python package, a CLI, or a FastAPI service, and a hosted web demo (OpenPod) is built on it, making it easy to embed podcast generation into products, research workflows or content pipelines.
Free. Podcastfy is open source under the Apache 2.0 license; you bring your own LLM and text-to-speech API keys (or run local models) and pay only those providers.
Content creators, educators, researchers and developers who want NotebookLM-style audio overviews inside their own tools and workflows โ with full control over sources, voices, languages and models.