Podcastfy 🧩 Workflow Open source

What it is

Podcastfy by Tharsis Souza (Apache-2.0) is an open-source Python package and CLI that turns multi-modal content (websites, PDFs, images, YouTube videos, text) into engaging multi-speaker audio conversations, in the style of NotebookLM's audio overviews but programmable. You control the length, tone, language, speaker roles and voices, and can generate long-form episodes. Because it is a library rather than an app, you can run it on a schedule, for example turning each new blog post or newsletter into an audio version, and publish the resulting MP3 wherever you host your show.

Key features

• Inputs: URLs, PDFs, images, YouTube videos and raw text
• Transcript generation with 100+ LLMs (OpenAI, Anthropic, Google and more)
• TTS with OpenAI, Google, ElevenLabs or Microsoft Edge
• Custom conversation styles, roles and languages
• Long-form mode for extended episodes
• Python API, CLI and a hosted demo web app

How to install

Install FFmpeg, then pip install podcastfy, and set API keys for your LLM and TTS providers.

Pricing and rights

Free under Apache-2.0. LLM and TTS usage is billed by your providers; Edge TTS is free but lower quality. Summarising third-party content into podcasts may raise copyright issues when published.

Who it is for

Developers and newsletter or course creators who want to generate audio versions of written content automatically.

Verdict

A flexible, well-documented library for programmatic podcast generation. Updates have slowed (last push May 2026), and output quality depends heavily on the TTS provider.