NotebookLM can now digest YouTube videos and audio files

Over the past two weeks, your feed has probably been flooded with NotebookLM's eerily lifelike AI podcasts. Two hosts bantering back and forth, turning dense whitepapers into digestible radio banter.

There was just one catch: it only accepted text. If your core research was trapped in a two-hour lecture video or a voice memo, you had to transcribe it elsewhere first.

Google plugged that hole. NotebookLM now lets you plug in public YouTube URLs and audio files directly, alongside your existing PDFs and Google Docs.

Video and audio get the grounding treatment

Powered by Gemini 1.5's multimodal capabilities, NotebookLM can now watch and listen to your sources directly.

Paste a public YouTube link, and the tool summarizes key concepts while showing an embedded player. Crucially, its inline citations link directly to the video transcript, so you can jump straight to the exact moment an argument was made without scrubbing a timeline.

Audio recordings work the same way. You can dump in team interviews or lecture recordings, and NotebookLM will search across the transcribed conversations to pull out specific nuggets. You can even blend class recordings, lecture slides, and notes into an automated study guide with a single click.

Google also fixed a major sharing bottleneck. That viral Audio Overview feature can now be shared via a direct public link. The only catch: Workspace users on enterprise or education plans can't generate these public share links just yet.

Google noted that personal data is never used to train NotebookLM, and Workspace admins can now enable the platform domain-wide.

Why it matters

We think this update turns NotebookLM from a niche paper-summarizer into a genuinely formidable research assistant.

For students, researchers, and creators, it is well worth testing immediately. Taking a dense 90-minute technical talk on YouTube, distilling it into cross-referenced notes, and then generating a 10-minute conversational podcast for your walk home is an unmatched workflow right now.

Strategically, Google is playing to its unfair advantages. While competitors scramble to build complex RAG pipelines for text, Google can seamlessly pipe its multimodal frontier model into its YouTube monopoly.

That said, there are obvious friction points. It currently relies on public YouTube URLs rather than arbitrary video uploads, and heavy enterprise users are boxed out of the easiest sharing mechanics. Running multimodal inference at this scale is also eye-wateringly expensive, raising questions about how long Google can keep this tool completely free before slapping on hard limits.