Google's AI podcast generator just got more customizable, and I put it to the test

NotebookLM, Google’s experimental AI-first notebook and research assistant, has gone viral for its Audio Overview feature, which transforms a user’s material into a podcast. The feature just got an upgrade to fine-tune the experience, and I put it to the test. 

The new features allow users to quickly instruct the two AI podcast hosts on the topics they want them to discuss. For example, if you upload a large document, you can instruct AI Overviews to make the podcast focus on only one section instead of the entire thing. 

Also: Google’s stunning AI podcast tool gets new features that make it even better

To test out the technology, after I signed in, I started a new notebook and uploaded a PDF of my best AI image generator roundup for ZDNET. I have been curating the list for over two years and making it as comprehensive as possible, so it’s a massive document. 

Then, I dropped the PDF into the “Add sources to get started” box and immediately discovered the materials AI could generate, including a FAQ, Study Guide, Table of Contents, Timeline, Briefing Doc, and Audio Overview. 

Also: I tested Meta’s limited edition Ray-Ban smart glasses, and they’re near-perfect

In Audio Overview, there are now two options: Customize and Generate. To test its ability to zero in on a topic, I clicked on “Customize”, where I could enter specific instructions. Google gives you suggestions for the type of customization prompts you can include, such as “Focus on a specific source”, “Focus on a specific topic”, and “Target a specific audience”. 

I typed, “Can you focus on the best image generator overall?” Within minutes, it generated a seven-minute podcast episode where both AI hosts chatted through my best pics for AI generators, summarizing all my points and suggesting why each generator got the best pick. You can listen to two of those minutes below: