Google’s NotebookLM is rolling out a cutting-edge feature that transforms documents into dynamic, podcast-style audio conversations. Announced on Wednesday, this new feature, called Audio Overviews, is now available to all users participating in the experimental NotebookLM tool. Unlike traditional text-to-speech systems, Audio Overviews leverages AI to create lively discussions, complete with humor and banter.
In a blog post, Google detailed how Audio Overviews works. Users can experience it by uploading files such as PDFs, TXT, markdown files, or by directly pasting text into the platform. Once the content is uploaded, users can click the Notebook guide icon to access summaries and suggested prompts. A new "Audio Overviews" option appears in the top right corner, enabling users to generate an audio discussion with just a tap.
The generated audio features AI hosts, both male and female, who engage in realistic, human-like conversations about the content. Early tests reveal that the AI voices deliver expressive dialogues, using emphasis, voice modulation, and natural pauses. The hosts even interrupt each other to add context or humor, creating an engaging and conversational atmosphere. Users can also download the audio files for later use.
Interestingly, the AI often pulls in additional information from the web, providing context beyond the original document. However, Google warns that these discussions are not always comprehensive or fully objective, as they are based solely on the uploaded content.
There are some limitations. As an experimental feature, generating audio may take a few minutes. At present, the tool only supports English, and users cannot interact with the AI hosts during playback. Additionally, occasional inaccuracies in the generated conversations are possible.