Google LLC today rolled out new features for its artificial intelligence note-taking and research assistant NotebookLM, including the ability to upload videos from YouTube URLs and audio files directly, in addition to text, PDFs, Google Docs and web pages.
NotebookLM is an experimental tool from Google Labs that uses the company’s powerful Gemini 1.5 large language model’s multimodal capabilities to enhance note-taking, organize ideas and extract insights from source materials. It allows users to upload multiple sources including numerous documents, now including audio and video. Sources are limited to 500,000 words. Currently, NotebookLM is limited to a maximum of 50 sources and 1,000 notes per notebook.
Over the summer, the company expanded access to NotebookLM globally, and the new options will open up its capabilities for even more note-taking opportunities such as the analysis of video essays and lectures. It would also make group projects and discussions more productive by allowing it to listen to audio recordings from group study sessions or team meetings.
NotebookLM automatically generates a summary and source guide when documents are uploaded, including key topics and suggested questions designed to help users make deeper dives into the subject material. This allows users to interact with the sources more easily and provides passage highlights and citations that lead directly to the source material that generated responses.
The application produces two types of notes: saved model-generated responses from chat interactions or direct quotes, and user-composed written notes. The saved responses can be cited directly from the source material that generated them so that users can quickly locate their context. NotebookLM also has additional functions for generating study guides, short answer quizzes, essay questions, frequently asked questions and glossaries.
Earlier this month, Google introduced Audio Overview to the application, which allows users to turn their notes into an engaging audio discussion between two English speakers. The speakers, voiced by AI, discuss the subject like a pair of podcast hosts who go over the source material in depth. They summarize the material, make connections between topics and make friendly banter back and forth as they go.
The Audio Overview capability is still experimental and has some known limitations. For large notebooks, it may take several minutes to produce, but users need only click the generate button and walk away until they’re done. Overviews are currently only available in English and can sometimes introduce inaccuracies.
With today’s feature update, Google added an option to share Overviews with a URL. Users can now click on a share icon on the Audio Overview generated in NotebookLM and it will produce a public URL that can be shared on the web, social media or text message – even with users who do not have a NotebookLM account.