Skip to content
TheChatArchiveThe Chat Archive
Perplexitysonar-reasoning· PerplexityHugging Face LMArena Search Arena 24K

Google released the Audio Overviews feature on Gemini. Now, users can generate audio…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with gpt-4o-search-preview’s answer.
  1. User

    Google released the Audio Overviews feature on Gemini. Now, users can generate audio podcasts from uploaded documents straight from the Gemini UI

  2. Perplexity Assistant

    Google's Audio Overviews feature, now integrated into Gemini, enables users to convert uploaded documents, slides, or Deep Research reports into podcast-style audio clips. Here's how it works:

    Key Features

    • Podcast-style conversion: Upload files (PDFs, Google Docs, etc.), and Gemini generates a 10-minute audio discussion between two AI hosts. The hosts summarize content, draw connections, and provide dynamic perspectives[1][3][5].
    • Use cases: Ideal for students, professionals, or anyone needing to digest complex information quickly. Examples include turning class notes into audio for later review or transforming lengthy reports into digestible podcasts[1][3].
    • Integration: Available in both the Gemini web interface and mobile app. Users can trigger the feature via a suggestion chip that appears after uploading a document[2][4].

    Availability

    • Rollout: Globally available starting March 18, 2025, for Gemini and Gemini Advanced subscribers in English, with additional languages planned[1][4].
    • Export options: Generated audio can be downloaded for offline listening[2][4].

    Technical Notes

    • Source fidelity: Audio content reflects the uploaded material or Deep Research outputs, not comprehensive topic overviews[4].
    • Enhanced accessibility: Combines with Gemini’s Deep Research tool for deeper insights, such as analyzing YouTube videos or academic papers[3][5].

    This feature builds on Google’s earlier NotebookLM implementation, now streamlined for broader Gemini integration[3][5].

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.