NotebookLM: Audio Overview — audio summary from your sources (2026)

Audio Overview is an AI-generated audio summary built from sources you already uploaded to the notebook — a discussion or monologue that reflects that material, not free-floating model opinions. In about 18 minutes you start from a notebook with ready sources (if you do not have one yet, follow NotebookLM: your first useful briefing first), pick the format that fits your commute or meeting, add a focus prompt, generate in Studio and listen while noting gaps before you share or download.

1. Steps 1

   Open a notebook that already has sources ready

   Go to notebooklm.google.com and open your project notebook — do not start an Audio Overview on an empty notebook. Check that sources show ready state and the topic is coherent (one deliverable, one client, one study). If you mixed PDFs from another project, create a fresh notebook and upload only what you want summarized in audio.

   What you should see: A sources panel with several indexed items and no persistent load errors.

2. Steps 2

   Studio → Audio Overview and pick a format

   In the Studio panel, select Audio Overview. Before generating, choose the format that fits: Brief — single speaker, key takeaways in under two minutes, great for a short commute; Deep Dive (default) — two hosts in a longer, lively conversation; Critique — two voices with constructive evaluation (useful for an essay or design doc in your sources); Debate — formal back-and-forth on the topic. Do not generate all four the same day — one per goal.

   What you should see: Audio Overview card in Studio with a format picker (Brief, Deep Dive, Critique, Debate).

3. Steps 3

   Set language, length and a focus prompt

   Pick output language (80+ languages supported; default often follows your Google Account preferred language). Set length: Shorter, Default or Longer — note that Longer is English only. Optionally add a short prompt: «focus on regulatory risk», «executive level, no jargon» or «compare the two proposals in doc 3». That steers the script without replacing your sources.

   What you should see: Language and length controls plus an optional instructions field before Generate.

4. Steps 4

   Generate and let it finish in the background

   Click generate. NotebookLM builds the audio in the background: you can stay in the notebook, chat or start another Studio artifact meanwhile. Generation usually takes a couple of minutes. Output may include inaccuracies or minor audio glitches — do not treat it as a certified transcript. If you created one before, Studio → Load brings back a previous Audio Overview without redoing the job.

   What you should see: A generating indicator, then a player with audio ready to play.

5. Steps 5

   Listen critically and open citations while it plays

   Play the overview and check whether topics match your sources, not what merely sounds good. You can keep working in NotebookLM: when a sensitive figure comes up, pause and open the citation in the PDF or transcript. Note in an external doc: OK / UNSURE / NOT IN SOURCES. Audio helps you spot gaps; it does not replace reading the passage.

   What you should see: Active player and a citation panel with highlighted source text after you pause.

6. Steps 6

   Download or share with permissions and context

   If someone else needs the audio, use share link (the notebook must be shared appropriately), share the whole notebook or Download to save the audio file. Always add written context: which sources were included, generation date and what they do not cover. Do not ship the MP3 as an audited report or a substitute for a written briefing.

   What you should see: Share or download menu by the player and your note with date and source list.

7. Steps 7

   Optional: interactive mode (English only, new overviews)

   On newly generated overviews, NotebookLM may offer interactive mode in English: join with your voice, ask questions grounded in sources and hosts answer before resuming the thread. Use it to clarify one point, not to outsource diligence. Glitches happen; if interactive mode is unavailable for your audio or language, return to cited chat in the notebook.

   What you should see: Join conversation with microphone or an interactive mode control on the player when available.