Google's AI Podcast Generator

How to Create an AI Podcast with Google's NotebookLM

Google's NotebookLM turns documents, website links, YouTube videos, or pasted text into a realistic, conversational two-host audio discussion, and it's free with any Google account. For the rest of my coverage in this space, see AI Audio & Music Tools.

1. Go to NotebookLM and sign in with your Google account.
2. Create a new notebook by clicking "Create."
3. Add your sources: upload PDFs or text files, paste a website or YouTube link, or paste text directly. You can add up to 50 sources, though 1-5 focused, related sources works best.
4. Open the Studio section and find "Audio Overview."
5. Click Generate and wait a few minutes. The finished audio typically runs somewhere between 5 and 15 minutes long.
6. Play, download, or share your finished Audio Overview once it's ready.

You Can't Reliably Control Length or Focus: Here's Proof

Before you invest time trying to steer the output: one documented test gave NotebookLM's Audio Overview custom instructions to keep it under 6 minutes, avoid certain overused phrases, and focus on a specific topic first. The result: the new audio ran longer than the original (over 12 minutes, despite the instruction to stay under 6), and it ignored the topic-focus instruction entirely. It did avoid the specific banned phrases, though that may have been coincidental rather than the instructions working. The practical takeaway: treat NotebookLM's Audio Overview as a one-click summary generator you don't get fine control over, not a scriptable podcast tool. If you need precise length or focus, plan on editing the result afterward rather than trying to prompt your way there.

Real Limitations to Know Before You Start

• No control over speaker count or gender. You get two AI hosts, can't choose how many or which voices, and there's no way to export a custom cloned voice into the mix. For that, see my ElevenLabs Review
• English only at this time
• Length is unpredictable. Most Audio Overviews run 5-15 minutes, but you can't set an exact target
• Single combined audio track. There's no way to split the two hosts into separate tracks, which complicates syncing with video avatars or further editing
• A recognizable "AI tell." In my testing, nearly every Audio Overview leans on phrases like "deep dive" or "delving into," which can feel repetitive across multiple episodes
• Hosts can sound overly enthusiastic about mundane details, which works for some content and feels odd for others

The Free Path From Article to Published Podcast

If you want to go beyond generating audio and distribute a podcast, one documented workflow does the whole thing with entirely free tools in under 15 minutes: generate the audio in NotebookLM from your article or notes, then use Headliner (also free) to turn that audio into a captioned audiogram video. From there, upload to Spotify for Podcasters, which automatically syncs your episode to Apple Podcasts, Amazon Music, and YouTube: distribution across all the major platforms from a single upload, plus built-in analytics to track performance.

Want More Control? Try Google AI Studio + Adobe Podcast Studio Instead

If what you want is narration control rather than a generated two-host format, that's a different job entirely. See Free AI Voiceover Tools. If NotebookLM's lack of control is a dealbreaker but you still want the podcast format, a different pipeline gives you more of a real production workflow: use Google AI Studio to brainstorm topics, build an outline, and draft a conversational script yourself (rather than letting NotebookLM auto-generate one from source material), then generate or record the voice, and import the result into Adobe Podcast Studio for professional noise removal, leveling, and mastering. This takes meaningfully longer, budget roughly 1-2 hours end to end across scripting, voice generation, and editing, but gives you real control over length, tone, and structure that NotebookLM's one-click approach doesn't offer. In my experience, a hybrid approach works well here too: use AI for the first draft and time-consuming cleanup, then have a human do the final voice read for warmth and personality.

Where This Fits in the Wider AI Audio Toolkit

This page sits in the AI Audio & Music section. NotebookLM writes and voices the discussion for you; for narration you script yourself, see How to Use Text to Speech and The Best Text to Speech Software. The ElevenLabs Tutorial covers the workflow for a voice you control end to end.

Frequently asked questions

Can I turn my notes or documents into a podcast with AI for free?

Yes. Google's NotebookLM does this directly and for free: upload your documents, notes, or links, and its Audio Overview feature generates a conversational discussion based on that material in a few minutes.

Can I control how long the generated podcast is?

Not reliably. A documented test found that custom instructions asking for a specific length were ignored: the output ran longer than requested despite explicit instructions to keep it short. Most Audio Overviews run between 5 and 15 minutes without a way to set an exact target.

Does AI replace podcast producers?

No. AI speeds up production and handles routine tasks like initial drafting and cleanup, but producers still make editorial choices, set tone, and are responsible for verifying that factual claims in AI-drafted content are correct before publishing.

What's the fastest way to publish an AI-generated podcast?

Generate the audio in NotebookLM, create a captioned video with a free tool like Headliner, then upload through Spotify for Podcasters, which automatically distributes to Apple Podcasts, Amazon Music, and YouTube from one upload.