creative
Best AI Podcast Generators
Gemini Notebook (formerly NotebookLM) Audio Overviews and ElevenLabs GenFM compared on price, voices, script editing and commercial rights, with the unranked alternatives, the accuracy studies and the platform disclosure rules.
Quick answer: The best AI podcast generator for most people is Gemini Notebook (formerly NotebookLM), because it is free, it turns up to 50 of your own sources into a two-host Audio Overview in more than 80 languages, and it has by far the largest independent user record in the category, 4.9 out of 5 from about 62,000 ratings on the App Store. For a podcast you will publish, with a script you can edit and voices you choose, ElevenLabs Studio with its GenFM podcast generator is the better choice, because ElevenLabs lets you rewrite the script before any audio is made, its newest voice model ranks first on the independent Artificial Analysis text-to-speech leaderboard, and paid plans from $6 a month include a commercial licence. The one caveat: AI podcast hosts get facts wrong. University of Pittsburgh researchers found errors in every Audio Overview they made from three research papers (Futurity), so check every episode against its sources before anyone else hears it.
This page ranks end-to-end AI podcast generators: tools that take a prompt, a document, a link or a script and produce a finished episode, writing the conversation and voicing the hosts. That is a different job from the pages next to it:
- Best AI voice generators ranks text-to-speech: turning text you have already written into a spoken voice, one narrator at a time.
- Best AI voice clone ranks tools that copy a specific person’s voice.
- Best AI video editors ranks Descript and other editors you use to cut a show you recorded yourself.
- This page ranks tools that write and voice the whole episode for you.
Which one should you pick?
| If you want | Pick | Price | Why |
|---|---|---|---|
| A free podcast-style summary of your own notes, papers or documents | Gemini Notebook Audio Overviews | Free; higher limits from $4.99 a month (Google AI Plus) | 4.9 from about 62,000 App Store ratings; four formats; more than 80 languages |
| An episode you will publish, with an editable script and your choice of voices | ElevenLabs Studio (GenFM) | Free 10,000 credits a month; Starter $6 a month with a commercial licence | Script editable before audio; ElevenLabs’ newest voice model ranks first on the Artificial Analysis leaderboard |
| A hosted AI show with an RSS feed, up to four hosts | Jellypod (not ranked) | From $25 a month; 7-day free trial | Its own RSS feed and website; no user rating with real volume yet |
| A podcast generator you run yourself | Open Notebook (not ranked) | Free, open source | One to four speakers; can run entirely on your own machine |
| To record your own voice and use AI to edit it | An editor, not a generator | Varies | See best AI video editors |
The state of AI podcast generation
Google’s tool started the category and still leads it. NotebookLM’s Audio Overviews launched in September 2024, added an interactive mode in December 2024, and powered Spotify’s 2024 Wrapped podcast (Wikipedia). Google renamed NotebookLM Gemini Notebook on 16 July 2026; the address is still notebooklm.google and old links redirect (Google).
Generated shows now make up a large share of new podcasts. On 15 April 2026, Podcast Index’s new-feeds report flagged 45.7% of new shows in one 24-hour period as potentially AI-produced, using an AI classifier of its own (Podnews). A separate snapshot reported by Gizmodo on 3 May 2026 put the figure at 35.4% of one day’s new feeds. Both are single-day counts, and the share moves day to day.
The best-known AI podcast network shrank. Inception Point AI produced about 3,000 episodes a week at roughly $1 each in late 2025 (TheWrap). By 29 September 2026 its network had about 300 active shows, which it no longer tries to monetise, and the company now trades as Liquid Studios (Podnews).
Listeners are wary of AI hosts. In Sounds Profitable’s Podcast Landscape 2025 study of Americans who have listened to a podcast, 47% said they would be less likely to keep listening to a favourite show if it used AI voices, against 21% who said more likely (Sounds Profitable).
Microsoft left the category. Microsoft retired the Podcasts feature in Copilot on 18 August 2026 (Microsoft).
Platforms now set rules. Apple Podcasts requires shows with synthetic voices or AI hosts to disclose it in every episode and in the show’s metadata, Spotify removes shows that clone another creator’s voice without permission, and EU transparency duties for AI-generated audio have applied since 2 August 2026. The rules section below has the detail.
How we ranked these tools
A tool is ranked only when independent assessment of it exists from more than one kind of source: an independent benchmark, a hands-on review by an established outlet, a peer-reviewed study, or user ratings with real volume. Vendor claims are reported as claims. Most tools in this category are less than two years old and have thin independent records, so two are ranked and the rest are described without a rank. Much of the independent record for ElevenLabs covers its voices and the company as a whole, not GenFM alone, and its section says which is which. The general method is on how we rank.
The best AI podcast generators, ranked
| Rank | Tool | Free plan | Paid from | Hosts | Languages | Edit the script | Commercial use | Best for |
|---|---|---|---|---|---|---|---|---|
| 1 | Gemini Notebook Audio Overviews | Yes | $4.99 a month (Google AI Plus) | 1 or 2 | 80+ | No (steer with a prompt) | Not stated by Google | Most people; learning from your own sources |
| 2 | ElevenLabs Studio (GenFM) | Yes, 10,000 credits a month (no commercial licence) | $6 a month (Starter) | 1 or 2 | 32 at launch | Yes | Yes, on paid plans | Publishing an episode |
1. Gemini Notebook (formerly NotebookLM): best AI podcast generator for most people
Gemini Notebook is Google’s research notebook. You add sources, such as PDFs, Google Docs, slides, web pages and YouTube links, and its Audio Overview turns them into a podcast-style episode grounded only in what you uploaded.
Formats (Google help):
| Format | What you get |
|---|---|
| Deep Dive (default) | Two hosts unpack and connect the topics in your sources |
| The Brief | One speaker, under two minutes |
| The Critique | Two hosts give a constructive evaluation of your material |
| The Debate | Two hosts hold a formal back-and-forth debate |
- Length: Shorter, Default or Longer; length control works in English only.
- Languages: more than 80.
- Steering: you can type instructions on focus and expertise level before generating.
- Interactive mode: in English, you can join the conversation and ask the hosts questions while the episode plays.
- Output: download the audio file or share a link; public sharing is disabled for Workspace Enterprise and Education accounts.
- Not offered: script editing, a choice of voices, custom or cloned voices, music, and RSS publishing.
Independent evidence:
- User ratings: the iOS app is rated 4.9 out of 5 from about 62,000 ratings (App Store), the largest user record of any tool on this page.
- Hands-on comparison: XDA Developers (23 June 2025) tried three alternatives, ElevenLabs GenFM among them, and concluded that “NotebookLM still doesn’t have many real competitors”.
- Accuracy studies: University of Pittsburgh and Middlebury College researchers made Audio Overviews from three research papers on Mars and found errors in every one, always at the end of the overview (Futurity, 24 February 2026). A study in Radiology ran 21 research papers through NotebookLM, counted factually incorrect statements, and concluded the podcasts were limited by hallucinations.
Google’s own caveat: “Audio Overviews, including voices, are AI-generated and may contain inaccuracies or audio glitches” (Google help).
Price and limits: Gemini Notebook is free with a Google account. Since 2 September 2026, consumer usage limits are compute-based and refresh every five hours rather than daily (Google, 28 August 2026). Google publishes no fixed number of Audio Overviews per day.
| Plan | Price | Notebooks | Sources per notebook |
|---|---|---|---|
| Free | $0 | 100 | 50 |
| Google AI Plus | $4.99 a month | 200 | 100 |
| Google AI Pro | $19.99 a month | 500 | 300 |
| Google AI Ultra | From $99.99 a month | 500 | 500 to 600 |
Source: Google help.
Watch for:
- Commercial use: Google says it “won’t claim ownership” of content you generate (Google help), but that page gives no explicit licence to publish Audio Overviews commercially. If you plan to publish, check Google’s terms first.
- The voices: former NPR host David Greene is suing Google in Santa Clara County, California, alleging the male Audio Overview host copies his voice; Google calls the claim baseless and says the voice belongs to a paid professional actor (Inside Radio). The case is ongoing.
For using Gemini Notebook to study, see best AI study tools; for research across many PDFs, see best AI for PDF.
2. ElevenLabs Studio (GenFM): best AI podcast generator for publishing
ElevenLabs is a voice AI company, and GenFM is the podcast generator inside its Studio editor and its ElevenReader app. You upload a file or a link, GenFM writes a conversation or a single-host bulletin, and Studio voices it.
How it works (ElevenLabs help):
- Inputs: EPUB, PDF, TXT, HTML, DOCX, XML and FDX files, a URL, or an existing Studio project.
- Formats: conversation (a host and a guest) or bulletin (one host).
- Length: short, default or long.
- Voices: use the suggested voices or choose your own.
- Script: you can check and edit the script before any audio is generated, the main difference from Gemini Notebook.
- Language: detected automatically, with a manual override; GenFM launched in 32 languages on 6 December 2024 (ElevenLabs).
Independent evidence:
- Hands-on test: XDA Developers (23 June 2025) tried GenFM in ElevenReader and found the podcast “sounded natural, and was engaging to listen to”. At the time it allowed two free podcasts a month and one source per podcast, and the reviewer stayed with NotebookLM.
- Voice quality: ElevenLabs’ Eleven v4 model ranks first on the independent Artificial Analysis text-to-speech leaderboard, which ranks voices by blind listener preference. That leaderboard measures ElevenLabs’ text-to-speech models, not GenFM or its scripts.
- User ratings: ElevenLabs is rated 3.1 out of 5 from about 1,170 reviews on Trustpilot, with 48% five-star and 38% one-star reviews; complaints centre on billing, credit use and glitches. That rating covers all of ElevenLabs, not GenFM alone.
Price: writing the script is free; you pay the normal credit cost of converting the script to audio (ElevenLabs help).
| Plan | Price | Credits a month | Commercial licence |
|---|---|---|---|
| Free | $0 | 10,000 | No |
| Starter | $6 a month | 30,000 | Yes |
| Creator | $22 a month | 121,000 | Yes |
| Pro | $99 a month | 600,000 | Yes |
| Scale | $299 a month | 1,800,000 | Yes |
| Business | $990 a month | 6,000,000 | Yes |
Source: ElevenLabs pricing. ElevenLabs discounts the first month of Starter and Creator; the prices above are the regular monthly rates.
Why it ranks second: ElevenLabs gives you control that Gemini Notebook does not, an editable script, a choice of voices and a commercial licence, but its free plan carries no commercial licence, it takes one source at a time where Gemini Notebook takes up to 50, and its user ratings are mixed. Gemini Notebook is free and has a far larger and more positive user record.
For ElevenLabs as a text-to-speech tool, see best AI voice generators.
Other AI podcast generators worth knowing (not ranked)
These tools are available today, but we did not find enough independent assessment to rank them. Descriptions come from each company’s own pages.
| Tool | What it does | Price | Independent record we found |
|---|---|---|---|
| Wondercraft | Turns a document, URL or ideas into an episode with AI hosts, voice cloning, music and sound effects. | Free 150 credits (no commercial rights); Creator $25 a month ($21 billed annually); Pro $45 a month | Trustpilot 2.1 from 18 reviews |
| Jellypod | Up to four AI hosts, a script editor, its own RSS feed and website, and voice clones. Claims more than 3,300 voices in 121 languages. | Starter $25 a month ($300 a year); Creator $50; Business $125; 7-day free trial, no free plan | No rating with real volume found |
| Speechify AI Podcasts | iOS app that makes Lecture, Debate, Talk Show and Late Night Show episodes. | In-app purchases of $9.99 and $79.99 | App Store 4.5 from 44 ratings |
| Adobe Acrobat Studio | Turns documents into podcast overviews. | $24.99 a month | None found specific to the podcast feature |
| Podcastfy | Open-source Python library that turns text, websites, YouTube videos and PDFs into audio conversations, using OpenAI, Google, ElevenLabs or Microsoft Edge voices. | Free (Apache 2.0); you pay for any voice API you use | About 6,600 GitHub stars |
| Open Notebook | Open-source, self-hosted alternative to Gemini Notebook with one to four podcast speakers; can run fully on your own machine. | Free (MIT licence) | About 39,800 GitHub stars |
If you record human hosts yourself, you need an editor and a recording tool with AI features, not a generator. Those are compared on best AI video editors and best AI for transcription.
How accurate are AI-generated podcasts?
AI podcast generators summarise your sources in conversation, and the independent tests we found all reported errors.
| Study | What was tested | Finding |
|---|---|---|
| University of Pittsburgh and Middlebury College (Futurity, 24 February 2026) | Three research papers on Mars through NotebookLM | Errors in every overview, always at the end, including an unsupported inference of liquid water and possible life on Mars |
| Tejani et al., Radiology (2025) | 21 radiology research papers through NotebookLM | Podcasts limited by hallucinations and by content too simplified for radiologists |
| Rettberg, arXiv 2511.08654 (11 November 2025) | Cultural framing in NotebookLM audio | The system “translates cultural contexts to a white, educated, middle-class American default” |
What this means in practice:
- Treat an AI podcast as a study aid or a first draft, not a finished authority.
- In the Pittsburgh test the errors came at the end of each overview, where the hosts draw conclusions, and some were subtle enough that a non-expert would miss them.
- If you publish, use a tool with an editable script, such as ElevenLabs Studio, and check the script against your sources before generating audio.
Rules for publishing an AI podcast
| Platform or law | Rule | Source |
|---|---|---|
| Apple Podcasts | Shows using synthetic voices, AI-generated hosts or AI replicas must prominently disclose it in the content and metadata of every episode and the show; AI must not be used to mislead | Apple Podcasts guidelines, sections 1.11 and 1.12 |
| Spotify | Removes shows that impersonate another creator’s or host’s likeness without permission, including through AI voice cloning; extended Verified badges to podcasts on 20 May 2026 | Music Business Worldwide |
| YouTube | Creators must label realistic synthetic content; using AI to write a script needs no label. Since 15 July 2025, mass-produced content made from generic templates cannot earn money from the YouTube Partner Programme. | YouTube help; YouTube monetisation policy |
| EU AI Act, Article 50 | Since 2 August 2026, providers of AI that generates audio must mark it as AI-generated in machine-readable form, and anyone publishing deepfake audio must disclose it. Systems already on the market before that date have until 2 December 2026 to add marking. | Kingsley Napley, 14 August 2026 |
Voice rights: cloning a real person’s voice without consent can break platform rules and, in many places, the law. Tools for cloning your own voice are compared on best AI voice clone.
How to make an AI podcast
- Gather your sources. Upload the documents, links or notes the episode should cover.
- Pick the format. A two-host conversation suits explainers; a single host suits news briefings.
- Steer it. In Gemini Notebook, type the focus and audience before generating; in ElevenLabs Studio, edit the script directly.
- Check the facts. Compare the episode or script against your sources; the independent tests above found errors in generated episodes.
- Choose the voices. Gemini Notebook uses Google’s preset voices; ElevenLabs lets you choose, and Jellypod and Wondercraft offer voice clones.
- Add music if you want it. Background tracks can come from an AI music tool; see best AI music generators.
- Disclose and publish. Label the show as AI-generated where Apple Podcasts or the EU AI Act requires it, and check the tool’s licence permits commercial use.
Best AI podcast generator by need
Best overall: Gemini Notebook
Gemini Notebook is the best AI podcast generator for most people, because it is free, works in more than 80 languages and has 4.9 out of 5 from about 62,000 App Store ratings.
Best for publishing: ElevenLabs Studio
ElevenLabs Studio is the best AI podcast generator for episodes you will publish, because you can edit the script before audio is made, choose the voices, and get a commercial licence from $6 a month.
Best free option: Gemini Notebook
Gemini Notebook is the best free AI podcast generator, because every Audio Overview format is available on the free plan.
Best for studying: Gemini Notebook
Gemini Notebook is the best choice for turning lecture notes and readings into audio, because its overviews draw only on the sources you upload and its interactive mode lets you ask the hosts questions.
Best for recording your own show: an editor
If you record human hosts, a generator is the wrong tool; see best AI video editors.
Frequently asked questions
What is the best AI podcast generator?
The best AI podcast generator for most people is Gemini Notebook, formerly NotebookLM, because it is free, turns your own documents into two-host Audio Overviews in more than 80 languages, and is rated 4.9 out of 5 from about 62,000 App Store ratings. ElevenLabs Studio with GenFM is the better choice for publishing, because it lets you edit the script, pick voices and use the audio commercially on paid plans from $6 a month.
Is NotebookLM still free?
Yes. NotebookLM, renamed Gemini Notebook on 16 July 2026, is free with a Google account, including Audio Overviews. Since 2 September 2026, usage limits are compute-based and refresh every five hours. Google AI Plus ($4.99 a month) and Google AI Pro ($19.99 a month) raise the number of notebooks and sources.
What happened to NotebookLM?
Google renamed NotebookLM to Gemini Notebook on 16 July 2026. It remains a standalone product at notebooklm.google, old links redirect, and Audio Overviews work as before.
Can I publish a NotebookLM podcast commercially?
Google says it won’t claim ownership of content you generate in Gemini Notebook, but its help page gives no explicit licence covering commercial publication of Audio Overviews, and the voices are the subject of a lawsuit by former NPR host David Greene, which Google calls baseless. For a podcast you plan to publish commercially, ElevenLabs Studio on a paid plan includes a commercial licence.
Are AI-generated podcasts accurate?
Not reliably. University of Pittsburgh researchers found errors in every NotebookLM overview they made from three research papers, and a study in Radiology concluded that NotebookLM podcasts of 21 research papers were limited by hallucinations. Check any AI podcast against its sources before relying on it or publishing it.
Can I make a podcast with my own voice using AI?
Yes. ElevenLabs offers instant voice cloning from its $6 Starter plan, and Jellypod’s plans include voice clones. Tools for cloning your own voice are compared on our best AI voice clone page.
Do I have to say a podcast is AI-generated?
On Apple Podcasts, yes: shows with synthetic voices or AI hosts must disclose it in every episode and in the show’s metadata. In the EU, deepfake audio must be disclosed under Article 50 of the AI Act, which has applied since 2 August 2026. YouTube requires a label for realistic synthetic content, and Spotify removes shows that clone another person’s voice without permission.
How many podcasts are AI-generated?
Between about 35% and 46% of new podcast feeds on a given day were flagged as likely AI-generated in Podcast Index data reported in April and May 2026. Both figures are single-day counts.
Do people listen to AI podcasts?
Many listeners are wary. Sounds Profitable’s Podcast Landscape 2025 study found 47% of Americans who have listened to a podcast would be less likely to keep listening to a favourite show if it used AI voices, against 21% who said more likely.
What happened to Copilot Podcasts?
Microsoft retired the Podcasts feature in Copilot on 18 August 2026.
What is the difference between an AI podcast generator and a text-to-speech tool?
An AI podcast generator writes the episode and voices it, turning a document or prompt into a conversation between hosts. A text-to-speech tool only reads text you have already written in one voice. For text-to-speech, see our best AI voice generators page.