Listening to text with AI now splits into two very different approaches. One has a synthetic voice read the text exactly as written. The other has the AI reinterpret it into a conversation between hosts. The first is faithful; the second is convenient but no longer the original text. For anyone who wants to hear articles, notes or PDFs in the car, at the gym or for accessibility, the choice matters more than the marketing suggests.
- Gemini Notebook audio summaries: Google’s tool generates podcast-style discussions from uploaded sources in four formats, with adjustable length and interactive mode limited to English.
- Mobile screen readers: iPhone and Android built-in features read text verbatim, free and without uploading anything to the cloud.
- ElevenLabs and ElevenReader: More natural voices across 90+ languages and 17,500+ voice library, but terms of use restrict uploading content you don’t have rights to.
- Copyright caveat: Converting a book you don’t own for personal listening sits in a legal grey area; sharing the audio publicly is clearly communication to the public.
Gemini Notebook: a podcast built from your documents
Gemini Notebook is the name Google gave NotebookLM in July 2026, according to its official announcement; its help center still carries the old name in the header. Its audio summary feature generates, per Google, “in-depth debates between AI hosts” that summarize the sources you have uploaded to a notebook.
You can choose between four formats:
- Detailed information (default): two hosts converse and connect the themes of your sources.
- Summary: a single speaker presents the key points in under two minutes.
- Critique: two hosts evaluate the material, for example an essay.
- Debate: a formal discussion between two positions.
You also pick the language and can write instructions to focus on a topic. The tool supports more than 80 languages, including Spain Spanish, Catalan, Galician and Basque, and is available in Spain, according to its information page. There are limits: adjustable duration (shorter or longer) only works in English, the interactive mode, where you interrupt the hosts by voice, is also English-only, and Google warns that summaries, voices included, “may contain errors or audio glitches.”
Usage has a cap Google does not quantify in minutes: as of November 2026, Gemini Notebook limits depend on computing resources, renew every 5 hours up to a weekly limit, and are double with AI Plus and four times higher with AI Pro.
And in the Gemini app, without creating a notebook
The Gemini app can also generate audio summaries from documents, presentations, notebooks and Deep Research reports. You upload the file at gemini.google.com and ask for an audio summary. Google estimates it takes between 3 and 5 minutes and can be downloaded or shared via a link. The feature is not available to users under 18.
A practical difference: according to the help on notebooks in Gemini, the Gemini app cannot generate items from the Gemini Notebook Studio panel, such as its audio summaries, which are only in Gemini Notebook.
My advice: treat the summary as a trailer, not the movie. It helps you decide whether a report deserves your afternoon, not to study something high-stakes. It fails like any AI summary.
The phone’s voice reader: free, faithful and nothing uploaded to the cloud
If you want to hear the exact text, your phone already does it. It does not summarize or interpret: it reads.
On iPhone, Apple’s guide places it in Settings > Accessibility > Spoken Content, with three options:
- Speak Selection: select text and tap “Speak.”
- Speak Screen: swipe two fingers down from the top and it reads everything on screen, with pause and speed controls. You can also ask Siri to “Speak Screen.”
- Accessibility Reader: shows text full-screen and can read it aloud automatically.
On Android, the feature is called Select to Speak, according to Google’s accessibility help: enable it in Settings > Accessibility and launch it by swiping two fingers up. In addition, on websites and apps with selectable text you can choose the text and tap “Read aloud” in the pop-up menu without enabling anything. With the camera, it can read printed text via optical recognition in languages such as Spanish and Catalan. Exact paths may vary by manufacturer and Android version.
These features were born for accessibility and are, in my opinion, the best option for listening to long texts faithfully.
ElevenLabs and ElevenReader: more natural voices, with fine print
Specialized services offer more expressive voices. ElevenLabs announces on its text-to-speech page, as of November 2026, more than 90 languages with its Eleven v4 model, up to 10,000 characters per request and a library of more than 17,500 voices. Its reading app, ElevenReader, lets you upload PDF, DOCX, EPUB, Markdown and TXT files to listen to them.
The important part is in the terms. ElevenLabs’ terms of use for the European Economic Area say you retain rights to what you upload, but also that you cannot upload content for which you do not have all necessary rights, and that you guarantee its use does not infringe third-party rights. Gemini Notebook’s terms ask the same: do not share protected content without the necessary rights.
Can I convert a book or article that isn’t mine into audio?
To listen to it yourself, at home, there is some leeway; to share it, none. Spain’s Intellectual Property Law reserves to the author the reproduction, distribution, public communication and transformation of their work (article 17), and transformation includes adapting or modifying it (article 21).
The private copying exception in article 31.2 allows reproducing an already disclosed work without authorization if several conditions are met simultaneously: it is done by a natural person for private use and without commercial purposes, from a lawful source, without bypassing access conditions, and without the copy being used collectively or for profit. The article also requires the reproduction to be done “without assistance from third parties,” and its section 3 excludes from this exception works made available to the public on demand when a contract governs their reproduction. My reading is that this leaves in doubtful territory uploading an entire book to a cloud service that processes it for you, that an e-book bought on a platform is also governed by its terms, and that the law does not resolve it expressly. For a specific case, consult a lawyer.
What is beyond doubt is the other extreme. Publishing the audio of someone else’s article or book on a podcast or social network is public communication, and article 20 only excludes the strictly domestic sphere. If the audio is also an AI reinterpretation, you enter the territory of who holds rights over AI-created content, which is also unsettled.
My position: the phone reader is undervalued and podcast summaries are overvalued as a study tool. For a text you need to understand, listen to the original at 1.25x. To know whether a text deserves your time, the synthetic podcast is an excellent filter. What you should not do is confuse one with the other.