How to Turn an Ebook into an Audiobook
Three realistic ways to turn a book into something you can listen to, compared honestly on cost, quality, speed, and what happens to your text.
Published 2026-09-23 · All guides
The short answer
You have three realistic options, and the right one depends on whether you need a file or just need to listen, and on how you feel about uploading the book:
- Listen live with built-in voices. Free and instant, but it doesn't make a file. (Read Aloud in your browser, or your device's accessibility reader.)
- Generate audio files on your own device with an open neural voice. Free and private. The quality is good, and the speed depends on your hardware. (EPUB to Audiobook, or command-line tools.)
- Send the text to a cloud text-to-speech service. The best voices and fast, but you pay per character and the text leaves your hands.
Human narration is a fourth option, and it's in a different league for quality and cost. It's covered at the end.
Before you start: DRM and rights
Books bought from Kindle, Apple Books, Kobo (mostly), and Google Play are DRM-protected. No legitimate tool will open them for conversion. Text-to-speech works on DRM-free files: Project Gutenberg and Standard Ebooks classics, many indie and publisher-direct purchases, and your own writing. Making an audio version for your own listening is generally treated like any personal format-shift. Publishing or sharing a narration of a book you don't hold the rights to isn't.
Option 1: Built-in voices, live
Every major platform can read text aloud with no download:
- In any browser: Read Aloud opens EPUB, PDF, and TXT files, highlights each sentence, and remembers your place.
- Microsoft Edge: the built-in "Read aloud" works on web pages and PDFs, with some of the most natural free voices available.
- iPhone and iPad: Settings → Accessibility → Spoken Content → Speak Screen, then swipe down with two fingers in any reading app that shows the text.
- Android: Select to Speak, or Google's "Reading mode" app.
Good for: starting right now and occasional listening. Not good for: offline listening, car stereos, or anything with the screen off. There's no file to take with you.
Option 2: A neural voice on your own device
Since 2024, small open-weight text-to-speech models have become good enough to rival commercial voices. The best known is Kokoro, an 82-million-parameter model under the Apache 2.0 license. It's small enough to run in a browser tab. EPUB to Audiobook runs it locally: pick chapters and a voice, and it writes an MP3 per chapter or one M4B. The book is never uploaded, and finished chapters are saved as you go, so a long book can be done over several sessions.
If you're comfortable with a terminal, open-source projects wrap the same kinds of models in command-line tools that process a whole EPUB in one run. They can use a graphics card, which is much faster, but they take more setup.
Speed is the trade-off. In the browser, the model runs on your processor, and on many machines that's slower than real time, so a 10-hour book can take more than 10 hours. It's a start-it-and-leave-it job. The browser tool measures your device with a short preview before you commit, and saves each chapter as it finishes.
Good for: private documents, manuscripts, public-domain classics, and anyone who'd rather not pay per book. Weak spots: mostly English-only, occasional mispronounced names, and one voice for the whole book.
Option 3: Cloud text-to-speech services
Google Cloud, Amazon Polly, Microsoft Azure, OpenAI, and ElevenLabs all sell text-to-speech by the character. A typical 90,000-word novel is roughly half a million characters. On basic neural voices that's a few dollars. On premium expressive voices, especially ElevenLabs-style narration, it can run well past $100. You also need an account, often an API key, and some way to split the book and stitch the audio back together. Several consumer apps wrap this for you at a markup.
Good for: the most expressive voices, multiple languages, and speed. Downsides: cost, and the full text of your book goes to a third party under its data-retention terms. That's a real consideration for unpublished work.
Comparison
| Built-in voices (live) | On-device neural | Cloud TTS | |
|---|---|---|---|
| Cost | Free | Free | Per character |
| Makes a file | No | Yes (MP3/M4B) | Yes |
| Text leaves your device | Only with cloud voices | No | Yes |
| Voice quality | Fair to good | Good | Good to excellent |
| Speed | Instant | Depends on hardware | Fast |
| Languages | Many | Mostly English | Many |
Getting the best result from any TTS
- Start from an EPUB if you can. A PDF's running headers, page numbers, and footnote markers get read out loud. If you only have a PDF, convert it to EPUB first and skim the text.
- Skip the front matter. Copyright pages and tables of contents make terrible listening.
- Pick one voice and keep it. Switching narrators mid-book is jarring.
- Package it as M4B. One file with chapters beats a folder of MP3s in every audiobook app. See What is an M4B?
What about human narration?
If you're an author publishing an audiobook for sale, human narration is still the standard listeners expect. Rates are typically quoted per finished hour and vary widely by narrator. Some retailers now accept AI-narrated titles, but their rules on disclosure and eligibility change often, so check each store's current policy before you publish. For personal listening, proofreading a manuscript, or accessibility, on-device TTS is more than good enough.