Skip to content
epub2audiobook Convert

EPUB to audiobook converters compared: where the book goes, and what comes out

Published

There are a dozen ways to turn an EPUB into speech, and they differ on three things that matter more than the voice: where the speech is made, whether your book leaves your machine, and whether you get a file at the end. Several popular options never give you a file at all.

The short version. With a graphics card and some patience for Python, ebook2audiobook makes the best-sounding audiobooks. To stay in a browser with nothing to install and nothing uploaded, use this site. If you only want to listen on your phone and never need a file, ElevenReader is the most polished app.

Prices and features below were checked in October 2026, against each project’s own README or pricing page. We have no affiliation with any of them.

At a glance

Tool What it is Where the voice is made Book uploaded? What you get Price
This site Web page Your CPU (Piper) No M4B with named chapters, or MP3 per chapter Free
ebook2audiobook Python or Docker, with a web UI Your CPU or GPU (XTTSv2 and others) No, unless you use a hosted copy M4B with chapters, or MP3, FLAC and others Free, open source
Audiblez Python, command line or GUI Your CPU or GPU (Kokoro) No M4B, chapters numbered rather than named Free, open source
epub_to_audiobook Python or Docker, with a web UI Azure, OpenAI or Edge in the cloud, or Piper and Kokoro locally Text goes to the cloud provider unless you choose a local one MP3 per chapter Free; cloud voices billed by the provider
OfflineTTS Web page Your device (Kokoro) No WAV or MP3, no M4B Free
ElevenReader Phone and web app ElevenLabs’ servers Yes Nothing: listening is in the app only 10 hours a month free; $11 a month
Speechify Phone, desktop and web app Speechify’s servers Yes Listening in the app; file export unclear Limited free plan; $29 a month or $139 a year
NaturalReader Web and phone app NaturalReader’s servers Yes MP3, on paid plans From $13.90 a month or $79 a year
Calibre Desktop ebook manager Your CPU (Piper) or your system’s voice No Nothing: it reads aloud in the viewer Free, open source
tts.ai Web page tts.ai’s servers Yes, up to 50 MB MP3 per chapter, or a zip 5,000 characters a day free; $5 to $99 a month

If you want the best voice

ebook2audiobook is the most capable tool here and the most popular by a distance, with over twenty thousand GitHub stars. It runs on your own machine, can use several speech engines including XTTSv2, can clone a voice from a sample, and covers more than a thousand languages. Its default output is an M4B with chapters and metadata.

The cost is hardware and setup. Its README recommends 8 GB of RAM and a graphics card with 4 GB of memory, and says plainly that CPU conversion is slow while a GPU can approach real time. You install it as Python scripts or a Docker container. There are hosted versions on Colab and Hugging Face too, but using them means uploading the book.

Audiblez is the lighter-weight alternative: pip install audiblez, plus ffmpeg and espeak-ng, and it reads with Kokoro-82M, a small model that sounds more natural than most. It writes an M4B, though its chapter marks are numbered rather than taken from the book’s own titles. On an M2 Mac it manages about 60 characters a second on the CPU; a GPU is roughly ten times faster.

If you want a file and nothing to install

This site runs Piper in the browser tab, so the EPUB is read from your disk and never uploaded. You get a chaptered M4B with the book’s own chapter names and cover, or one MP3 per chapter. It takes about an hour for a ten-hour book on an ordinary laptop, with no graphics card.

OfflineTTS also works on-device in the browser, using Kokoro’s voices, and reads text-based PDFs as well as EPUB. It exports WAV or MP3 rather than an M4B, and doesn’t mention chapter marks. Its own page notes that model downloads, analytics and some languages make network requests.

If you only want to listen

ElevenReader, Speechify and NaturalReader are reading apps rather than converters. You import an EPUB, their servers make the speech, and you listen in the app. The voices are among the best available, because they run on datacentre hardware.

What you don’t get is a file you keep. ElevenReader’s pricing page says its audio can’t be exported on any plan. NaturalReader exports MP3 on its paid plans. Speechify lists an offline MP3 download on Premium, but doesn’t say whether that means a file you can take out of the app.

If you already use Calibre

Since version 8, Calibre’s viewer reads aloud using Piper, the same engine this site uses, running locally. It speaks; it doesn’t save. The older “TTS to MP3” plugin is Windows-only, unmaintained since 2022 and reported broken on recent versions.

Calibre is still useful alongside any of these. It converts a DRM-free MOBI or AZW3 to EPUB, and it refuses to convert a locked book, which tells you quickly whether a file has DRM.

If you run Audiobookshelf

epub_to_audiobook was built for it. It writes one MP3 per chapter in the layout Audiobookshelf expects, and estimates the cost before converting when you use a paid voice. Its default voice is Azure’s, which is a cloud service, so pick Piper or Kokoro if you don’t want the text leaving your machine. Edge TTS is free but is also a Microsoft cloud service.

Where this site loses

It is slower than any GPU tool, and slower than every cloud service. The Piper voices are good but not the best here: Kokoro sounds more natural to most ears, and XTTSv2 and the commercial apps more so. We tried Kokoro in the browser and dropped it, because on a CPU it ran at about a third of real time against Piper’s eleven times.

It is English only, has eleven voices and no voice cloning, and reads EPUB but not PDF. If any of those is a dealbreaker, one of the tools above is the better choice, and the table says which.

The one with our name

There is a small open-source command-line script on GitHub also called epub2audiobook. It is unrelated to this site. It sends text to Microsoft Edge’s online speech service, is tuned for Russian-language books, and writes MP3s rather than an M4B. If you’ve read a description of “epub2audiobook” mentioning Edge TTS or Python, it was about that project.

What none of them can do

Every tool in this table fails on the same books: ones with DRM. An encrypted EPUB has no readable text in it, for any converter. Where to find EPUBs you can convert covers getting books that were never locked, and the Kindle guide covers the books most people are actually asking about.

Sources

← All guides