ebook2audiobook
DrewThomasson/ebook2audiobookReviewed by AI
Checked by RepoPicks Radar on
What is ebook2audiobook?
ebook2audiobook reads an e-book (EPUB, MOBI, PDF, Word, text and more, with OCR for scanned pages) and reads it out loud into an audiobook file with chapters, in .m4b, .mp3, .wav and other formats. Voices come from free speech models that run on your computer, in many languages, and you can upload a one to five minute recording to copy a voice. You work in a page that opens in your browser. It needs a download of the project and a launcher script that installs what is missing; a graphics card makes it far faster, as the models are very slow on a plain processor.
ebook2audiobook is free and open source (Apache-2.0) and can replace ElevenLabs, Speechify, and Audible.
How easy is it to install ebook2audiobook?
B · installer plus one simple step · Web
- Windows C
- macOS C
- Linux C
- Web B
The gentlest way in is the web: the author's free demo on Hugging Face Spaces, or the Google Colab notebook where you press the run buttons. On Windows, Mac and Linux you download the project folder and double-click a launcher (ebook2audiobook.cmd, or ebook2audiobook.command on Mac and Linux); it then installs the missing pieces itself (scoop on Windows, Homebrew on Mac) and opens the page at localhost:7860, which still means a download, a terminal window and a long first setup.
ebook2audiobook on GitHub
- Stars
- 20,526on Nov 10, 2026
- Downloads
- 992all releases
- Growth
- +29stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux, Web
- You need to install
- Python, Graphics card
- To build from source
- Python, Docker, NVIDIA graphics card
- Disk space
- 837 MB–2.5 GB · from the installer
- Accounts or keys
- None
- Cost
- Free · Free (the LICENSE file is Apache-2.0). No paid service is needed: the voices run on your own machine. The free hosted demo on Hugging Face and the Google Colab notebook are shared machines.
- License
- Apache-2.0 (Permissive)
- Can replace
- ElevenLabs, Speechify, Audible
- Last change on GitHub
- Nov 9, 2026
ebook2audiobook risk signals (6)
- Attentionvoice cloning of real people
The README's demo list includes "David Attenborough Voice" and "Scarlett Voice", and the feature is "Optional voice cloning using your own voice file", with the README saying the ideal recording is "around 1 to 5 mn". As a rule, many countries have laws that protect a person's voice and likeness.
AI review
- Attentioninstalls other software on your machine
README: "Note for MacOS users: homebrew is installed to install missing programs" and "Note for Windows users: scoop is installed to install missing programs without administrator privileges". The Dockerfile also fetches Calibre from "https://download.calibre-ebook.com/linux-installer.sh" and the Rust installer with "curl https://sh.rustup.rs -sSf | sh". So running the launcher brings in software from outside this repository.
AI review
- Attentionoptional public link exposes the app
README: "For Public Link: ./ebook2audiobook.command --share". That flag publishes the Gradio page on a public address, so anyone with the link can upload files and use your machine while it runs.
AI review
- Infoowner warns legally acquired books only
README, first notice: "This tool is intended for use with non-DRM, legally acquired eBooks only. The authors are not responsible for any misuse of this software or any resulting legal consequences." The project does not remove copy protection; it reads files you can already open.
AI review
- Infolicense text disagrees across files
The LICENSE file and GitHub say Apache-2.0, but the Dockerfile has LABEL org.opencontainers.image.licenses="MIT" and pyproject.toml classifies it as "License :: OSI Approved :: MIT License". Both are permissive licences, but the project does not state one consistently.
AI review
- Infomodels downloaded after install
requirements.txt pulls "coqui-tts[languages]", "piper-tts" and "huggingface-hub", and docker runs mount a "./models:/app/models" folder. The speech models are fetched when you first use a voice or language, on top of the 837 MB to 2512 MB Radar's check estimates for the install.
AI review
Evidence is shown in its original language.
What Radar could not check about ebook2audiobook
- Real disk use and download time, because each speech model and language is fetched the first time you use it.
- How long a whole book takes: the README says modern engines are "very slow on CPU" without saying how slow.
- Whether the automatic installation of scoop or Homebrew goes through cleanly on your machine.
- How good the result sounds in your language, out of the 1158 the project claims.
- Whether the free hosted demo on Hugging Face is running and how long its queue is.
- Whether chapter detection works on your file; the README warns that EPUB has no standard chapter structure and that you should delete unwanted text by hand first.