recito

recito turns a PDF or EPUB ebook into an audiobook you can listen to anywhere — narrated entirely on your own machine (an unpacked-epub directory, such as an ebook source tree, works too). It exists so that listening to a book you own doesn't mean renting a cloud narrator: no accounts, no uploads, no per-hour fees, nothing leaves your computer. The voice is an open-weight AI model and recito itself is free software.

Project page: recito.org Source: spacecruft.org/books/recito

recito reads PDFs the way a person does — two-column layouts, running headers, watermarks and page numbers are all handled — and epubs are read chapter by chapter, following the book's own table of contents. Numbers and abbreviations are spoken properly ("$3.5mn" is read as "three point five million dollars"). And every passage is checked by listening back to it: anything that comes out wrong is re-narrated automatically. The result is a single .m4b file with chapters, title, author and cover art, ready for any audiobook player.

What you need

• A computer with an NVIDIA graphics card (about 8 GB of video memory)
• Python 3.12 and ffmpeg
• About 5 GB of disk space for the voice model, downloaded once on first use

Install

pip install recito

Best in a virtualenv. Detailed setup and other ways to install: the install guide.

Quick start

recito speak mybook.pdf

A while later, mybook-ryan.m4b sits next to the book. If anything interrupts the render, run the same command again — finished parts are never redone.

ryan is the default voice; recito voices lists all nine and --voice serena picks another. To check the narration before committing to a whole book, skim the text it will read (fast, no GPU needed), or render a single page:

recito extract mybook.pdf --dump-text mybook.txt
recito speak mybook.pdf --voice serena --pages 10 -o preview.m4b

You can also narrate in your own voice from a few seconds of clean audio (--voice clone:ref.wav), and books in German, French, Spanish, Italian, Portuguese, Russian, Chinese and Japanese work with --language.

Learn more

• Usage guide — voices and cloning, other languages, reading part of a book, other output formats, per-book settings, covers and titles, troubleshooting
• Install guide
• For developers: DEVELOP.md (internals), ISSUES.md (known bugs), FUTURE.md (planned features), TTS.md (voice engines)

License

AGPL-3.0-or-later. See LICENSE.txt.

Copyright © 2026 Jeff Moe — Loveland, Colorado, USA