About
What Sandbook is, and what it will never do
Sandbook is an offline text-to-speech book reader for iPhone and iPad. It reads EPUB, PDF, TXT, RTF, Markdown and HTML files, web pages, scanned pages and typed or pasted text aloud in 41 voices (US and UK English, Spanish, French, Italian, Brazilian Portuguese and Hindi), with the Kokoro-82M speech model running on the device itself through Apple MLX. Free. No in-app purchase, no subscription, no character or minute cap.
iOS 18.2 or later · iPhone 12 or later & iPad · No account
Four things this app will never do
Every app says it respects you. The four below are the version of that claim you can check, because each one is a property of how the app is built rather than a promise about how it will be run.
One clarification, since it would otherwise be a half-truth: the four statements above are about the iPhone and iPad app. This website is a separate thing — a static site that does load Google Analytics — and the App Store page is served by Apple under Apple's own terms.
Why those four, and not a longer list
Most apps in this category are a thin client in front of a speech server. That arrangement decides almost everything downstream: if the voice is generated somewhere else, your text has to go somewhere else; if your text goes somewhere else, there has to be an account to attach it to; if there is an account and a GPU bill, there has to be a meter, and a plan, and a reason to watch how much you listen.
Sandbook is built the other way round, and the four statements above fall out of that single decision rather than being bolted on as policy. The whole Kokoro-82M model is in the app bundle: the download is about 276 MB, with the model and all 41 voices included. Speech is generated on the Apple silicon GPU through Apple MLX. There is nothing on the other end to meter, nobody to bill for characters, and no pipe for your reading to travel down.
This is also why the list stops at four. A promise that costs nothing to keep is worth nothing to make. Each of these four would require rewriting how the app works to break, which is a far better guarantee than a paragraph of intent.
What it deliberately does not do
The same decision that produces the guarantees produces the gaps. These are not oversights waiting to be fixed in the next release; they are the other side of running a real neural model on a phone, and you should know about them before you install anything.
Real-time reading pauses when the screen locks
The model holds real GPU memory while it is speaking, and iOS does not let an app keep that kind of work running once it leaves the foreground. So live reading pauses rather than pretending otherwise. The workaround is built in: prepare a chapter's audio ahead of time, and that chapter then keeps playing with the screen locked, with 15-second skip and the chapter on your lock screen. Continuous listening across a whole book with the phone in a pocket, or while driving, is still better served by a cloud reader.
Six voice languages, not thirty
41 voices: 28 in US and UK English, plus Spanish (3), French (1), Italian (2), Brazilian Portuguese (3) and Hindi (4). The voice decides the language, so a Spanish book gets a Spanish voice, and each book remembers its own. The pronunciation clean-up for dates, prices and units applies to English text alone. Readers that run in a data centre, and some other on-device readers, cover far more languages.
It cannot read an image-only PDF
Printed pages scanned with the camera, and photos or screenshots you pick, are recognized on the device and read aloud. A scanned PDF file with no text layer is not: Sandbook does not run text recognition on PDF files yet. DOCX, MOBI and AZW3 files are not supported either.
No whole-book audio export
Text in the Reader tab exports as an M4A audio file through the share sheet. There is no way to turn a whole EPUB into an audiobook file; if that is what you came for, this is not the app.
One device, no sync
No iCloud sync, no handoff, no shared library, no sleep timer, bookmarks or notes. Your books and your position in each of them live on the device you imported them to. It is iPhone and iPad only, on iOS 18.2 or later, and the voices need an iPhone 12 or later or an iPad with an A14 or M1 chip or later. There is no Android, Windows or web version.
How to check all of this yourself
None of the claims on this page require trusting us, which is the point of making them structural. Three checks take about two minutes between them:
- Turn on airplane mode and read a whole chapter. Speech that is generated somewhere else stops when the connection does.
- Open the App Store listing and look for the In-App Purchases section. On Sandbook's listing there is not one, because there is nothing inside the app to buy.
- Read the policy. The privacy policy says the same thing this page does, in the language a policy has to use.
To judge the voices without installing anything, the same Kokoro model runs in a browser tab on a computer: type a sentence and listen.
There is also a plain-text fact sheet at /llms.txt, kept deliberately boring and dated, for anyone — or any assistant — that would rather read the specification than the marketing.
The app is the argument
Everything on this page is checkable in about two minutes with the app installed and the phone in airplane mode. Sandbook is free, has no account and no in-app purchase, and starts speaking the first time you open a book.
iOS 18.2 or later · iPhone 12 or later & iPad