The disagreement is about where reading starts
Most reader apps start from the file. You pick a format, then a file, then a reading mode. OmniShelf starts from the reader: you pick up the book you were already in. Everything else in the app follows from that one reversal, including the parts that are inconvenient.
Format is a property of a file, not of a reading life
A shelf that separates PDF from EPUB is a shelf organised by the people who built it, not by the person using it. Real reading is mixed. A week might contain a scanned paper, a purchased novel, a public-domain text pulled from a catalogue, and a comic someone sent you. Those are not four activities. They are one evening.
Putting them together is easy to draw and hard to build. A PDF has fixed pages: the page is the unit, and page 214 means the same rectangle of paper on every device. An EPUB has no pages at all until the reading system decides on a screen size, a font size and a margin, and then manufactures them. A shelf that holds both has to store two different notions of “where I stopped” and present them as one progress indicator that does not lie.
We chose to keep both models underneath and unify only what the reader sees: the cover, the title, the position, the way back in. The cost is that some operations are deliberately format-specific — text reflow exists for PDFs and is meaningless for EPUBs — and the interface has to admit that rather than pretend otherwise.
What we removed is the argument
OmniShelf has no advertising, no in-app purchases, no subscription, no account and no analytics. This is usually presented as a virtue. It is better understood as a constraint we were willing to accept.
Advertising is not a rectangle at the bottom of a screen. It is a permanent reduction of the reading area, paid for on every page of every book, and it introduces an incentive that runs against the reader: the longer you stay on a screen, the more inventory is sold, and the most reliable way to keep someone on a screen is not to finish the chapter. A purchase prompt is worse in a different way — it turns a missing feature into a sales moment.
Removing them also removed a class of bug that is hard to see from the outside: code paths that only exist to decide whether to show an ad, and the state machines they bring with them. What is left is smaller, and it is auditable. You can read the whole application’s source and confirm that nothing is sent anywhere.
Offline is the default, not a mode
A reader that requires a connection to open a file you already own is not a reader. OmniShelf reads from local storage, indexes local folders, and renders without a network. Online catalogues (OPDS) exist, but they are a way to find books, not a way to read them: once a file is downloaded it behaves exactly like a file that was always there.
This is why the app asks for broad file access, and it is the only honest justification for it. A library has to be found where it already lives — in Downloads, on removable storage, in a directory a desktop machine syncs — without asking the reader to re-authorise each folder through a system picker on every launch. If a future version can find books with narrower access, the permission should be removed rather than defended.
Typography is not decoration
Reading speed and comprehension change with line length, leading and hyphenation. A reader that ships one font and one margin is asking every reader to adapt to the defaults of whoever wrote the app. OmniShelf exposes margins, line spacing, alignment, hyphenation and font choice, and it ships a serif family as the default because that is what long-form text is set in nearly everywhere else.
The defaults matter more than the controls. Most people never open the settings panel, so the first page has to be readable as shipped: a measure that does not run the full width of a phone, enough leading that lines do not collide, and hyphenation off by default because automatic hyphenation in a language it does not understand produces worse text than ragged right.
What the app will not do
It will not ask you to create an account. It will not sync your library to a server we operate, because we do not operate one. It will not show you a recommendation feed, because a recommendation feed is an advertising product with a reading interface attached.
It will also not pretend that every format is equally good. A comic rendered as a reflowable document is a worse comic; a scanned PDF forced into a single column is a worse paper. Where a format has a native way to be read, that is the default, and the alternatives are offered rather than imposed.
Who this is for
Someone with a folder of books they did not buy from a single store. Someone who reads a paper in the morning and a novel at night and does not want to think about which application to open. Someone who has noticed that the reading experience degrades a little every year as the business model behind the app becomes more important than the reading.
If that is you, the app is free and the source is public. If it is not, there are excellent readers that will sell you a subscription and recommend books to you; they are just solving a different problem.