Build an accessibility-first reader that imports permitted documents and articles, reconstructs their reading order, and produces stable speech with sentence-level highlighting, bookmarks, rewind, skip, sleep timers, and persistent position. Begin with EPUB^?, PDF^?, pasted text, and share-sheet imports rather than promising unrestricted access to protected book libraries. Include screen-reader-tested controls and a correction interface for headings, page artifacts, pronunciation, and repeated text.

Natural speech generation has improved enough to make long listening sessions practical, while recent reviews still show that document parsing, navigation, synchronization, and accessibility remain unreliable. The recurring failures across unrelated products indicate a workflow gap rather than one application's temporary defect.