Narra

The problem is finding your place, not playing audio

Switching between a paper book and its audiobook means hunting through a timeline for a passage you can already see on the page. Chapter markers are too coarse to help: a chapter can be forty minutes long, and the sentence you stopped at is somewhere inside it.

Narra treats the page itself as the bookmark. Point the camera at the page you are on, tap a word, and the app resolves that word to a position in the recording and opens the reading view there.

• Scan a physical or e-ink page with the camera.
• Tap any word to choose the exact resume point.
• Playback opens at the matched passage, not the chapter start.
• Books download as complete packages and play offline.

Why matching happens against a transcript

Each book is imported as audio plus a transcript, and the transcript is what makes a word on paper addressable. Scanning produces candidate text, that text is matched against the transcript index, and the match resolves to a timestamp in the recording.

The work is done on a downloaded package, so a book keeps working on a plane or a train. Audio and transcripts are stored as private objects; the database keeps only their identities, sizes, and checksums.

What is available today

The backend, import pipeline, and client reading flow exist, and the clip on this page is a recording of that working flow rather than a mockup. The repository is private and there is no public build to install yet.

The launch list decides who hears first when there is something to install. It also indicates whether the first release should prioritise the iPhone scanning flow or the desktop reader.