Paper to searchable text without leaving the phone
An Android scanner that crops, reads and exports on device, with no internet permission in the manifest.
What was in the way
Most phone scanners send the page to a server to read it. For a contract, a payslip or a letter from a clinic, that turns a two second chore into a question about where the document went and who kept a copy.
The honest answer for anyone handling other people's paperwork is that they cannot tell. The upload is invisible, and the privacy policy is the only evidence they get.
How it was built
Recognition runs on the handset through ML Kit, so the page never needs a network hop. The manifest asks for the camera and nothing else, no internet permission at all, and a network security config blocks cleartext in case a dependency ever reaches out.
The capture path reuses Google's document scanner, then assembles an A4 PDF that carries the recognised lines as an invisible text layer under the page image.
Room holds the scan record and the cached text. A background worker fills in text for older scans, so library search matches on body content instead of re-running recognition.
How it works
Crop and confirm before anything is saved
Edge detection and cropping happen in the capture flow. The page is kept or rescanned before a record exists.
Every scan stays re-readable
A saved scan keeps its page image and PDF, so text can be pulled again later without going back to the paper.
Text out in the formats people ask for
Recognised text copies, shares, or writes to Downloads as .txt. Word, Excel and CSV sit behind a one time unlock.