OmniScanner: AI Scan & Redact

OmniScanner puts a privacy checkpoint between paper and Share. Capture or import a page. One on-device model proposes masks using both the page image and Apple Vision text recognition. You inspect the actual boxes. When you approve them, OmniScanner rebuilds a safe copy without the hidden pixels, searchable words, or covered barcodes. A PRIVACY WORKFLOW, NOT A LAST-MINUTE TOOL 1. Bring in paper from the camera, Photos, Files, or a network scanner. 2. Review the private areas already marked on the page. 3. Move, resize, delete, or add any box. 4. Apply only what you approve. 5. At export, Privacy Check uses the same model before the file leaves the app. ONE MODEL LOOKS TWO WAYS The visual path finds page regions that are not words at all, including a face photograph or signature panel. The multilingual text path reasons over Apple Vision OCR, its layout, and its confidence. Their results become one set of editable suggestions rather than two competing modes or a list of patterns you must configure. WHAT IT CAN MARK • Names, addresses, email addresses and phone numbers • Account, payment and government identifiers • Dates of birth, passwords, network identifiers and other sensitive text • Signatures, face photographs and fields on identity documents MEASURED BEFORE RELEASE The shipping Core ML threshold was selected on 1,692 validation documents under a clean-document false-positive budget, then evaluated once on a frozen 793-document test set. It found 89.17% of labelled text entities, flagged 2.86% of clean documents, and left 72.67% of positive documents with every labelled entity found. A separate 2,000-page visual evaluation measured 95.18% maximum recall. These are test results, not a guarantee. The model can miss something important or mark something ordinary. Every suggestion is editable, nothing is applied until you approve it, and the last review is yours. REDACTION THAT REMOVES CONTENT • Approved boxes are burned into the page; they are not removable overlays • Covered words leave the searchable PDF text layer • Covered barcodes and QR codes stop decoding • The untouched original remains on your device, so the edit stays reversible inside OmniScanner MORE THAN ONE PAGE On Mac, place a whole folder of PDFs on one canvas. Every page is inspected once, all calibrated levels are ready without rerunning the model, and the selected masks can be applied across the batch. THE DOCUMENT TOOLS UNDERNEATH • Edge detection, deskew, crop, rotation, enhancement, brightness and contrast • Searchable OCR, folders, tags and smart folders • Searchable or password-protected PDF, image, long-image, Word, PowerPoint and text export • Fill and flatten forms, annotate PDFs, reorder or extract pages • Sign a PDF with your own certificate • Optional iCloud sync across iPhone, iPad and Mac PRIVATE BY CONSTRUCTION The redaction model is inside the app and runs offline. No account, analytics, advertising, or third-party tracking SDK is required. Optional iCloud sync uses your private database. BASIC AND PRO Scanning, OCR, organization, PDF editing and searchable PDF export work without a subscription. Automatic redaction covers 100 pages each month free, and hand-drawn boxes are never limited. Pro makes automatic redaction unlimited and adds Fill & Sign, password protection, iCloud sync, smart folders, advanced conversion, long-image export and certificate signing. Terms of Use: https://www.apple.com/legal/internet-services/itunes/dev/stdeula/ Privacy Policy: https://omniscanner.echo.cool/privacy/