A private document scanner for Android. Paper to clean, shareable PDFs in a few taps, with everything happening on the phone: no account, no ads, no tracking, no watermark of its own.
Licensed under the GNU GPL v3. Third-party components are listed in
licenses/THIRD_PARTY.md. Product goals live in
PRODUCT.md and the full plan, including the clean-room rules every contributor
follows, in PLAN.md.
- Scan: live page detection (with a small learned page finder for white pages on pale or busy surfaces) with auto-capture, a glare warning, batch scanning, crop with a magnifier, and the looks Original, Auto, Bright, Gray, Ink and Board, with shadow and grain removal and neutral paper under tinted light; old photos have fading, dust, scratches and torn corners mended. Curved pages are flattened, lines straightened and text squeezed toward a book's spine widened back out; photos of screens can have their moiré patterns reduced. Modes for ID cards (printed at true size), books (split at the spine), whiteboards, and QR codes or barcodes.
- Text: on-device OCR (PP-OCR models for English, with doubtful words corrected from word statistics; Tesseract for other languages), full-text search, searchable PDFs, copy, and export to plain text, a spreadsheet, or a Word file that keeps the page's headings, paragraphs and pictures; documents can also be shared as PowerPoint slides, and a portrait made into a passport or ID photo with a sheet of copies to print.
- Translation on the phone: Text → Translate puts a document's text into another of 58 languages with Mozilla's Firefox Translations models (each language downloaded once, about 20 to 70 MB); the text itself never leaves the phone. The page's picture shows the translation in place of its own words, in the same layout, to read full screen or share, and a translation is kept with the document to show again next time.
- Documents: folders (renamed or deleted with a long press; All groups documents by folder), tags and notes, Trash, merge, reorder/rotate/copy out pages, sign, draw, type text and stamps, erase, remove handwriting (found pixel by pixel by a small on-device network) or fingers, black out private details (found automatically or by hand, and removed from the text layer too), password-protected PDFs (AES-256), page and file size options, watermark, App lock. Edits can be applied to other pages at once. Test papers: split into one question a page, mark the ones answered wrong and gather them to practise again.
- PDFs and Office files you already have: import a PDF, Word document, Excel workbook or PowerPoint deck (or open or share one to Oscanner) and its pages become a document to sign, mark up, black out, clean and merge like a scan. An Office file's text is kept exactly, and a PDF's too on Android 15 and later, instead of being read again.
- Everywhere: share sheet in and out as PDF, a smaller PDF, separate pictures or one long picture; save to any file provider, print, a Scan shortcut, Quick Settings tile and widget. Send to a computer on the same Wi-Fi: it downloads the PDF in a browser, no account or cable.
- Your defaults from the start: after the welcome tour, four short steps set the theme, the look and fixes for new scans, which way up pages are taken, PDF page size, file size and watermark, reading text, the app lock and library order. Settings → Set up defaults runs it again.
What's next, measured against CamScanner, is in ROADMAP.md.
Download the APK for your phone from the latest release
(arm64-v8a for almost every phone from the last few years) and open it to install.
Requirements: Android Studio (or its bundled JDK 17+) and the Android SDK with platform 36.
./gradlew assembleDebug # app/build/outputs/apk/debug/
./gradlew testDebugUnitTest # unit tests
./gradlew assembleRelease # per-ABI release APKs (unsigned)Release builds are split per CPU type; the arm64 APK is about 18 MB.
app/ navigation, theme host, app lock, shortcut/tile/widget
core/model plain data types (documents, pages, edit recipes)
core/data Room database, page files, rendering queue, OCR jobs, export
core/imaging OpenCV: detection, straightening, looks, clean-up, book split
core/ocr PP-OCR reader and word fixer, Tesseract wrapper, language downloads
core/pdf dependency-free PDF writer (text layer, AES-256) and DOCX writer
core/design Soft Card, Night design system: theme, type, components
core/translate on-device translation: Mozilla's bergamot engine (WebAssembly) in a hidden WebView
feature/* library, capture, review, document, text, settings
Pages are never destroyed: each page keeps its original photo plus an edit recipe (crop, look, sliders, marks, clean-up), and what you see is a render that can always be rebuilt.
- Model-based filling of removed marks (LaMa inpainting)
- App-level encryption at rest (Android's file-based encryption protects data meanwhile)
- Source builds of OpenCV, Tesseract and ONNX Runtime for F-Droid; a baseline profile for faster cold start
- A freedom-to-operate review of the areas flagged ⚠ in
PLAN.mdbefore any public release