Scan, search, read, edit, summarize, and ask — every step on your device. The AI is Apple Intelligence on the Neural Engine; the search index is local; the only thing that ever leaves is an optional iCloud sync through your own Apple account.




screens captured on the iPhone 17 Pro simulator with seeded scans
The + button opens the document camera directly. Multi-page scans combine into one PDF.
A post-capture pass nudges contrast and gently deskews tilted pages. On by default.
High-contrast output for receipts, faint print, and signed documents.
Pull up to 20 existing photos — each gets edge-detect, perspective correction, and OCR, just like a fresh scan.
Pull PDFs or images from Files, iCloud Drive, Dropbox. PDFs come in as-is; images get auto-cropped and OCR'd.
Suggests a filename from the largest text on the first page, so a new scan lands as "Lease Agreement", not a timestamp.
Every new scan gets text recognition, embedded as an invisible PDF layer so it's searchable everywhere.
One-tap OCR backfill for PDFs created before OCR existed, or imported without a text layer.
Find any word across every scan — name and contents. Tap a result to jump to the matched page with the text highlighted.
Find a word inside one open document; press Return to jump to the next match.
An AI summary, key points, and a follow-up Q&A about any scan. Reads a column-preserving rendering of the page so receipts and forms don't scramble. Long-press to copy.
Ask in plain language and get an answer pulled from your whole library, with tappable citations back to the source page.
An on-device embedding index matches by meaning, not just keyword overlap — built incrementally and persisted, so unchanged docs aren't re-processed.
Customize the system and user prompts for Summary, Key Points, and Ask. A CUSTOM badge shows when you've diverged; reset one or all.
On devices without Apple Intelligence the AI surfaces hide cleanly with a one-time card. Everything else works the same.
Reorder by dragging, rotate, and delete individual pages. Saves back to the original PDF with the OCR layer intact.
Pen, eraser, undo, redo, clear — real ink colors and pinch-to-zoom for tight signatures. Re-runs OCR after save so search still works.
Drop a text box anywhere, resize, pick a size and anchor. One-tap presets insert today's date or your saved name for forms.
Re-apply Color, Black & White, or Maximum Punch to a scan you already saved — no need to rescan.
Group scans by project or topic. Scanning from inside a folder drops the new scan there.
Move one or many scans between folders or back to the root; multi-select to delete, share, or move at once.
Newest, oldest, or name A→Z / Z→A. Rename inline from a row or inside a document.
Soft delete with a 30-day restore window, then automatic permanent purge.
The standard iOS share sheet — Mail, Messages, AirDrop, Files, anywhere.
Encrypt a single doc with a password before sharing. Paperix never stores or transmits it — and it can't recover it.
AirPrint to any compatible printer on your network.
Pull the OCR'd text to the clipboard for pasting elsewhere.
After a scan, Paperix surfaces follow-ups found in the text — contacts, dates, addresses, links, receipt totals — to add to Contacts/Calendar, open in Maps/Safari, or copy. On-device; quiet when there's nothing.
Scan a receipt and Paperix reads the labeled grand total (skipping subtotal and tax). One tap copies the amount and currency.
Run the same detection on an older scan or imported PDF, with an explicit "nothing found" state when there's no follow-up.
No SDKs, no analytics, no telemetry — the app has no networking code of its own and never phones home. Verifiable with Little Snitch or Lulu.
Re-authenticate every time you reopen Paperix. Opt-in.
Off by default. Turn it on and scans sync through your own Apple iCloud — no Paperix account, no third-party server. The search index stays local per device.
Universal app with an adaptive iPad layout, plus a Siri Shortcut, a Quick Scan widget, and the paperix://scan URL scheme.
Every AI step is stamped “on device.” Summaries, key points, answers, embeddings, OCR — all generated on the Neural Engine. No document content, query, or model output is transmitted off the device, ever. See the privacy policy.