Android · Coming soon
A document scanner that keeps your documents. It finds the page, flattens it, reads the text on your phone so you can search inside your scans, and writes a PDF out of the photographs it already took. Nothing is uploaded, because there is nowhere for it to go.
Paper Lens is in internal testing. It is a signed release build on real phones, it is scanning real documents, and the feedback from that is still being folded back in. It is not on Google Play yet.
There is no release date, and there will not be one until it deserves one. Shipping the date and then the app is how software gets worse. If you would like an email on the day it lands, ask for one — that is the entire mailing list, and it will be used for exactly that.
It looks for the page in every frame while you hold the phone, draws the edges it has found, and takes the photograph itself once the page is steady and square enough. If you would rather decide that yourself, switch to manual — the detection keeps running, so the crop is still waiting for you afterwards.
The perspective is flattened, and every corner can be dragged with a magnifier under your fingertip so you can see what is beneath it. If dragging is difficult, each corner can be nudged a step at a time instead. Filters, brightness and contrast are one control away, and nothing is ever written over your original photograph.
Scans are ordinary JPEGs in a folder you chose, described by a plain text file beside them. Open them with anything. Copy them to a computer. Uninstall Paper Lens and they are still there, still readable, still yours.
There is no cloud folder, no team workspace, no subscription tier and no document assistant. What is here is what a scanner is for.
Live edge detection, an automatic shutter when the page is steady, and a manual mode when you want the decision.
Four corners, a magnifier under your finger, and a nudge control for anyone who cannot drag precisely.
Colour, greyscale, black-and-white and the untouched original, plus brightness and contrast. Apply one to every page at once, or reset back to the photograph.
Add pages, drag them into order, retake a bad one or delete it — without disturbing any of the others.
The text of every page is read on the device and indexed, so a search finds words that appear in no title, and shows you which page they are on.
Pages are embedded exactly as captured. A five-page export measured a few hundred bytes per page beyond the photographs themselves — no generation of quality is lost.
Export as numbered image files, or hand a copy to another app through the system share sheet without it ever touching your scans folder.
Every scan and its description in a single zip file you control, and a restore that puts the whole library back — titles, page order and crops included.
That is a statement about the binary, not an intention. Android requires an app to declare the internet permission before it can open a connection, and Paper Lens does not declare it.
A scanner is mostly invisible engineering. These are the parts worth telling you about, because each of them is a reason the app behaves the way it does.
There is no computer-vision library in the app. The pipeline that finds the page is Kotlin, and every change to it is measured against a corpus of real captures. A change that lowers the median overlap with the true page is rejected, however good it looks on the one desk it was tested on.
Detection runs on the camera preview, and phones vary far more than any tier list admits. Rather than sorting devices into fast and slow, it watches its own frame times and spends exactly what the hardware in your hand can afford.
Every library that would have saved the work decodes your image and encodes it again, which costs a generation of quality on every export. Paper Lens embeds the JPEG it already has, byte for byte. A five-page export came out a few hundred bytes per page larger than the images inside it, and every embedded stream is identical to its source.
Recognition runs after the crop, so the engine is never asked to read words off the desk behind your document. The text is kept twice on purpose: in a file beside your scan, so a restored library is searchable again without re-reading anything, and in an index, so searching answers as fast as you can type.
Everything needed to rebuild the library lives in your folder. Clear the app's data, reinstall it, point it back at the same place, and it reads the folder and returns your scans with their titles, their page order and their crops. This was tested by deleting everything and getting it all back.
The crop canvas is drawn rather than assembled out of ordinary controls, which made it invisible to a screen reader — a scanner nobody could crop is only finished for some people. It now describes itself and offers a way to select and move each corner, and the release checklist includes a screen-reader walk, 48 dp targets, measured contrast and a pass at 200% text.
Everything worked in debug and then the optimised build crashed as soon as it recognised text — and, because the app retries unfinished work at launch, it then crashed at every launch after. It was diagnosed by retracing the phone's own crash log against the build, and fixed with rules narrow enough to keep the size gain. This is why "verified on a device" is the last line of the checklist, and not "compiles".
No advertising, no subscription, and no feature held back behind one. If it turns out to be useful to you, there will be a way to say so.
06 — Contact
Support for Paper Lens is one person and one address. Bug reports, accessibility problems and arguments about features are all welcome.
The support page lists what to include so a problem can be fixed on the first attempt, and answers the questions that come up most often.
Please do not attach a scanned document unless it has been asked for. Almost every bug can be found without one.
