// bring_it_in
Pick a photo, scan a page, share a file from another app, drop it on a chat, or type @ and its name. Each one becomes a document on your device, and its text is read there.
// photos
On iPhone and iPad, Photos in the attach menu opens Apple's photo picker. It runs outside Mortise and hands over only the photos you choose, so Mortise never asks for access to your photo library. Pick up to ten at a time. Each becomes its own document, named after the photo: IMG_1234 arrives as IMG_1234.pdf. PNG, JPEG, HEIC and HEIF are read.
// scan_document
Scan document opens the same camera Notes and Files use: it finds the page's edges, flattens it, and lets you add as many pages as the document has. The pages become one document, named for when you scanned it (Scan 2026-11-11 at 10.42.pdf). It asks for the camera the first time. If camera access is off, the composer says so and offers Allow camera, which opens the setting. Scan document appears only where the document camera exists, so not on the Mac.
// one_page
Before anything is read, the image is turned upright using the orientation the camera recorded, held to 4,096 pixels on its long edge, and laid on white, so a transparent screenshot doesn't come out black. It is saved as a JPEG with none of the original's metadata: no location, no camera details, no capture date. That copy, not the original, goes into your library and to any device you sync with. The JPEG is then wrapped as a one-page PDF with no text layer, so it goes down exactly the path a scanned PDF does. The same photo always makes the same file, so adding it twice gives you the document you already have, not a second copy.
Recognition is Apple's on-device text recognition, in English, at its accurate setting, with the page drawn at 2,048 pixels on its long edge.
// reading_order
Text recognition returns lines, not an order. Sorting them top to bottom interleaves anything with two columns, so a Settings screenshot would read a sidebar item, then a serial number, then its label. Mortise splits the page at a gutter instead, but only when the text on both sides runs alongside: a sidebar beside a pane is read one side, then the other. A label with its value at the far edge stays on one line, a table whose cells share rows stays in rows, and a chat screenshot stays in the order the messages were sent. A title across two columns is read first, then each column below it.
A page scanned or photographed sideways is turned upright before it is ordered, so it reads top to bottom instead of bottom to top. Mortise goes by the way most of the page's text runs, so one sideways caption doesn't turn the page.
// share_sheet
On iPhone and iPad, Mortise appears in the Share sheet for PDF, Word (.docx), PowerPoint (.pptx), Excel (.xlsx, .xlsm), CSV and TSV files, and for HEIC, HEIF, JPEG and PNG images. It does not appear for links or plain text: nothing is fetched from a share. Pages, Keynote and Numbers files are not offered, since Mortise can't read them; those apps can send a Word, PowerPoint, Excel or PDF copy instead.
The share itself loads no model and touches no network. It copies the files into a folder only Mortise can open, says Added to Mortise, and closes. A share arrives whole or not at all, and each file can be up to 200 MB. Mortise reads the folder the next time it opens or comes to the front, or straight away when it is already open beside the other app on iPad. A notice names what arrived, with Attach to put it on the open chat, and says when a file is already in your library. Sharing a PDF from Safari's Options → PDF also sends a blank page; Mortise leaves that page out.
// drop_and_mention
Drag files from Finder, Files or another app onto a chat, on the Mac and on iPad, and they are attached to it. On the Mac, images come in this way or through New file. A file Mortise can't read is turned away with the reason, and for a spreadsheet format it doesn't read, how to re-save it.
Type @ at the start of a word and a list opens above the message box: knowledge bases first, then files already in your library, up to eight of each. A name that starts with what you typed comes before one with a word that does, and that comes before a match in the middle. Spaces count, because file names have them. Choosing one attaches it and takes the @ text out of your message. Knowledge bases are listed only when the model in use can search one; otherwise the list says to pick a model that can.
// honest_limits
Mortise reads the text in a photo. It does not describe what the picture shows, and a photo with no words in it adds nothing to search. Recognition is English. Turning off Settings → “Read scanned PDFs” turns it off for photos and scans too.
A table in a photo or scan is read as rows of text, not as a spreadsheet the model can query. For figures you need exact, attach the Excel, CSV or TSV file itself.
Several photos picked together become several documents. To make one document of several pages, use Scan document. A sideways scan is read upright, but its preview still shows it the way it was taken. On the Mac there is no photo picker, document camera or Share sheet entry: drop the file, or use New file.
Once it is in, a photo or scan is a document like any other. How it is searched and cited is on how documents work.
// buy_once
No subscription, no account, no server of ours. Every update included.