How it works

Four stages, all on your device.

A capture runs through a pipeline of small, testable steps. Every one of them runs locally.

  1. Capture

    Scan with the camera (multi-page, auto-cropped), or import from Photos and Files. Images and PDFs are accepted.

  2. Extract

    Vision OCR reads images. PDFs with real text are read natively; image-only pages fall back to OCR.

  3. Structure

    An on-device intelligence provider infers headings, paragraphs, lists, task lists, tables, code blocks and quotes. When Apple Intelligence is available it refines the result; otherwise a deterministic built-in converter does the work. The result screen tells you which one ran.

  4. Render and export

    The structure becomes GitHub-flavoured Markdown. Review it in Preview, Markdown and Original tabs, edit by hand, then copy, share or save.

Conversion modes

Automatic is the default and detects structure on its own. The other modes bias the structuring toward a kind of source:

Editing and reprocessing

Tap Edit on the Markdown tab to fix anything OCR or the model got wrong; Done commits the change to the preview and every export. Edited captures are marked, and "Revert to original" puts them back. Reprocess re-runs the pipeline in another mode without touching the source image, it asks first if it would discard an edit.

Share or save

Copy the Markdown, share the .md file (AirDrop included), or save to Files and iCloud. The Share or save sheet also offers the page you captured: a single scan or imported PDF goes out as the stored file; a multi-page scan is combined into one PDF, or sent as separate images if you untick that. The stored original is never modified.

History

The three most recent captures sit on Home; Captures holds the rest, grouped by day, with a thumbnail of the real page. Metadata and Markdown live in a local database; originals are stored as files in the app's container and kept until you delete the capture.