Capture
Preserve the original page and record the best source available.
Document reconstruction
A scanner captures an image of a page. Rebuild aims to recreate the document itself.
The goal is a genuine digital document: selectable text, clean lines, meaningful structure, editable output and a clear link back to the source.
Source first
The original remains available. Reconstruction should never quietly replace evidence with a guess.
The reconstruction path
Four stages, with validation between them.
Understand
Identify text, layout, page regions and the kind of document.
Rebuild
Create real digital text, lines and structure where confidence is sufficient.
Export
Produce a useful PDF or bounded editable output with provenance and review.
What works today
The current iOS beta already performs local reconstruction.
It uses on-device text recognition, reconstructs supported text and layout into PDFs, retains confidence and processing metadata, and preserves a page when safe digital replacement cannot be proven.
Current outputs include searchable PDFs and bounded Word export modes. Results still require review, especially for handwriting, complex tables, unusual type and specialist notation.
What is not public on the web yet
Reconstruction is not a fake upload box on this website.
The existing development web pipeline requires a local Node and Python worker. It is not compatible with this static Cloudflare Pages release, so the public website does not claim that online Rebuild processing is available.
A future web release needs a deliberate privacy, retention, deletion and specialist-engine architecture before document upload is enabled.
Learn how scanned-text editing worksUse what is available now
Start with the local web tools.
Merge, organise, compress and convert without sending your file away.