PixelOkay
Files stay on device
Tools
Tools›Document Layout Reconstructor
DOCUMENT INTELLIGENCE SCAN · STRUCTURE · EDIT

Document Layout Reconstructor

Find words, rebuild text and tables, capture illustrations, then refine everything in an editable page workspace. No paid API or document uploads.

✦ Browser-side OCR✦ Editable layout blocks✦ Word & HTML export
YOUR WORKSPACE

Reconstruct a document ✳

Import, recognize, inspect and refine — all in your browser.

01

Source document

⇧Drop a scanned documentor click to browse your filesPDF, PNG, JPEG, WebP · 24 MB max each · up to 12 pages
No document loaded.
02

Recognition

OCR downloads free engine/language files only when you analyze.

Recognition estimates positions. Correct text, tables and reading order before publishing.

03

Edit the structure

Draw on the page to create regions. Drag existing regions to move them.

LIVE DOCUMENT CANVASDocument workspace
▤

Start with a document

Import a scan or use the sample to explore text recognition, table reconstruction and editable regions.

No pages
04

Selected element

None
Select a text block or table to edit its contents, or draw a new region on the page.
DELIVERABLES

Keep it editable.

Export document content, not just a screenshot.

LOCAL OUTPUT

DOCX preserves text, tables and captured image regions in reading order; HTML keeps approximate page coordinates. Exact typography and complex layouts require review.

BUILT FOR PRACTICAL WORKFLOWS

Scans are the starting point.
Your edits finish the job.

⌗

OCR + PDF text

Use a PDF's text layer when possible or recognize scanned pages with open-source Tesseract OCR running locally.

▦

Editable building blocks

Correct detected paragraphs, convert repeated columns to tables, and manually mark images or diagrams as captured regions.

↗

Useful exports

Create Word documents, position-aware editable HTML, text files and portable JSON projects without uploading originals.

About reconstruction

Does PixelOkay reconstruct complex documents perfectly?

No. Browser OCR and layout heuristics provide an editable first draft, but multi-column documents, merged tables, handwriting and unusual typefaces need review. Scanned images require user-marked image regions.

Does my document get uploaded to a server?

No. Document images are processed on your device. OCR and PDF libraries, and OCR language data, may download from public CDNs when first needed. Your document contents are not sent to those services by this tool.

Why might exported Word layouts differ?

Word reflows paragraphs and tables. The HTML export retains approximate absolute page positions, while Word exports prioritize editable content and reading order rather than pixel-perfect positioning.

What file types can I open?

PDF, PNG, JPEG and WebP sources are supported within the tool's page, dimension and file size limits. A saved PixelOkay JSON project may also be reopened.

Using Document Layout Reconstructor

Import supported images or a PDF and choose recognition settings. Review text/table blocks, select image crops manually and correct positions or reading order. Save a JSON project before exporting DOCX, HTML or text.

PDF text geometry and table detection are estimates; figures require manual selection. Earlier/Later changes export reading order rather than physical position. DOCX reflows paragraphs/tables and is not a pixel-perfect copy; positioned HTML also depends on fonts/browser rendering. PDF.js/Tesseract and OCR data download externally. Resource budgets apply and there is no autosave.

References: PDF.js · Tesseract.js

Rebuild a document layout with editable, reviewable regions

Import a supported PDF, inspect extracted text and OCR regions, then arrange text, tables and figures into an editable reconstruction. The tool estimates layout and leaves room for manual review rather than promising a pixel-identical conversion.

Example workflow

  1. Import a PDF and check whether text extraction or OCR is needed for its pages.
  2. Review detected text and table regions, correct content, and adjust reading order with the available Earlier/Later controls.
  3. Export DOCX or positioned HTML, then inspect the result in its target application and save project JSON if you need to reopen the work.

Authored example: Reconstruct a one-page report with a heading, two paragraphs, a small table and a logo. Check reading order and table cell content individually before comparing the reflowed DOCX with the source.

Limits and processing: PDF.js and Tesseract.js plus OCR data download from external CDNs. Text/table regions are heuristic; manual figures need review. DOCX reflows and does not preserve exact source positions; positioned HTML depends on fonts and browser rendering. Projects are not autosaved.

Document Layout Reconstructor questions

Will the DOCX keep the PDF's exact page layout?

No. DOCX output reflows in the target editor and does not preserve exact source positions.

Are detected tables and figures guaranteed to be correct?

No. Detection is heuristic and figures are manual; review the reconstructed content before relying on it.

Related tools