Meaning and layout are separate deliverables
Preserving meaning is easier than preserving geometry. Keep the source and choose reconstruction only when the deliverable needs it.
Choose the kind of preservation you need
Source requirement
The source image plus text boxes or table cells that can grow.
Production decision
Translation changes text length, so preserving meaning does not preserve geometry.
Final inspection
Use side-by-side output when review matters more than a visual overlay.
What OCR layout means
OCR may detect lines, blocks, tables, and reading order, but plain extracted text cannot preserve exact coordinates, typefaces, or spacing. A visually convincing overlay needs bounding boxes and a rendering stage in addition to recognized text.
Tables and multi-column pages need structural output. If cells or columns are flattened into one sequence, translation cannot reliably put them back.
Translation changes geometry
The same meaning can require more or fewer characters in another language. Line breaks move, headings wrap, and narrow buttons may no longer fit. Vertical scripts and right-to-left scripts can also require a different layout direction.
For editable reconstruction, treat every text block as a flexible region and allow it to grow. For a faithful archive, preserve the original image and attach the translation separately.
Before rebuilding a design, use the translation text expansion estimator to identify containers that may need more room.
Choose the right deliverable
Use side-by-side source and translation when auditability matters. Use a translated overlay for quick visual orientation, not as proof that every label stayed connected to the correct object. Use desktop publishing or specialist document software when the result must be printed or delivered professionally.
Always retain the source image. It is the evidence needed to resolve a misplaced caption, number, footnote, or table cell.
Understand which parts of a photo layout OCR can retain, why translated text changes length, and when document reconstruction is required. Continue with the image translation guide collection for adjacent workflows.