# English image-only PDF: Word and searchable PDF outputs

2026-10-14. The source is the self-authored two-page English Willow supplies record, rasterized locally at180DPI into twoJPEGs. The inputPDF has zero extracted characters. It is a digitally generated scan simulation, not a physical scanner or phone photo.

Actual Doc2X OCR of this two-page input produced the shared result. No manual OCR text edits or image enhancement were performed. The published DOCX and searchablePDF are unchanged native downloads with public filenames.

Reflow DOCX:2native tables,each4x6;0textboxes;0embedded images. IDs001–006,blank003quantity/amount,zero values,-5.00 and total44.60 remain. Source footer sentences are absent. Local LibreOfficeDev26.8 preview is1page; not a Word/WPS interactive edit test.

SearchablePDF:2pages,470001→482460bytes. Page boxes and embedded JPEG bytes match the input. Both pages rendered pixel-identically at2x with PDFium(1191x1684perpage). This is a fixed sample result,not a guarantee for all files/readers.
PDFium searches input→output:Willow0→4,33.600→2,44.600→1,Returned gloves0→1. Independent text extraction includes the source footers and table values; selection order/cell semantics still need checking for other files.

Actual Chrome native Find:44.60 shows1/1 and openspage2. Actual text-layer selection copied via native Command-C and pasted via native Command-V into a local textarea yielded exactly“Recorded total: 44.60”. Browser virtual clipboard calls did not supply this result; the successful test used native system copy/paste. Only that line was paste-verified,not the whole table.

The input/output PDF appearance is unchanged. Screenshots show actual export controls and selected PDF text; no generated UI. JPG screenshots are re-encoded from originals. The local preview/notes/check files and proofZIP are assembled locally. An extra preservedWord export restores the original persisted default; it is not one of the two compared outputs.