PDF Lab / Original specimens

OBSERVED · REPRODUCIBLE · LIMITED

One page, three different kinds of text

We authored a page as real text, a raster scan and a raster with synthetic hidden OCR. Similar appearance does not imply the same search or editing behavior.

Start with the problem

Symptom: Letters are visible, but you cannot select them, or copied text differs from the page.
Main cause: Text commands, scanned pixels and hidden OCR text behind an image are different data. Changing hidden text alone does not change the visible image.
What to do now: Search and copy the same date from the three PDFs below and compare it with the page. Then choose by your purpose and whether you have the source document. Adding an explanation, extracting searchable text and changing visible wording are different tasks. PDF Desk provides no OCR.

The visible date can match while extracted text differs

Compare real text, an image-only scan and a scan with one synthetic hidden date line. The OCR simulation is not the output of a recognition service.

Page representation	Extracted characters	Date extracted by independent parser
Original with real text	119	Delivery date: 2026-10-22
Image-only scan	0	No date extracted
Synthetic hidden-OCR date	25	Delivery date: 2026-10-22

What changed

The extraction counts differ because the stored text differs. The OCR simulation stores the visible date as pixels and the searchable date as separate hidden text.

Why it matters

Search or copy success does not prove that the visible wording is editable. Changing only hidden text can leave the date in the image unchanged.

What to do

Compare copied or searched text with the visible wording, then choose below by your purpose and access to the source. Editing the hidden date does not establish a change to the visible date.

Compare full pages and open the original PDFs

Real text · the date is stored as text. The independent parser extracts the visible date and other wording. — **Real text · the date is stored as text**
The independent parser extracts the visible date and other wording.

Scanned image · the letters are pixels. The date is visible, but this PDF has no extractable text commands. — **Scanned image · the letters are pixels**
The date is visible, but this PDF has no extractable text commands.

Choose by your goal and access to the source

You have the source and need to change visible wording

Edit a copy of the originating document and export a new PDF. Adding text over a scan is not replacement of the original wording.
Reopen the new PDF and check the changed phrase, line breaks and page count, both visually and through search and copy. Exporting from the source does not guarantee identical fonts or structures.

You only have the PDF and need search or reusable text

Try selecting and copying one sentence. If real text matches the page, extract it and check numbers, tables and reading order against the original.
An image-only page needs OCR. Compare recognized text line by line with the image and correct misread numbers or names. Recognition does not by itself edit visible scanned pixels or preserve the original layout.
Recheck existing hidden OCR text too. This experiment uses a manually inserted hidden date; it does not measure an OCR product’s recognition accuracy.

You only have the PDF and must change the visible scan

Ask the author for an editable source if possible. Otherwise, recreate the necessary content in a new document and review it as a reconstruction, distinct from the original.
Even when OCR supplies text, the visible wording, fonts and layout may need rebuilding. A change to hidden text alone does not establish a change to the image.

You want to keep the original and add an explanation

Read the PDF editing compatibility guide, then use Open PDF → New text box in the workbench. Do not convert blocked documents to bypass structural checks.
Place the explanation in unused space. Review → Download checked PDF, then reopen it. This adds separate explanatory text; it is not OCR or replacement of the original.

These choices help identify the task. We have not tested or recommended a particular external OCR or reconstruction product, version, price or upload behavior. Check its official processing and cost information before use. The observed experiment here compares display and extraction in our own PDFs; it contains no successful external OCR run.

Problem and conditions

When text cannot be selected, or copied text differs from the page, inspect its representation. text-source contains text operators and artwork. scan-only is a rendered image of that page. scan-ocr adds only one invisible date line as an OCR simulation. No OCR service was used.

Observed differences

The independent parser extracts the date from real text and no text from the scan. It extracts just the date line from the OCR simulation. The visible date in that last file remains an image: changing only the hidden text would not update the page. PDF Desk does not support replacing that hidden original text.

Check the difference

Search for Delivery date and select and copy the date in all three files.
Zoom into the scan to see that its letters are also image pixels.
Compare the workbench original-text lists. The scan is not presented as editable text.
Check extracted numbers and punctuation against the visible page.

Verification scope and official explanation

Searchability does not prove accessibility tags, correct reading order or accurate OCR. The official pypdf guide distinguishes recognition from extraction of existing text. Our files demonstrate that distinction; they do not verify successful source reconstruction or external OCR processing.

Files to inspect yourself

Independent observations and method (JSON) · Authorship and file hashes

Crop coordinates, scale, source hashes and fresh extraction · Rechecked: 2026-10-24 · pdftoppm version 26.07.0 · pypdf 6.10.0 · pdfplumber 0.11.9. Original PDFs were not modified.

Environment, date and evidence boundaries

Checked: 2026-10-23 · Windows / Chrome 155 · pdf-lib 1.17.1 · PDF.js 6.4.299 · fontkit 1.1.1 · pypdf 6.10.0 · pdftoppm version 26.07.0. PDFs and artwork are original; NanumGothic uses OFL-1.1. Tables and files record direct observations; explanations interpret those structures. Physical OS keyboard input, phones and every PDF reader are outside this test.

Official reference explanations (separate from measurements): pypdf: text extraction and OCR

Reproduction code: scripts/lab-inputs.py, lab-outputs.mjs, lab-derived.py, tests/readiness-browser.mjs and tests/lab-pdf.py.

Choose your file with Open PDF and read any restriction. Supported PDFs can receive explanatory text; the workbench cannot perform OCR or replace lettering inside a scan.

How to add explanations to supported PDFs Open a PDF to check support

Keep this guide handy

Keep the guide address or share it with someone who needs these instructions. It contains no personal PDF or editable project.

Public guide address

One page, three different kinds of text

Start with the problem

The visible date can match while extracted text differs

What changed

Why it matters

What to do

Choose by your goal and access to the source

You have the source and need to change visible wording

You only have the PDF and need search or reusable text

You only have the PDF and must change the visible scan

You want to keep the original and add an explanation

Problem and conditions

Observed differences

Check the difference

Verification scope and official explanation

Files to inspect yourself

Environment, date and evidence boundaries

If you get stuck on this task

Keep this guide handy