PDF to Text

report-a4-3pages.pdf has three pages of real text. Page 1 reads "report-a4-3pages.pdf page 1", and pages 2 and 3 follow the same pattern. meta-card.pdf has one page and the words "Harbour report". pdf.js reads text operators. It does not OCR a picture. A page that was rebuilt as a JPEG of letters has no text layer, and this page will say so rather than invent words. A password stops the read. Nothing is uploaded. If you need the picture of a page, use PDF to JPG. If you need words from a scan, this page will not invent them.

Three report lines and the words Harbour report are real text operators. A JPEG picture of the same page would come back without those words. That empty result is correct. OCR is a different page and a different kind of mistake. A password stops the read. The PDF is not uploaded.

The report yields three lines, one per page, each naming that page. The card yields the words Harbour report. Those are text operators. A JPEG picture of the same page would come back without them, and that empty result is correct. OCR is the image page, a different job with different mistakes. A password stops the read. The PDF is not uploaded.

Page 1 of the report is the sentence report-a4-3pages.pdf page 1. Pages 2 and 3 follow. The card is the sentence Harbour report. That is the text layer pdf.js can see. A page rebuilt as a JPEG of those same letters has no text layer, and the line will say there is no text rather than guessing. OCR of a picture is the image-to-text page, with its own English limit and its own confidence number. A password stops this read. The PDF stays in the tab.

The report’s three pages each contain one sentence, and the card contains Harbour report. Copy those if you need the text layer. Do not expect the same words from a scanned page or from a JPEG rebuild. Those files have paint, not text operators. OCR is a separate page. A password ends this read immediately. The PDF never leaves the browser.

Work in this order. If the PDF was exported from a writer, expect real sentences. The report gives one sentence a page, naming itself. The card gives Harbour report. If the PDF is a scan or a JPEG rebuild, expect little or no text, and treat that as the truth. This page does not OCR. The image-to-text page is the OCR path, with its own limits. A password stops the extraction. You do not send the file out to be unlocked. Copy the text from the page, or download the text file, and leave the PDF where it was.

Info fields are on PDF metadata viewer. A picture of a page is PDF to JPG.

Three pages of the report

Inputs: p2t-f: report-a4-3pages.pdf

On this page, take report-a4-3pages.pdf and press Extract text. A picture of letters has no text layer.

Result: PDF text, 3 pages. Page 1: report-a4-3pages.pdf page 1 Page 2: report-a4-3pages.pdf page 2 Page 3: report-a4-3pages.pdf page 3 A page that is only a picture of letters has no text to extract.

The harbour card

Inputs: p2t-f: meta-card.pdf

On this page, take meta-card.pdf and press Extract text. A picture of letters has no text layer.

Result: PDF text, 1 page. Page 1: Harbour report A page that is only a picture of letters has no text to extract.

What is page 1 of the report?

report-a4-3pages.pdf page 1.

What is the card?

Harbour report, one page.

Will a scan produce words?

No. A picture of letters has no text operators.

Is this OCR?

No. OCR is the image-to-text page. This page reads a PDF text layer.

Password?

The read stops.

Uploaded?

No.