OCR PDF

What this reading returns

The server turns each page into a PNG at 200 dots per inch, then asks Tesseract for an English text layer. The language flag is eng. There is no language menu. pdftoppm runs once, with -png, -r 200, and a page range from 1 through the page count. Tesseract runs once per page and writes a PDF for that page. One page is returned as that PDF. More than one page is merged in order. The download name is always ocr.pdf. It is a PDF, not a Word file. The original file on the computer is not replaced.

The scan and ocr.pdf both go into MongoDB GridFS, bucket uploads, with tool ocr-pdf. The source uses role source. The result uses role result. The working PNGs live in a temporary directory and are removed when the job finishes or fails. Retention deletes GridFS objects whose upload time is at least RETENTION_MINUTES old. Missing, zero, or non-positive values use 30. The deletions row stores the file id, filename, tool, role, upload time, and delete time, not the bytes. The server deletes the scan and ocr.pdf within 30 minutes.

There is no free OCR file. The gate is called with the free-file count already set to 1, so a visitor without an active purchase is refused. The cap is 20 pages even when a pass is active. A day pass is $6, charged once, for 24 hours, and it does not renew. A year is $60 and runs 365 days. The email must match a purchase whose status is active, from its start up to but not including its end. The free server file is PDF to Word, one file, up to 20 pages, on its own page. Each of pdftoppm and each page's Tesseract is killed after 60 seconds. This form takes one file.

No file, or bytes without %PDF- in the first 1024 characters, fails with "Choose a PDF." A file whose page count cannot be read fails with "This PDF could not be read." No pages fails with "This PDF has no pages." More than 20 pages fails with "OCR stops at 20 pages." A missing pass fails with "OCR needs a $6 day pass or the $60 year. The free server file is PDF to Word. Enter the email from checkout." If pdftoppm or Tesseract is not installed, the page says "OCR is not available on this server." A timed-out or failed process also ends as "This PDF could not be read."

Questions

What is inside ocr.pdf?

A PDF. Tesseract writes an English text layer from the 200 DPI picture of each page. The page does not rebuild columns, and it does not emit a .docx. A PDF that already has selectable text is still sent through the same picture path. PDF to Word is the page that copies an existing text layer into Word.

Is the first scan free?

No. OCR is never the free server file. The free file is one PDF to Word conversion, up to 20 pages, when that PDF has selectable text. This page asks for the checkout email of an active pass before it reads the scan.

What remains after deletion?

GridFS removes the scan and ocr.pdf after the retention clock, 30 minutes unless RETENTION_MINUTES is a positive number. The deletions row keeps ids and times, not bytes. The jobs row keeps the filenames, the page count, the typed email, and the network hash. The temporary PNGs are already gone.

What does a pass change here?

An active pass is required before the first scan. The day pass is $6, charged once, for 24 hours, and it does not renew. The year is $60 for 365 days. Cancel stops the next year charge. The 20-page cap remains for both. A pass does not add a language other than English, and it does not turn the result into Word.

Also

• PDF to Word
• Heavy compress
• Compress PDF