What changed on this site and when, newest first. Changes to how files are converted or handled are listed here. Found a problem? Write to contact@pdf-to-word.online; corrections get an entry too.
Fully scanned PDFs now come back as normal, editable paragraphs. Before, the recognised text was placed in dozens of floating text boxes on top of the page picture, so the page looked printed twice. On our scanned test page the new output kept every word and took about half the time (roughly 5 seconds instead of 11 to 12).
“Searchable” scans, the kind phone scanner apps and Google Drive save (a page picture with invisible recognised text underneath), are now detected and their existing text is turned into normal paragraphs. Before, they ended up as hundreds of floating text boxes over the page picture.
Scanned pages are now read with Greek as well as English. On our scanned Greek test page, the share of words recognised correctly went from about 3% to 94%.
A PDF that needs a password to open is now refused straight away with a clear message, instead of failing after the wait with a generic error.
The server no longer writes file names into its error log; failed jobs are logged by a random job number only. Server logs are now size-capped.
New guides, each based on our own test files: test results, keeping tables, Greek and other scripts, protected PDFs, Word, LibreOffice and Google Docs compared and what happens to your files. The five older guides were rewritten with measured results, sources and a review date.
The converter page and the About page now explain how the conversion works, its limits and who runs the site, in all seven languages.
Shorter page titles and descriptions so search results show them in full; a share image, apple-touch-icon and favicon.ico.
About and privacy pages in all seven languages, a contact address on every page, and Google’s consent message for visitors in the EEA, UK and Switzerland.
The converter page is now rendered to plain HTML at build time, so its text is readable without JavaScript.
Unknown addresses return a real “not found” page instead of the home page, and www addresses redirect to the main address in one step.
When a file fails, the reason is shown under that file; the drop area works with the keyboard.
The converter page became available in Greek, the seventh interface language.
Each interface language (English, Spanish, Italian, French, German, Chinese) got its own address, plus a guide for converting on a phone.
Words that came out glued together (“ΑΙΤΗΣΗΓΙΑ” instead of “ΑΙΤΗΣΗ ΓΙΑ”) are now separated: the cause was missing spaces between text runs in some PDFs.
Tables are kept inside the page margins, Word now respects the narrowed column widths, and table rows can grow to fit their text instead of overlapping.
Updated the main conversion engine after a bug made it drop almost all the text of some PDFs while still producing a file (one test document went from 11 recovered characters to 4,929). The converter now also checks that most of the text actually arrived and switches engine if not.
Text recognised on scanned pages is no longer hidden behind the page picture.
Conversions run as background jobs that the page checks every two seconds, so long scans no longer hit connection time limits.
First guides: scanned PDFs, Mac, batch conversion and formatting problems.
pdf-to-word.online went live with batch upload (up to 20 files), OCR for scanned PDFs and automatic deletion of files.