FAQ

The questions people ask before downloading, answered from the project files. If an answer is not written here, it is because the files do not support it — not because it was left out on purpose.

Do my documents get uploaded anywhere?

No. The application is a local Windows program: its server binds to 127.0.0.1 only, so it is reachable from this computer and nothing else. A grep across the server and interface code on 2026-10-16 found no HTTP client, no telemetry and no analytics code. Your documents, thumbnails, drafts and settings live in the local data folder. See the privacy policy and the security page.

Do I need Python or an internet connection?

Neither. The installer and the Professional package carry their own Python runtime and libraries, so there is nothing to install alongside them: no Python, no pip, no internet connection, no account, no cloud service. The only exception is running from a source checkout, which needs Python 3.10 or newer and pip install -r requirements.txt.

Is it free?

Yes. No payment, subscription or licensing code exists anywhere in the project files, and the shipped test report records that no paid or network service is required by the software. The licensing terms themselves: Cannot verify from project files — there is no LICENSE file; the README states only “Original code and design built for this project”. See the terms of use.

Will OCR change how my document looks?

It depends which mode you pick. Searchable OCR adds an invisible text layer: the page still looks exactly like the scan. Editable OCR replaces the printed glyphs with real text so you can correct words — layout changes may occur and formatting may differ. Full detail in the OCR guide.

Which languages can OCR read?

English only, as shipped. The official English PP-OCRv5 model is the one bundled and verified. An “OCR language” setting exists in the app, but recognition in any other language is Not Verified and no other model ships in the package.

What happens to digitally signed PDFs?

They are refused, not modified. Adding or changing anything in a signed file would invalidate the signature, so the app explains the situation and writes nothing — verified in the regression gate (latest run 2026-10-17, 43/43, test_document_integrity 52/52), which checked that a signed PDF is refused without confirmation and that nothing was written when it was refused. The project never claims signature validity is preserved after a modification.

What about password-protected (encrypted) PDFs?

They are refused with a message that names the fix. If you know the password, the Remove password tool produces a fully working copy — a wrong password is reported, never silently ignored — and the locked file still shows a LOCKED badge in the library. Opening a file you do not have the password for is not implemented; that is on the project’s own honest limitations list.

What is the difference between searchable and editable OCR?

Searchable OCR tags the page: the picture stays untouched and an invisible text layer makes the words findable. Editable OCR rebuilds the page: the printed glyphs are covered and real, editable text is inserted in their place, which is what lets you fix a wrong word — at the cost of a layout that may differ from the scan. See the comparison table in the OCR guide.

Does it work completely offline?

Yes. Recognition, conversion, search, the document assistant and printing all run on your machine with local models and libraries. The module that analyses documents contains no API key, no URL and no network call, and a test enforces that. Even the interface is served from the local process — no fonts, icons or scripts are fetched from the internet.

Is there a mobile or web version?

No. Irfan PDF is a Windows desktop application for Windows 10 (1809 or newer) or Windows 11, 64-bit. There is no cloud service, no sync, no mobile app and no hosted web version — the verification report lists all of those as things that simply do not exist. Opening the app in a browser on the same PC is the same local application, not a separate product.

How do I know a conversion worked?

The app checks its own output before reporting success: a PDF must exist, open, keep its page count and render its first page; DOCX and XLSX files are reopened and read back (the Excel test opens the workbook in real Excel and requires a cell to read as a number, not text); extracted images are re-read and their pixel dimensions compared; compressed sizes are re-measured. Your own check is the last word — see how to verify any conversion.

How accurate is OCR?

Honest answer: no OCR system is perfect, and this one does not claim to be. Recognition errors happen — especially on faint or badly skewed scans — and the English model’s performance on other scripts is Not Verified. Always check critical numbers, names and dates against the original page, and never use OCR output alone for legal, medical or financial figures.

Next: Installation · Release notes · Download