PDF

A PDF is a description of where marks go on a page. Some carry a text layer (anything exported from a word processor); some are scans, which are pictures of pages with no text at all. PDF handles both, on your device, and gives back a PDF.

Text-layer PDFs

The text is extracted page by page with its positions, translated sentence by sentence, and laid back onto the page in place of the original text. Because a translated sentence is rarely the same length, the tool fits the new text into the original box, shrinking the type slightly when it must; the page keeps its images, rules and layout. Each page appears as it finishes and the status line counts sentences.

Scanned PDFs

A scan has no text, so the page is first read by the recognizer, on your device. The recognizer's language data (1 to 5 MB per language) is downloaded once through traxlate.com the first time you use it and kept locally; the page asks before that download. Recognition quality follows scan quality: a clean office scan reads well, a phone photo of a curled page less so. The result is translated like any text and composed onto a new page over the scan.

Reviewing and exporting

The viewer shows the translated page over the original; Text only shows the sentences as a list, and a sentence can be edited in place — an edit is remembered in your translation memory so the same sentence comes back corrected. Export offers a PDF file, a styled PDF, or plain text when you only need the words.

Limits, honestly

• Tables and multi-column layouts are followed by position; an unusual layout can put a sentence in the wrong reading order. The Text only view shows the order used so you can catch it.
• Forms are translated as text; field logic is not preserved.
• A password-protected PDF must be unlocked first.
• Very long documents run page by page and can be stopped at any point; the finished pages are kept.

Why does the first page take longer than the rest?

The first run loads the translator into memory (and for a scan, the recognizer). The status line says so; later pages skip it.

Is the PDF uploaded to be converted?

No. Extraction, recognition, translation and composition all run in your browser or app.