▮Scanned and Image-Only Documents

Legal and official documents arrive as scans far more often than as files. Powers of attorney, court decisions, notarised agreements, apostilled certificates, corporate registry extracts, older contracts pulled from an archive. They come as PDF images, phone photographs, or JPEGs forwarded from a client's client.

 

• Nothing can be counted automatically. No word count means no instant quote. We assess the pages by eye before we price, which is why a scanned job takes longer to quote than an editable one.
• OCR is a starting point, not a solution. Automatic recognition handles a clean printed page reasonably well. It fails on stamps, handwritten annotations in the margin, signatures over text, faded thermal-paper originals, tables that lose their column structure, and Turkish diacritics misread as their unaccented forms. Every OCR output has to be checked against the image before translation begins.
• Layout has to be rebuilt from nothing. There is no source file to return. We reconstruct the document so the translation mirrors the original page for page, with clause numbering, table structure, and signature blocks in the same positions. For anything going to a court, a notary, or a public authority, that correspondence is not cosmetic. Someone will compare the two documents side by side.
• Illegible sections are marked, not guessed. Where the image genuinely cannot be read, the translation says so at that point. It does not quietly invent a plausible sentence. This matters more than it sounds: a guessed figure in a scanned contract is worse than a marked gap.

We have handled apostilled documents, notarised agreements, powers of attorney, affidavits and court decisions in Turkish, French and Arabic, alongside routine corporate documentation, across the full span of our project history.

 ▮What Breaks in Editable Files

Long-lived documentation is where the money is either saved or wasted.

 

We keep a translation memory per client, not per project. When revision three of a manual arrives, the segments that did not change are already translated and already approved. What you pay for is the delta.

 

This only works if the memory is maintained across years and across translators. It is one reason our longest client relationships are the documentation-heavy ones: a horticulture software and documentation programme running in 14 languages since 2011, and a personal protective equipment documentation programme past 2.1 million words since 2012.

 

The memory is yours. We return it in TMX on request, at any point.