Reglyph now recognizes table-of-contents pages in scanned and image-only PDFs and translates them entry by entry. Each translated entry stays on its original row, dot leaders run to the right edge, page numbers stay right-aligned, and chapter headings stay inside their heading bars.
Previously a whole contents page could come back as one run-on paragraph. We tested the change on the multi-page table of contents of a 50-page Japanese equipment manual translated into English.
Line breaks inside a block of text are now kept in the translated PDF. Numbered and bulleted list items, “Note:” lines and parenthetical remarks each start on their own line, as they do in the original, instead of being merged into a single paragraph. In the 50-page manual we tested, this affected almost half of the pages.
Pages that mix text with diagrams, photos or tables now finish much sooner. In tests on a 50-page illustrated equipment manual, figure-heavy pages were processed about three times faster than before, so long illustrated documents such as manuals and reports are ready sooner.
The Reglyph website is now available in Arabic (right-to-left), Turkish, Hindi and Indonesian, alongside English, Japanese, Simplified and Traditional Chinese, Spanish, Portuguese, French and German. A language menu in the header switches between them.
Translations are now worded to stay close to the length of the original text, so the translated text fits the space the source text used on the page and needs less shrinking. Pages that are scanned rotated by a quarter turn are also detected more reliably.
Reglyph now checks whether the recognized text of each page matches the script the document is written in, and re-reads pages where it doesn’t. This fixes garbled text on scans in less widely supported scripts. Georgian is the first such language and can be chosen as the source language.
Pages that were scanned sideways, and landscape pages, are now detected and turned upright before translation, then returned in their original orientation in the translated PDF.
Printed text that sits on top of a watermark, logo or background pattern is now translated like any other text, while real signatures and stamps are kept exactly as they are. Headers and letterheads that carry meaning, such as an issuing authority, a document title or a footnote, are translated; page numbers and reference IDs are left untouched.
Japanese pages set in vertical columns (tategaki) are now translated and laid out as horizontal text, so the translated page reads naturally in English and other horizontal languages.
Alongside the translated PDF you can download a bilingual PDF with each original page and its translation side by side, which makes it easy to check a translation against the source. Choose it at upload and it is ready together with the translation.
Reglyph