DjVu to PDF Converter With Text Layer

DjVu stores backgrounds and pictures with the IW44 wavelet codec, which PDF does not have, so they are re-encoded as JPEG, which is heavier at the same quality. Text carries over at almost no size cost: letter masks go into JBIG2, the same idea as in DjVu — one copy of each letter per book. In our measurements a text reference grew from 462 KB to 0.85 MB and a 62-page color book from 2.8 to 5.8 MB. For comparison, the PDF of page pictures that the reference DjVuLibre library produces for the same color book weighs 135 MB. If you need a smaller file, choose “Smaller file”: the background drops to 100 dpi while the text stays just as sharp.

If the DjVu has a text layer, yes: it is carried over as invisible text on top of the page, so search, selection and copying work as in any PDF. The summary after upload shows how many pages have it. If there is no text (common for scans), turn on “Recognize text”: a recognition model reads the pages right in the browser and the result goes into the same invisible layer. We also checked that the text comes out the same in the Chrome engine and in MuPDF: across the 71 pages of the DjVu3 reference it matches the original text by 99.7% and 100%.

Yes. DjVu bookmarks become a PDF outline with the same nesting, and the PDF opens with the outline panel shown. Internal links go to the right pages, external ones to their addresses. If the book’s pages have their own labels (Roman numerals in the preface, “Chapter 1”), they become the PDF page numbers, so going to a page number in the viewer matches the book.

Such books are stored as index.djvu plus one file per page. Select all of them at once: in the file dialog open the book’s folder and press Ctrl+A (Cmd+A on a Mac). If a part is missing, the tool names it instead of building a PDF with holes.

We checked the DjVu decoder behind this tool against the reference DjVuLibre library: 444 purpose-made files covering every kind of compression and a dozen real books matched it pixel for pixel in every layer. Text colors carry over exactly when a page has up to 48 different inks; on scans with hundreds of shades, very close colors are merged — a test page with 1,244 shades of brown became 25 groups, each within 24 of 255 of its shades in every channel.

“As in the original” keeps everything as in the DjVu: color backgrounds and pictures at their own resolution. “Smaller file” reduces backgrounds and pictures to 100 dpi (150 dpi for photo pages); the text is not touched and stays sharp. “Black and white for printing” removes the background and color, leaving only letters and lines without yellowed paper or scan stains — such pages print cleaner and weigh the least.

The file uses ordinary PDF 1.5 features — JPEG for the background and JBIG2 for text. Adobe Reader, Chrome, Firefox, Safari, Preview on a Mac and PDF reading apps on Android and iPhone understand them; we checked the output with three independent viewing engines. If the PDF has to open in very old software, “More options” has a CCITT variant: it is 3–4 times larger but reads even in software from the early 2000s.

Yes: pick a range in the “Pages” row. The outline then keeps only the pages that made it into the PDF, links to left-out pages are removed, and page numbers continue the book’s numbering — a chapter starting at page 120 starts at page 120 in the PDF too.