Translating a PDF Without Wrecking Its Layout

Paste a page into a translator and you get text. Paste a book in and you get a wall of text where a designed document used to be: no columns, no captions under their pictures, no page breaks in the right places.

This tool rebuilds the pages instead. Same layout, same fonts, same images, different language.

---

What "keeps the design" actually means

The output is a PDF that mirrors the original page for page. Text sits where it sat. Images stay where they were. A caption under a diagram is still under that diagram.

That matters most for exactly the documents people need translated and least want to rebuild:

• Manuals and instructions, where a step's text has to stay beside its illustration
• Books with real typesetting
• Reports and decks with charts and tables
• Catalogues, where every product sits in a grid

Without layout preservation, translating any of these means re-laying-out the entire document by hand afterwards. That's the job this removes.

---

The practical limits

50 credits per page. Billed on the page count, so the cost is knowable before you start: a 40-page manual is 2,000 credits, a 300-page book is 15,000.

Up to 500MB per file. That's generous, most books are far under it.

PDF in, PDF out.

Source and target must differ, and the source can be left on auto-detect if you're not sure.

You can pause and resume. Useful on a long book, and it means you're not committing to the whole run in one sitting.

---

Dual-language documents

There's a second pair of language fields, and it exists for documents that are already bilingual.

You can set a secondary source language and its own secondary target, so a document mixing, say, Korean and English can have each part sent to a different destination language, in one job, rather than being forced through a single pairing that mangles half of it.

Most people won't need it. The people who do, need it badly.

---

Test before you commit

This is the important habit, because 50 credits a page adds up fast on a book.

Extract 3 to 5 representative pages into a small PDF and translate that first. For 150–250 credits you learn:

• whether the layout survives your document's particular structure
• whether the terminology comes out the way your field uses it
• how it handles your tables, captions and footnotes
• whether the fonts hold up in the target language: this is the one that surprises people, because a language that runs longer than the original has to fit in the same box

Pick pages that are hard: the one with the complicated table, the one with text wrapped around an image. A clean page of body text will pass and tell you nothing.

---

What to expect it to struggle with

• Text baked into images. If a diagram's labels are part of the picture rather than the PDF's text layer, they aren't text and won't be translated.
• Scanned documents for the same reason: a scan is an image of a page, not a page.
• Heavy expansion. Some language pairs run substantially longer, and text that has to fit a fixed box will be tight.
• Very specialised terminology. Check a sample in your field before trusting a whole book.
• Handwriting.

---

After it's done

Read the result, particularly anything that carries risk, safety instructions, legal text, dosages, specifications. Layout-preserving translation makes a document look finished, which is exactly why it deserves a proof-read rather than less of one.

For a document that matters, treat this as a very good first pass that saves the layout work, not as a final translation.

---

Try it

Open book translation with a five-page extract of something you actually need translated. Around 250 credits to find out whether it handles your kind of document, far better spent than discovering it on page 200.

Then: what everything costs