Paperly / PDF translation to English / Keep invoice or manual layout

Keep invoice or manual layout when translating a PDF to English.

Invoices and manuals are layout documents: the position of a line item or a warning box carries meaning. Paperly translates the text and rebuilds the page so headings, tables and footers stay where they were. This guide explains what to expect, where it falls short, and when you should not upload the file at all.

What stays in place

The layout translator rebuilds the page rather than reflowing it. For an invoice or a manual, that means the following parts of the page keep their positions.

• Headings. Title blocks, section numbers and chapter headings stay at the same place on the page, with English text in them.
• Tables and line items. Description, quantity, unit price and total columns keep their grid. English text is fitted into the original cell regions.
• Footers and headers. Page footers, terms lines and running headers stay aligned to the page edge.
• Lists and callouts. Numbered steps and warning or note boxes in a manual stay in their regions.

You review the result side by side, with the original page on the left and the English page on the right, so a misplaced line item is easy to spot.

Why some cells overflow and stay in the original language

English is frequently longer than the source language, especially in narrow description cells and short labels. Paperly does not force the longer text in by clipping it. If the English cannot fit its original region, that region keeps the original text and is listed in the review notes.

On an invoice this usually shows up in the item description column. On a manual it shows up in captions, table headers and short warnings. Open the review notes, shorten the English in the editor, then click Update PDF after editing. A shorter wording that fits is better than a faithful sentence that does not.

What does not change: images and text inside images

The tool translates text that is in the PDF as text. Everything that is a picture stays a picture.

• Company logos and letterhead artwork
• Stamps, seals and signatures
• Screenshots and photos in a manual
• Charts and diagrams whose labels are part of the image

If a stamp says "paid" in the source language, it still says that in the English PDF. Tell the reader in your covering note, or annotate the file yourself.

Scanned invoices are a beta limitation

A scanned invoice is a photo of paper. It has no text layer, so there is nothing to extract and translate until the text is reconstructed with OCR. OCR reconstruction on scanned pages is a limitation of the beta. If the invoice came from a scanner or a phone camera, expect weaker results than for a PDF exported from accounting software. Check first whether you can select text in the PDF viewer. If you cannot, treat the document as a scan.

Numbers, part numbers and names must be checked

On these documents, small errors are expensive. Machine translation is not reliable on identifiers, so compare the English page with the original for:

• Invoice and order numbers, part numbers, SKUs and model codes. They should be unchanged. Restore any that were altered.
• Quantities, prices, tax lines and totals. Check decimal and thousands separators and currency markers.
• Company, person and product names. These should usually be left alone.
• Safety warnings and torque or voltage values in manuals. Have a person who knows the product read them.
• Payment terms and legal clauses. The wording matters, so do not rely on the machine output for contractual meaning.

Honest limits

• Scanned pages need OCR reconstruction before translation. That is a beta limitation: a scan has no text layer, so there is nothing to translate until it is recognised.
• Images and text inside images are not translated. Logos, stamps, charts and screenshots stay as they are.
• Mixed inline styling may need review. Bold spans, superscripts and footnote markers can shift after translation.
• Machine translation needs a human read for names, numbers and legal wording. The output is a draft you check against the original.

Confidential documents: where the file goes

Invoices and manuals often contain customer names, bank details or unreleased product information, so be clear about what this tool does. When you click Upload and read PDF, the PDF is uploaded to Paperly for layout processing. It expires after one hour of inactivity. When you click Translate to English, the extracted text is sent to Alibaba Cloud Machine Translation through Paperly.

If you are not allowed to share the document with those services, do not upload it. Use the local document converter on the home page instead. It runs in your browser and does not send the file anywhere, but it produces Markdown rather than an English PDF. The comparison guide sets the two side by side. See the privacy page for details, and the free beta note for status.

A practical order of work

1. Confirm the PDF has selectable text and that you may share it.
2. Upload, set the source language if detection is wrong, and click Translate to English.
3. Go through the review notes first, then every table and footer side by side.
4. Fix identifiers, shorten overflowing cells, click Update PDF after editing, and download the English PDF.

For the general workflow, see the main guide. For tables in Spanish documents specifically, see the Spanish tables guide.

Questions

Related

• PDF layout translator - keep layout, keep formatting, side by side
• Translate a PDF to English and keep layout - the main guide
• Translate a Spanish PDF to English without breaking tables
• Local PDF to Markdown vs cloud layout translator
• Chinese → English sample - desensitized before/after
• Free beta - not for sale
• Privacy - uploads and retention