Translate foreign-language PDF files locally vs. cloud server uploads
Translating confidential contracts, medical summaries, or internal memos on standard web translators uploads every sensitive page to third-party servers. Using local browser translation in desktop Chrome and Edge instead of cloud server uploads keeps your entire PDF on your own hardware with 0 bytes transmitted externally.
Detecting source language across 30 supported languages
When you drop a PDF or paste text, the tool samples the opening 20 blocks locally and refines the guess using the browser’s language detector. It lets you translate between 30 major world languages in the From and To menus, and selects a sensible default target language from your browser locale.
Clicking Translate invokes the Chrome for Developers Translator API documentation interface directly from your gesture so the browser can download or load the local language pair. You can halt long documents mid-way with Stop or flip directions anytime using Swap languages.
Comparing source and translated blocks side by side
As each block finishes, the main panel populates a two-column comparison table showing original text beside the translated output. Right-to-left scripts such as Arabic and Hebrew automatically receive dir="auto" alignment in the preview cells for natural reading.
After reviewing the side-by-side table, you can also summarize long translated reports into key points on the same device without leaving your browser.
How to preserve original document page geometry in PDF, Word, and text
Translated PDF export preserves original page dimensions, background graphics, and spatial bounding box coordinate text placement on flattened pages, whereas Word (.docx) and plain text exports keep every paragraph fully editable. Picking the right export format depends on your goal for visual fidelity versus post-translation editing.
| Export format | Layout preservation | Text selectability | Best use case |
|---|---|---|---|
| Translated PDF | Keeps page sizes, graphics, and block coordinates | Flattened rasterized pages | Visual review and reference copies |
Word (.docx) |
Linear headings and paragraphs by page | Fully editable text | Rewriting, formatting, or long blocks |
Plain Text (.txt) |
Paragraph breaks only | Fully editable UTF-8 text | Copying into emails or notes |
| Side-by-side preview | Dual-column browser table | Copyable directly from cells | Quick bilingual verification on screen |
Inside the PDF builder worker, each page renders onto an MDN OffscreenCanvas Web Worker API surface scaled up to 3x and bounded at 2,600 px on the longest side to protect browser memory. Spatial bounding box coordinate text placement ensures headings, captions, and body columns stay anchored where they appeared on the source sheet.
Because finished PDF pages are flattened, links and interactive form fields are no longer clickable. If you start with an untranslated PDF and only need an editable Word document without language conversion, use editable Word (.docx) conversion instead.
Fitting expanded translations, RTL scripts, and scanned PDFs
Languages rarely share identical sentence lengths: translating concise English headings into German or Spanish frequently expands character counts by 20% or more. Fitting expanded text back into tight PDF boxes without clipping requires grapheme-aware wrapping and controlled font scaling.
Auto-shrinking font sizes and wrapping with Intl.Segmenter
The layout engine wraps every translated block using MDN Intl.Segmenter Unicode grapheme reference rules so scripts without spaces, such as Japanese, Thai, or Chinese, break cleanly at character boundaries. If text still overflows its box height, the fitter steps down in 0.5 px increments to a 6 px minimum floor and notifies you how many blocks were shrunk.
Blocks exceeding 20,000 characters cannot fit inside a single fixed PDF box and prompt you to export as Word (.docx) instead. Keeping paragraphs concise ensures every translated block fits legibly inside its original frame.
Handling Arabic, Hebrew, and English OCR scans
When exporting to Arabic or Hebrew, the canvas renderer switches to right-to-left direction and right alignment inside each bounding box. Scanned PDFs run local English OCR first and paint a sampled solid background color over the scanned words before drawing translated text, which can visibly alter patterned paper textures.
For non-English scanned documents that need dedicated multi-language recognition before translating, make multi-language scanned PDFs searchable first.
Why is on-device PDF translation limited to desktop Chrome and Edge?
On-device PDF translation requires desktop Chrome or Edge because Safari, Firefox, and mobile browsers do not yet ship built-in local translation models. Rather than silently routing your private files to a cloud API, the tool disables translation on unsupported browsers to guarantee zero data leakage.
Browser requirements and first-run language pack downloads
Desktop Windows, macOS, Linux, and ChromeOS systems download each requested language pair once and display a live percentage progress indicator during caching. Once cached, subsequent translations between those two languages run completely offline even with Wi-Fi disabled.
If the browser reports that a user gesture is required before starting a language pack download, select Translate once more directly with your mouse or keyboard to grant permission.
Tip: Check Download translated PDF pages visually before sharing them externally. Multi-column layouts or dense tables can cause font shrinking to 6 px, making Word (.docx) much easier to read.
Reviewing machine translations before signing or submitting
Machine translation is a practical aid for reading foreign manuals, invoices, and research, but it should not replace certified human translation for legally binding contracts. For hands-free review of translated notes, you can also listen to translated text read aloud locally using your device’s built-in speech voices.