How to Translate a PDF: Text, Scans, Limits and Review Steps

To translate a PDF, first check whether it contains selectable text or scanned page images. Then choose a translation service that supports that kind of document, select the source and target languages, and review the downloaded result against the original. A translated file can look complete while leaving image labels, scanned pages or important details untranslated.

For an ordinary text-based PDF, Google Translate’s desktop Documents workflow is one option. Scans, confidential material, complex layouts and documents requiring professional review need a more deliberate process. This guide separates those cases so you can choose a route without repeatedly uploading the same unsuitable file.

Choose the route that matches your PDF

Start by selecting a sentence and copying it into a temporary text document. If the copy contains the expected words in the right order, you have a useful starting point for text translation. If nothing selects, or the characters are wrong, investigate the source before treating it as a translation problem.

Keep the original PDF unchanged as your comparison reference. A machine-translated working copy does not become an approved or certified translation merely because its pages resemble the source. Find out what the recipient requires before using an automated result as a final submission.

Check Google Translate’s document limits first

As checked October 17, 2026, Google’s consumer document-translation help lists a 10 MB file limit and a maximum of 300 pages for PDFs. It also says document translation is unavailable on smaller screens or mobile, and text in images and scanned PDF pages can remain visible in the output without being translated.

These are the consumer Documents workflow’s published limits, not universal limits for every product bearing Google’s translation name. Check the actual service before choosing a compression, splitting or OCR strategy. An upload error may reflect a file type, page count or device limitation rather than a problem with your internet connection.

Check the complete saved file’s size and page count. If it is too large, decide whether the receiver needs the whole document. Our PDF splitting guide can help prepare coherent sections, but review the original order afterward. Splitting a report into disconnected pages can remove context that helps a translator interpret terminology.

Translate a text-based PDF on a computer

1. Open Google Translate in your computer’s browser and choose Documents.
2. Choose the source and target languages. Use automatic detection only when appropriate, and check the detected language.
3. Select Browse your computer and choose the intended PDF revision.
4. Select Translate, wait for processing, and use Download translation when available.
5. Save the result under a new name and review it alongside the source.

Choose a clear destination name such as report-es-review.pdf, using a language label that makes sense to your team. Avoid saving every attempt as translation.pdf. A review folder containing both source and target versions helps prevent someone from distributing an earlier incomplete attempt.

This describes Google’s documented workflow, not a hands-on test of a particular uploaded document. It does not promise identical formatting, a particular processing time or unrestricted use for confidential files. Apply your organization’s approved upload and information-handling rules before sending content to any external service.

Handle scanned and mixed PDFs separately

A scanned page can look like ordinary writing while containing only image data. If the translator does not recognize and translate that image content, the output can retain the original language even though other pages changed. Review a representative scanned page early instead of waiting until the end of a long document.

When OCR is appropriate, use our searchable PDF and OCR guide to prepare readable text. Compare the recognized text with the image before translation. A misread product code, decimal point or surname can be carried into another language and become harder to detect afterward.

Do not assume that adding an OCR layer guarantees a particular document translator will process it correctly. Test a small authorized sample, inspect the actual translated text, and choose another approved route if the consumer workflow still leaves the scan unchanged.

A mixed document needs page-by-page classification. For example, an electronically generated report may have a scanned signed appendix and image-based chart labels. The first pages can translate successfully while the appendix and labels remain in the source language. Include those components explicitly in the review checklist.

Prefer the editable source when it is available

If the author can provide a Word document or presentation, consider translating that source before creating the final PDF. This can make paragraph cleanup and label editing more manageable. It also avoids asking a converter to reconstruct editable structure from fixed page positions.

For an existing Google Docs document, Google’s official instructions use Tools > Translate document, followed by a new name and target language. Google creates a translated copy. The help page notes that this command may be absent while working in Microsoft Office editing mode until the file is converted to Google Docs.

Do not treat that command as a direct PDF-import guarantee. If the source must first be converted into an editable document, review that intermediate version before translating. Missing text or incorrect reading order introduced at conversion should be corrected at that stage rather than blamed on translation.

Do not confuse consumer translation with Cloud Translation

Google Cloud’s Document Translation documentation describes a separate developer workflow with authentication, project setup and its own limits. It supports native and scanned PDFs with documented differences. It also warns about formatting loss for scans and complex layouts.

The Cloud documentation further distinguishes mixed native-and-scanned documents, where scanned content is not translated. That is an important reason to classify your source rather than assuming that a scan-support statement covers every mixed PDF. Do not transfer a limit or capability from this API to the consumer Documents tab.

For an occasional file, setting up a cloud integration may be unnecessary. For an organization processing repeated document batches, its administrators can assess cost, permissions, terminology handling and data requirements. This guide does not require purchasing a service or changing an account to translate a single suitable document.

Review meaning separately from page appearance

Start with completeness: compare section headings, the beginning and end of each section, captions, footnotes and table labels. Mark anything still in the source language. A document can have an attractive layout while omitting exactly the small note that qualifies a larger statement.

Then review high-impact details with someone competent in the languages and subject matter. Check names, dates, measurements, currency, negative statements and conditions. Do not “correct” an unfamiliar phrase simply because another wording sounds smoother to a monolingual reader.

Compare tables row by row when their associations matter. A translated column heading must still correspond to the values beneath it. Pay attention to units and footnotes rather than evaluating only whether a table’s border looks neat.

Finally inspect layout. A translated phrase may take more or less space than its source. Look for clipped endings, crowded cells, displaced captions and missing characters. Reopen the downloaded file in the viewer the recipient will use; a successful download is not the final usability check.

Troubleshoot common failures

The Documents tab is missing

Check the device and screen context against the service’s published availability. Switching to the supported computer workflow is more appropriate than repeatedly trying the same mobile preview.

The PDF uploads but some pages remain untranslated

Inspect whether those pages are scans, image-based labels or otherwise different from the translated pages. Classify the content before trying another language setting. A language selection cannot by itself turn an image-only page into reviewed editable text.

The text translates but becomes hard to read

Separate language quality from layout. If the wording is usable but lines overlap, work from the editable source or rebuild the affected sections in an approved editing workflow. Compare any reconstructed version with the original so a layout repair does not accidentally omit content.

The file is rejected even though it seems small

Check the actual file size, page count, extension and whether the file opens normally in a PDF viewer. Use the original saved file rather than a shortcut or incomplete download. For other upload failures, our under-limit PDF upload guide separates size from other acceptance conditions.

Example: prepare a bilingual project handout

Suppose a six-page handout contains ordinary paragraphs, a table and two image-based diagrams. This is an illustrative workflow, not a tested translation result. Classify those components before processing and agree whether the diagrams need translated labels or explanatory text beside them.

Translate the text-based content, then review the table’s headings, units and notes with a qualified reviewer. Handle the image labels through a separate approved editing process if the chosen translator leaves them unchanged. Keep a record of which pages and figures required manual work.

Save the reviewed target-language file and retain the source revision used for comparison. If you provide both languages together, label them clearly and check page order. Do not describe an automatically generated draft as reviewed until someone has actually completed the language and layout checks.

Keep unresolved items visible in the handoff

If a reviewer cannot confirm a term, mark that passage for review rather than silently choosing the most fluent alternative. Record its source page and enough surrounding context to locate it. A short list of unresolved terms is more actionable than a general note saying that machine translation may contain errors.

Distinguish untranslated content from intentionally preserved content. A product name, identifier or cited title may properly remain unchanged, while a diagram’s instructional labels may need work. Ask the language reviewer to make that decision in context. Automatically replacing every remaining source-language word can damage names and references.

Before delivery, check that the filename, cover note and document language label agree. State whether the file is a working draft or the reviewed deliverable, and identify any excluded appendix or separately supplied figure. This prevents the recipient from treating an incomplete translation package as a complete replacement for the original.

Official references

• Google Translate: Documents and websites
• Google Docs: Translate documents
• Google Cloud: Document Translation

Sources checked October 17, 2026. Featured photo by Maarten van den Heuvel, via Unsplash, under the Unsplash License. The folders illustrate document organization, not a translation-service screenshot.