Most translation tools skip the text inside images, charts, and diagrams — they only translate the live, selectable text layer, leaving labels, captions, and embedded graphics in the original language. To translate that locked text, you need a document translation platform with built-in OCR that reads the text inside images, translates it, and rebuilds the document so the layout stays intact. Native features in Microsoft 365, Google Workspace, and most AI assistants explicitly do not do this, which is why a chart's axis labels or a scanned page's body text often come back untranslated. As of June 2026, OCR-based document translation is the reliable way to close that gap.
Bluente is an AI-powered document translation platform used by 30,000+ professionals to translate files in 120+ languages while preserving original formatting, including text recognized inside images. This article explains why embedded text gets skipped and how to translate it without manually retyping anything.
Why Doesn't My Translation Tool Translate Text in Images?
Most translation tools only process the document's text layer — the selectable, machine-readable characters — and treat images as flat graphics to leave untouched. A chart exported as a picture, a scanned contract, a diagram with embedded labels, or an infographic is, to the tool, just pixels, so the text inside it is never seen and never translated.
This is a documented limitation across the tools professionals use every day. Native Microsoft 365 translation states that text embedded in images is not translated. Microsoft Copilot leaves text in images and charts in the original language. Google's built-in tools behave the same way. The result is a half-translated document: the paragraphs convert, but the figure captions, the org chart, and the screenshot stay in the source language — often the very parts a reader needs most.
What Is OCR and Why Does It Matter for Document Translation?
OCR (optical character recognition) is the technology that reads text inside an image and converts it into characters a system can translate. For document translation, OCR is what makes embedded text translatable at all — without it, anything stored as an image is invisible to the translation engine and passes through unchanged.
The hard part is not just reading the text; it is putting the translated text back in the right place. A good document translation platform recognizes the text in a chart or scanned page, translates it, and reconstructs the document so the translated text sits where the original did, with the surrounding layout preserved. Bluente handles this across PDF, image files (PNG, JPG, TIFF), and embedded graphics, so a scanned page or a chart-heavy report comes back fully translated, not partially. Charts notoriously lose their labels in naive translation; OCR-aware reconstruction is what prevents that.
How Do I Translate a Scanned Document or a Chart-Heavy Report?
Upload the file to a document translation platform with OCR, select the target language, and the platform reads the embedded text, translates it, and returns a document that keeps the original layout. There is no manual step of retyping labels or recreating charts — the work that makes embedded-text translation painful by hand is exactly what the platform automates.
This matters most for the documents where embedded text is unavoidable: scanned legal exhibits, financial reports full of charts, engineering diagrams, manufacturing schematics, and presentation decks built around visuals. For these, a text-only translator produces a document that looks done but is not. Bluente translates these files in 120+ languages in under 2 minutes on average, recognizing and translating the text inside images while keeping tables, charts, and structure intact — so the output is a complete, usable file.
Can I Translate Text in Images Without Losing the Layout?
Yes, with a platform built for format preservation. The challenge with embedded text is that translation often changes length — a translated label may be longer or shorter than the original — and a good platform handles that while keeping the chart, diagram, or page visually intact. Layout preservation is the difference between a translated document you can hand off and one you have to rebuild.
This is Bluente's core strength: the translated file looks exactly like the original, including the parts that were locked in images. For a regulated or client-facing document — where a mislabeled chart or an untranslated caption is a real error — that fidelity is the point. Format preservation is not a cosmetic nicety here; it is what makes the translation trustworthy enough to use for legal, financial, and technical work.
Is It Secure to Upload Scanned Confidential Documents for Translation?
It is secure only if the platform offers explicit guarantees, because scanned documents are often the most sensitive files an organization handles — signed contracts, identity documents, financial records. The translation tool you choose for these should treat the document as the unit of risk and tell you exactly how it is handled.
Bluente is SOC 2, GDPR, and ISO 27001 compliant, with zero data retention by default, automatic deletion within 24 hours, and documents never used to train AI models. For scanned exhibits, immigration paperwork, or financial statements with embedded figures, that posture is the baseline. Free image-translation tools rarely make such commitments, which makes them a poor fit for confidential scanned material regardless of how convenient they look.
Frequently Asked Questions
Q: Why does my translated document still have text in the original language inside images?
Most translation tools only translate the selectable text layer and skip text embedded in images, charts, and diagrams. Native Microsoft 365 and Copilot explicitly leave image text untranslated. You need a platform with OCR, like Bluente, to translate that text.
Q: What is OCR translation?
OCR translation uses optical character recognition to read text inside images, translate it, and place it back in the document. It is what makes scanned pages, charts, and infographics translatable while preserving the layout.
Q: Can Bluente translate a scanned PDF?
Yes. Bluente recognizes and translates text inside scanned PDFs and image files (PNG, JPG, TIFF), then reconstructs the document so the layout stays intact, across 120+ languages in under 2 minutes on average.
Q: Will translating text in a chart break the chart's layout?
Not with a format-preserving platform. Bluente translates embedded labels and captions while keeping the chart, diagram, and surrounding structure visually intact, even when the translated text length differs from the original.
Q: Is it safe to translate confidential scanned documents?
Only with a secure platform. Bluente is SOC 2, GDPR, and ISO 27001 compliant with zero data retention by default. Avoid free image-translation tools for confidential scanned files, as they rarely make security commitments.
---
Start translating documents for free. Bluente preserves your formatting across 120+ languages in under 2 minutes. Try BluTranslate free — no credit card required.