Smartcat translates the text in your PDF where it sits, and images, logos, icons and vector diagrams are detected and recreated in their original positions, preserving sharpness and placement. Text baked into an image is recognized, translated and written back into the picture, layout and graphics intact. A diagram comes out translated — not merely surrounded by translated text.
Smartcat translates picture-heavy documents for 1,000+ enterprise brands.
Two pipelines, one document
Native text — paragraphs, headings, tables — is translated directly and re-set in place, with the layout adjusted where the translation runs longer than the source. Image regions are detected and recreated in their original positions, and the text baked into them is recognized, translated and overlaid back onto the image. Figure captions stay attached to their figures because the document structure, not just its text, is what gets rebuilt.
1
Upload the PDF
Manuals, catalogs, reports and decks exported as PDF. Files up to 100 MB, with no page limit — documents of any length are accepted. For anything larger, see translate a large PDF.
2
Pick your languages
One source language, any number of targets, from 280+ supported languages. Your glossary and translation memory apply to the body text and to the labels inside diagrams alike, so terminology stays consistent across both.
3
Text inside the images is treated as text
A screenshot’s button labels, a diagram’s callouts, a chart’s axis titles — each detected region is read, translated with the same glossary as the body text, and re-set inside the image, preserving the original layout, fonts and graphics. That is the difference between a document that is translated and one that merely looks translated.
4
Review the hard spots
Add a source layout check stage before translation: text extraction does not always work perfectly on graphic-heavy pages, so someone checks the extracted document first. In-image text can be slightly less accurate than body text, so start your review with the chart labels — alone, or with a vetted reviewer from the Marketplace.
5
Download the translated PDF
Pictures where you left them, captions with their figures, and the labels inside the diagrams in the new language. Bold and italic formatting is generally preserved; some specialized fonts are substituted with similar alternatives.
We love having total visibility over our entire translation lifecycle on our Smartcat workspace due to the centralized nature of the platform and being able to gather all the linguistic assets we need, including translation memories and glossaries.
”Explore case study →
280+
languages
One source language into any number of targets, with your glossary and translation agent keeping terminology identical in the body text and inside the diagrams.
100 MB
maximum PDF size
No page limit — documents of any length are accepted. A typical 10–20 page document takes 15–45 minutes; 60+ pages can take significantly longer.
300 DPI
minimum for scanned pages
If the whole file is scanned pages, choose the pdfimageonly parsing method on upload and keep pages straight, evenly lit and high in contrast. The scanned document translator covers that job.
A PDF is more than its text layer. Here is what happens to each part of yours — and what is worth checking before you send the file on.
Start free trial
for ease of setup on G2
for ease of use on G2
enterprise brands
average rating on G2
Upload the file, pick the languages, and see what comes back — including the labels inside the diagrams. Free for 15 days with 15,000 Smartwords, no credit card, with full access to translation capabilities.
3 days
to deliver what once took 10
Smith+Nephew cut translation turnaround from 10 days to two or three by centralizing their complex translation workflows in Smartcat.
Up to 70%
Cost savings
Stanley Black & Decker cut translation costs by up to 70% while improving the quality of their multilingual content.
31 hours
of work time saved monthly
Babbel’s internal teams reclaimed 31 hours of work each month by automating their translation processes.
Upload the PDF, pick the languages, download a file your reader cannot tell was ever translated — with the labels inside the diagrams translated too. Free for 15 days with 15,000 Smartwords, no credit card.
Upload the PDF and Smartcat handles it in one pass: the native text layer is translated in place, and each detected image region is read, translated and overlaid back onto the image, preserving the original layout, fonts and graphics. No copy-paste, and no separate step for the diagrams.
Yes. Images, logos, icons and vector diagrams are detected and recreated in their original positions, preserving sharpness and placement, and figure captions stay attached to their figures. Most free converters extract the text, translate the extraction and generate a new file — which is why the pictures come back stripped or shuffled.
Documentation is explicit that in-image text can be slightly less accurate than body text, so it is worth a quick review. Dense chart labels, stylized logotype text and low-resolution screenshots are the hardest cases; the editor lets you correct those regions directly.
The documented formats include PNG, JPG and JPEG, GIF and TIFF. For an animated GIF only the first frame is translated. HEIC and WebP are not on the documented list.
Smartcat supports 280+ languages, including major global languages and a wide range of regional variants, with one source language and any number of targets in a single project.
The layout is rebuilt, not re-typed: text is re-set in place with the layout adjusted where the translation runs longer than the source. Plan for some loss on interactive parts — form fields, buttons and interactive elements may not be preserved — and use the editor for minor adjustments on very complex pages.
Then it is not a PDF with pictures — it is one big picture per page.
There is no automatic fallback: the AI PDF workflow does not support scanned PDFs, so in the upload dialog you choose pdfimageonly as the parsing method, which sends each page through the image-translation pipeline instead.
Aim for at least 300 DPI, with pages straight, evenly lit and high in contrast. The scanned document translator is built for that job, including handwritten content.
Up to 100 MB for the AI PDF workflow, and there is no page limit — documents of any length are accepted. Time is the trade: a typical 10–20 page document takes 15–45 minutes, and 60+ pages take longer.
A 200-page manual is a job you start and come back to. Over 100 MB, split the file with a PDF editor and see translate a large PDF.
Free for 15 days with 15,000 Smartwords and no credit card, with full access to translation capabilities. After the trial it is a paid plan. For one small text-only PDF a free converter will do; this matters when the document has pictures worth keeping and terminology worth getting right.
Smartcat is SOC 2 Type II compliant, and documents are encrypted in transit and at rest in an isolated workspace. Manuals and catalogs are usually pre-release material — the kind of file not to feed a free converter. Details on the security page.
Where the original font supports the target language’s script, closely; where it does not — a Latin display font asked to render Japanese — a substitute is used.
Text re-set inside an image is drawn from Smartcat’s set of supported fonts, matched to the original as closely as that set allows, and the source font is not embedded into the output.
Expect close, not identical.
There is no documented images-only toggle for a PDF. What the image editor does support is the reverse, region by region: select text you do not want translated — a logo or a brand element — and use Hide to exclude it while keeping the original background. For the whole-PDF picture, start from the PDF translator hub.