Smartcat extracts the text segment by segment — remembering where each one sits and how it's styled — translates it, and writes it back into the rebuilt layout. Your document is never flattened to plain text.
Upload it to Smartcat's AI document translator and the file comes back with its tables, fonts, images and columns where they were, exported in the format it arrived in, across more than 80 file types.
1,000+ enterprise brands translate documents in Smartcat
Most tools never keep formatting in the first place. They flatten the file to plain text, translate the text, and pour the result back into nothing. So:
Then language piles on: a Spanish or German translation runs longer than its English source, and text that no longer fits its old container overflows, shrinks or gets truncated.
Smartcat works the other way round — the file is parsed in its native format, and the layout is rebuilt around the translation in five steps.
Free for 15 days with 15,000 Smartwords and full access to translation capabilities — no credit card.
Each format has its own preservation profile — what is carried through, and what is worth a visual check in the export. Every card below links to that format's full workflow on Smartcat. Text that is part of an image inside any of these files is a different problem: it needs OCR and the image translator, not document parsing.
Digital PDFs are rebuilt: tables, columns and images are reflowed into the translated layout. Scanned, image-of-text PDFs run on the image pipeline instead — you select pdfimageonly at upload, and output quality follows scan quality.
Slide layout, image positions and formatting are preserved, and speaker notes are translated by default. Text inside images is not translated with the deck — that needs a separate image pass. SmartArt and embedded Excel graphs can cause processing failures.
Cell text, formatting and sheet structure, with upload toggles for sheet names, hidden cells and sheets, headers and footers, comments, graphics and shape text; translation can be restricted to column and row ranges (A:H, 1:50) per sheet. XLSM macros are preserved but not translated. Formulas are not carried through as working formulas, and numbers, currencies and dates are left as authored.
Tags, attributes and document structure stay intact, and HTML is one of the two formats where the editor can show you a target-language preview.
Tags, attributes and structure stay intact — only the translatable text between them changes.
For text baked into an image, select a bounding box and use the formatting toolbar to set font, size, color and alignment, or use Hide to keep an element such as a logo out of the translation.
Timecodes are untouched. Check line length against reading speed after expansion — subtitles are one of the two formats with a target-language preview.
Timecodes and cue structure are untouched; only the caption text is translated.
for simple setup
for ease of use
enterprise brands
languages available
Smartcat is SOC 2 Type II compliant: files are encrypted in transit and at rest, workspaces are isolated per account and access is role-based. Translation is metered in Smartwords — words, not pages or formats — and repetitions are discounted. See how expert-enabled AI translation fits into your document workflow.
50%
Of Translation Costs Saved at isEazy
isEazy, the e-learning platform, saved 50% of its translation costs in 2023 — on the most layout-critical content there is: slides, embedded quizzes and captioned media (case study).
50%
Higher Content Productivity at expondo
expondo, a 400-employee retailer operating in 18 countries, on product content — spec tables, manuals, listings: “We've been able to increase our productivity by 50% while reducing our outsourcing costs by 50%.” — Julia Emge, Director of Content Creation, expondo (case study).
80+
File Formats Rebuilt in Their Original Layout
Word, PowerPoint, Excel, PDF, HTML, InDesign (IDML) and subtitle files are all parsed natively and exported in the format they arrived in.
Up to 70%
Lower Translation Cost
Stanley Black & Decker moved from $200–$300 per 1,000 words to $1.20 on L&D content, and cut average turnaround by two weeks.
31 Hours
Of Monthly Work Time Saved
Savings of 31 hours every month for Babbel’s marketing and L&D teams.
2–3 days
Turnaround, Down From Ten
Smith+Nephew brought a translation cycle that used to take ten days down to two or three.
Upload it, review the translation segment by segment, export it in the format it arrived in and open it to see the rebuilt layout — across more than 80 file types, with fonts, tables and images in place. Free for 15 days with 15,000 Smartwords, no credit card.
Upload it to Smartcat, choose your languages, and export it in the original format. The file is parsed natively rather than flattened to text, so tables, fonts, images and columns are rebuilt in place. You review the translation as text in the editor before exporting; the rebuilt layout itself is checked in the exported file, because target-language preview covers subtitle and HTML files rather than documents.
Over 80, including Word, PowerPoint, Excel, PDF, HTML, XML, InDesign (IDML) and subtitle files. The cards above link to each format's workflow and set out what is preserved and what is worth a visual check.
Yes — their positions are recorded during extraction and rebuilt during reconstruction, with cells and boxes reflowed for longer target text. Images are kept in place; text inside an image is a separate OCR job for the image translator.
The layout reflows to absorb it — German and Spanish commonly run 15–30% longer than English. For very tight layouts where the source barely fit, export the file and check those pages, since target-language preview does not cover documents, then shorten the translation in the editor if a box is straining.
Scanned pages take a different route than digital ones: the AI PDF workflow does not support image-of-text PDFs, so at upload you select pdfimageonly as the parsing method and each page runs through the image-translation pipeline. The result follows the scan quality. Clean printed scans do well; for very long scans, see the large-PDF page for how quality is managed at page 300, not just page 1.
It depends what you are editing.
Free tools are usually the ones that flatten your file — that trade-off is often why you are here. Smartcat is free for 15 days with 15,000 Smartwords, full access to translation capabilities and no credit card; after that it is paid, with no permanently free tier.
Close, but not guaranteed identical. Some specialized fonts may be substituted with similar alternatives, and exact font matching is not guaranteed; bold and italic formatting is generally preserved. Script coverage is real where it counts — the supported set includes Noto Sans SC, TC, Japanese, Korean, Hebrew, Arabic and Devanagari — so non-Latin targets have glyphs. Fonts are carried through and substituted where needed; there is no separate embedding step.
No, and this one catches people out: locale conventions are not converted. A figure written 1,000.00 does not become 1.000,00 for a German file, $ does not become €, and 03/07/2026 is not rewritten as 07.03.2026. Values are translated as text and left as the source had them, so setting local conventions is a reviewer's job — on a price list or a spec table, a job worth assigning explicitly.
Three cases.
Smartcat is SOC 2 Type II compliant. Files are encrypted in transit and at rest, workspaces are isolated per account, and access is role-based. Details are on the security page.
Words, not pages or formats. Translation is metered in Smartwords, repetitions are discounted, and every plan starts free for 15 days with 15,000 Smartwords and full access to translation capabilities, no card — see the pricing page. Wrestling with one enormous PDF? That is the large-PDF page. A batch of heavy files in mixed formats? Translate large files.