Supported files and outputs
Choose the tool for the output you need.
| Tool | Input | Output |
|---|---|---|
| Text | Typed text and 43 file formats, listed below | Translated text · Copy or download TXT |
| Image | PNG, JPEG, WebP, GIF, BMP, TIFF, AVIF | PNG with translated text in detected positions |
| Document | Text-based PDF, DOCX/DOTX, DOC, PPTX/POTX, PPT, XLSX/XLTX, XLS, RTF, ODT/OTT, ODP/OTP, ODS/OTS, CSV, TSV, Markdown, TXT, HTML, reStructuredText, AsciiDoc, Org, EML, EPUB, FB2, SRT, WebVTT, ASS/SSA, LRC, SBV, SUB, SAMI, TTML/DFXP, JSON, JSON Lines, YAML, TOML, INI, Java properties, RESX, RESW, XML, Gettext PO/POT, XLIFF, Qt TS, Android strings.xml, Apple Strings, Stringsdict, String Catalog, Flutter ARB, TMX, Trados SDLXLIFF, memoQ MQXLIFF | Translated file in its original document format |
Text input formats
Files up to 10 MB. Text requests support up to 20,000 translatable source words and 120,000 characters, within your remaining word credits. The Text tool produces text; it does not recreate the original document layout.
| Category | Formats and extensions |
|---|---|
| Text | TXT — .txt Markdown — .md, .markdown reStructuredText — .rst AsciiDoc — .adoc, .asciidoc Org — .org |
| Documents | PDF — .pdf Word — .docx PowerPoint — .pptx Excel — .xlsx OpenDocument text — .odt OpenDocument spreadsheet — .ods OpenDocument presentation — .odp XPS / OpenXPS — .xps, .oxps |
| Ebooks | EPUB — .epub FictionBook — .fb2 |
| Web | HTML — .html, .htm, .xhtml |
| Data | CSV — .csv TSV — .tsv JSON — .json JSON Lines — .jsonl, .ndjson YAML — .yaml, .yml TOML — .toml INI — .ini, .cfg |
| Localization | XML — .xml XLIFF — .xlf, .xliff TMX — .tmx RESX — .resx Qt translation — .ts Gettext — .po, .pot Java properties — .properties |
| Subtitles | SubRip — .srt WebVTT — .vtt ASS / SSA — .ass, .ssa LRC — .lrc |
Email — .eml Mailbox — .mbox | |
| Images | PNG — .png JPEG — .jpg, .jpeg WebP — .webp GIF — .gif BMP — .bmp TIFF — .tif, .tiff AVIF — .avif |
How files are read
Document and ebook text is extracted with headings, lists, and tables where available. Data, localization and subtitle files go to the AI model as raw text and come back as text, so their keys, tags and timestamps may change. Every word of the raw text is counted, keys and tags included.
Code, embedded objects, macros, and text drawn as shapes are not translated. Encrypted ebooks and password-protected documents are unsupported. Text does not read DOC, XLS, PPT or RTF files; Document translates them.
PDF and image text
In the Text tool, PDFs and supported images are sent directly to the AI model for text translation, without a separate OCR step. Images are converted on the server when needed. Large or heavy images are resized and compressed before translation. Animated images and multi-page TIFFs use the first frame or page. PDFs can contain selectable or scanned text. Reading quality depends on the source clarity. Other file types use programmatic text extraction before translation.
Image requirements
Upload images up to 10 MB and 8 megapixels, with no side longer than 4,096 pixels. Clear text on a simple background works best. Translated text stays within detected text areas; fonts and detailed backgrounds can vary.
Document requirements
Guest and Free accounts support files up to 50 MB. Every paid tier supports up to 100 MB per document. The Document tool reads scanned PDF pages from their pictures. Password-protected files and macro-enabled Office files are currently unsupported for layout-preserving output.
Document layout preservation
Word files retain supported paragraph, table, and run formatting. PowerPoint translates supported text on visible slides. Excel translates supported text in visible cells while retaining numbers and formulas. PDF text is fitted into the original text areas. Longer translations can change line breaks or require layout review.
In ODT and ODP files, the text is translated wherever it shows: in tables, headers and footers, footnotes and speaker notes; field values, comments, hidden text and embedded charts stay as they were. DOC, PPT, XLS and RTF files come back in their own format, an XLS with its formulas still working, and templates come back as templates.
In JSON, JSON Lines, YAML, TOML, INI, Java properties, RESX and RESW files, only the text values are translated; every key, number, comment and line break stays as it was. JSON with comments or trailing commas is not supported.
In XML data files, such as a product feed, the text of each element is translated; tags, attributes, comments, numbers, prices and codes stay as they were.
In SRT, WebVTT, ASS/SSA, LRC, SBV, SUB, SAMI and TTML/DFXP files, each subtitle is translated as a whole and its lines are rewrapped to fit; its timing, style, position and speaker label stay as they were.
In Markdown, TXT, HTML, reStructuredText, AsciiDoc and Org files, the text is translated and the rest stays as it was: code, links, front matter keys, markup, scripts and styles.
In EML emails, the subject and the text a reader sees are translated, quoted replies included; the other headers, addresses, attachments and inline images stay as they were. Signed or encrypted emails are not supported.
In ODS files, the text in visible cells is translated, as in Excel; numbers, formulas and hidden sheets stay as they were. In CSV and TSV files, only the cells with words are translated; numbers, dates, codes such as SKUs, web addresses and formulas stay as they were.
In Gettext PO and POT, XLIFF and Qt TS files, each string is translated and its placeholders, tags, keys and notes stay as they were; plural strings get the target language’s plural forms. Into the file’s own language, the strings it already has stay as they are and are not counted.
In Android strings.xml, Flutter ARB and Apple Strings, Stringsdict and String Catalog files, each string is translated; its keys, placeholders, markup and comments stay as they were, and so do strings marked not to be translated. Plural strings get the target language’s plural forms.
In TMX translation memories, each unit gains a translation in the target language, and a unit that already has one keeps it and is not counted. In Trados SDLXLIFF and memoQ MQXLIFF files, each segment is translated with its tags and marked as translated; into the file’s own target language, confirmed and locked segments stay as they are.
Translate a document