Cantonese PDF & Document Translator — English ↔ Cantonese

Translate English to Cantonese or Cantonese to English while preserving your document’s layout. Upload your file, choose the target language, and get a natural translation that keeps the meaning of the original.

Loading credits…

How Cantonese documents come out

Upload a PDF, Word file, slide deck, spreadsheet, EPUB or a photo of a page and translate it between English and Cantonese. Cantonese is the spoken language of Hong Kong, Macau and much of Guangdong, but what people write down varies: most formal documents there are in standard written Chinese, while messages, comics, adverts, forum threads and transcripts of speech use written Cantonese, with characters and grammar of its own. This page is for the written-Cantonese side, and it reads standard written Chinese when that is what arrives.

Written Cantonese uses Traditional characters plus a layer of characters that standard written Chinese does not, and that layer is where documents go wrong. Everyday words have their own graphs: 嘅 (possessive, where standard Chinese writes 的), 咗 (completed action), 冇 (not have), 佢 (he or she, with no gender marked), 唔 (not), 係 (is), 喺 (at), 啲 (some, plural), 嗰 (that), 嘢 (thing). Some of the characters written Cantonese needs were missing from older character sets and reached Unicode only through the Hong Kong Supplementary Character Set, so a PDF or scan made with an older font may show them as blank boxes or substitute look-alikes, and a few, such as 𠮶, still fall outside common fonts. Colloquial Cantonese is also code-mixed: a Hong Kong message or memo routinely drops in English words (個 project 好 urgent), transliterated loans (巴士 bus, 的士 taxi, 士多 store) and local abbreviations that mean nothing to a reader who knows only Mandarin. Comics set dialogue in vertical balloons, chat screenshots come with timestamps and emoji, and older newspaper columns mix classical, standard and colloquial registers in one paragraph. In Guangdong the same words often appear in Simplified characters online, where 係 becomes 系 while 嘅, 冇 and 佢 have no simplified forms. Chinese runs shorter than English, so pagination shifts when an English file goes in.

Translate your document and keep its structure

The translator keeps your file’s structure while translating the text in context. The downloaded format depends on the source file.

PDFs and scanned pages
The translated PDF follows the page visually, with translated pages rendered as images. Selectable PDF text and identical fonts are not guaranteed.
Word documents and PowerPoint presentations
Start with DOCX or PPTX when you have it. Native paragraphs, table text and slide text can stay editable in the corresponding file format. Translated text inside pictures remains image content; review text wrapping and font changes in your editor.
Excel workbooks and EPUB books
Use the original XLSX or unprotected EPUB for a translated workbook or ebook. Review formulas, numbers and text cells in a workbook; check chapter order, navigation and text flow in your ebook reader.

What people translate to and from Cantonese

Messages, letters and social posts from Hong Kong

The most common upload is a screenshot: WhatsApp and Signal threads with relatives, group-chat announcements, Facebook and Instagram posts, and forum threads, all written the way people speak, with English words dropped in and sentence-final particles carrying the tone. Families who have moved to the UK, Canada, Australia or the US translate them for a partner or a child who grew up without the language, and community workers translate them for casework. Save the screenshots as PNG or JPG and upload them together. Use Advanced where tone matters: whether a message is teasing, annoyed or apologetic often sits in a single particle.

Transcripts, interviews and oral history

Cantonese typed as spoken turns up wherever speech is recorded: oral-history projects with older residents of Chinatowns in San Francisco, Vancouver, Toronto, Sydney or London, research interviews, focus groups, witness accounts and lawyers' notes, and subtitle or script files for Hong Kong film, television and YouTube. These documents are full of hesitations, repetitions and slang that a literal rendering turns into nonsense, and 佢 leaves the speaker's gender for the translator to infer from context. Upload the DOCX or the exported subtitle file as-is, so timing and speaker labels stay in place, and use Advanced for anything that will be cited or filed.

Comics, advertising and creative material

Hong Kong comics and web comics set dialogue in vertical balloons in colloquial Cantonese; advertising copy, shop signs, menus (茶餐廳 menus alone are a vocabulary of their own) and some song lyrics do the same. Going the other way, organisations translate English outreach into written Cantonese when they want a notice, a game script, a church bulletin or a health leaflet to read like speech rather than like a government form. That is the register this page produces: right for dialogue, wrong for a contract. Puns and slang need a reader's check either way, because a joke that works on the sound of a word rarely survives translation.

Formal Hong Kong documents: which page to use

Contracts, court documents, government letters, certificates, school reports and company papers in Hong Kong and Macau are written in standard written Chinese, in Traditional characters, even though everyone reads them aloud in Cantonese. You can upload them here and they translate normally, since the engine reads both registers. But when you are translating into Chinese for a formal Hong Kong audience, the Traditional Chinese document page produces the standard written register those documents use. Documents from Guangzhou and the rest of Guangdong are standard written Chinese in Simplified characters, and belong on the Simplified Chinese page.

What to watch for in Cantonese document translation

Most bad Cantonese translations come from reading it as Mandarin in Traditional characters. 冇 (not have) is 有 (have) with two strokes missing, so a scan error flips a sentence; 係 is the everyday word for 'is', where Mandarin writes 是; 佢 is gender-neutral, so the translator has to decide from context whether it is he, she or they. Grammar trips word-by-word rendering: the direct object comes before the indirect one (畀本書我 is 'give me the book'), comparison uses 過 (佢高過我, 'he is taller than me'), and 咗, 緊 and 過 carry the aspect that English puts into tense. Sentence-final particles (啦, 喎, 囉, 咩, 㗎) encode certainty, surprise, softening or a question, and translations either drop them, losing the tone, or spell each out into a clumsy phrase. Code-mixing is normal, with English words, abbreviations like OT and MTR, and transliterated loans (士多啤梨 for strawberry) inside Chinese sentences, and idioms are opaque: 食檸檬 is being turned down, not eating a lemon. Names need one system throughout, since Chan, Wong and Cheung in Cantonese romanisation are Chen, Huang and Zhang in pinyin. Going into Cantonese, the trap runs the other way: output that is really Mandarin (是, 不, 他, 的) reads as narration rather than speech, while slang in a semi-formal notice reads as flippant. Matching register to audience is the job, and it is what the Advanced tier is for.

Cantonese document translation FAQ

Is this Cantonese document translator free?
You can try it free, with no account needed, and signing in gives you a larger daily allowance. Pro and Scholar are there for longer documents and regular use.
Can it read screenshots, scans and photos?
Yes. Upload JPG, PNG or HEIC images, or a PDF that is really a scan, and the characters are read from the image before translating. Chat screenshots and clean print come through well. The characters specific to Cantonese are the ones to check: 冇 and 有 differ by two strokes, 嘅, 咗 and 啲 can be misread in small or stylised lettering, and comic balloons run vertically. Handwritten notes vary a lot with the hand, so treat those results as an aid rather than a finished transcription.
How accurate is it? What is the difference between Fast and Advanced?
Written Cantonese has far less published text behind it than standard written Chinese, so review the output more closely than you would a Mandarin document. Fast gives you the gist of a thread or a transcript. Advanced runs a frontier model and is the right choice for anything you will cite, file or publish: it is better at the things colloquial Cantonese lives on, such as sentence-final particles, aspect markers (咗, 緊, 過), English words mixed into Chinese sentences, and idioms like 放飛機 (to stand someone up) that mean nothing read literally.
Will the translation keep my document's original layout?
The translator produces a translated PDF, Word, PowerPoint, Excel or EPUB file that aims to preserve the original layout, including tables, columns and headers. Text size, fonts and line breaks may change, so review the result against the original. Chinese runs shorter than English, so an English file translated into Cantonese usually comes back with less text on each page.
Can I use this for immigration, court or other official purposes?
Not on its own. Immigration authorities and courts require a certified translation with a signed statement of accuracy, which machine output is not. Two further points matter for Cantonese. Official documents from Hong Kong and Macau are in standard written Chinese, so a translation prepared for an authority should be into that register, not into colloquial Cantonese. And when transcripts of Cantonese speech are used as evidence, translate them here to understand them and prepare, then have a certified translator produce the version you submit.
What comes out when I translate a document into Cantonese?
Colloquial written Cantonese in Traditional characters, Hong Kong style: 嘅 rather than 的, 唔 rather than 不, 佢 rather than 他 or 她. That is what you want for dialogue, subtitles, messages, social posts and outreach meant to sound like speech. For a contract, certificate, report or anything else a Hong Kong reader expects in formal written Chinese, use the Traditional Chinese page, whose output is the standard written register. Readers in Guangzhou understand Hong Kong-style written Cantonese, though they usually write it in Simplified characters. Going into English, upload either register and either character set.
What file types can I upload?
PDF, Word (DOCX), PowerPoint (PPTX), Excel (XLSX) and EPUB, plus JPG, PNG and HEIC images, which is how most chat screenshots and photographed pages arrive. Free and Pro accept up to 20MB and 100,000 extracted characters per file; Scholar accepts up to 50MB and 1,000,000 characters.

See “Keep formatting” in action

Real examples of the “Keep formatting” option: original on the left, translated document on the right. Click either page to enlarge.

1 / 6

Original

Translation

A civil complaint filed in Mexico City, Spanish to English. Caption block, numbered pleadings, the evidence table and the signature block all keep their places.

Original

Translation

An Italian birth certificate, photographed rather than scanned. The framed masthead, the field grid, the marginal annotations and the round seal all hold their positions.

Original

Translation

A handwritten archaeological field notebook from 1974. The engine reads the hand and keeps the ruled inventory table intact.

Original

Translation

A German hospital discharge summary. Diagnoses with ICD codes, lab values and the medication table survive the translation unchanged.

Original

Translation

A 1747 parish baptism register in Latin. Ruled columns, godparents and marginal notes stay aligned, so the genealogy still reads as a register.

Original

Translation

A scanned 17th-century Italian index of sermons. Two columns of dense small print, translated in place.

Translate documents in other languages

Mandarin Chinese documentsTraditional Chinese documentsPinyin documentsMalay (Jawi) documentsMalay documentsAkkadian documentsOld English documentsArabic documentsAramaic documentsMoroccan Arabic documentsCantonese text translatorAll document translation