Jyutping is the romanization scheme published by the Linguistic Society of Hong Kong. Adding it to a piece of Cantonese text means giving every character its reading — and the difficulty is not the lookup, it is that a character's reading depends on the word it sits in.
This page covers what actually goes wrong, how to do it by hand, and how to do it in one step.
The part that trips people up: one character, several readings
Many characters are read one way alone and another way inside a word. A converter that looks up one character at a time gives you the first, and the failures are not exotic — they are ordinary vocabulary:
| Character | On its own | In a word |
|---|---|---|
| 生 | sang1 | 學生 hok6 saang1 |
| 聽 | ting3 | 聽日 ting1 jat6 |
| 企 | kei5 | 屋企 uk1 kei2 |
| 行 | hong6 | 行路 haang4 lou6 |
| 話 | waa6 | 廣東話 gwong2 dung1 waa2 |
Both columns come from this site's own converter — the left by asking it for the single character, the right by asking it for the word. If you are testing a converter, those five words separate word-aware tools from lookup tables quickly.
Doing it by hand
Worth knowing even if you use a tool, because it is how you check one.
- Split the text into words rather than characters. 廣東話 is one word, not three lookups.
- Look each word up in a dictionary that gives Jyutping for the word — CC-Canto and the CantoDict-derived dictionaries do. A per-character chart will not.
- Write the reading under each character, keeping the alignment: the whole point is knowing which syllable belongs to which character.
- Check the tone numbers 1-6 against the six Cantonese tones. Tone is not decoration: 詩 si1, 史 si2, 試 si3, 時 si4, 市 si5 and 是 si6 are six different words spelled identically apart from the digit.
Doing it in one step
Paste the text, choose Jyutping, and press one of the export buttons. The romanization is aligned under each character and an English translation goes underneath, so the sheet is usable for study rather than being a wall of syllables.
Word segmentation is done with pycantonese, so the five words above come out with their in-word readings rather than their citation readings.
- Traditional and Simplified input both work; set the target language to Chinese (Hong Kong) if you also want the characters converted.
- Export as PDF to print, DOCX to keep editing, Markdown for notes.
- Yale and Yale with tone numbers are available if your course uses them.
Common questions
- Is Jyutping the same as Yale?
- No. They are different schemes for the same sounds. Jyutping marks tone with a digit 1-6 after the syllable; Yale uses diacritics plus an h for the lower register. 香港 is hoeng1 gong2 in Jyutping and hēung góng in Yale.
- Does it work on a photo of a page?
- Yes — a photo or a PDF. The text is extracted first, then romanized. Check the result, because extraction from a photo of printed Chinese is the step most likely to introduce errors.
- Is the romanization always right?
- It reads the word rather than the character, so it handles ordinary vocabulary — including the examples above — well. Unusual names and rare phrases are where segmentation can pick the wrong reading, so those are worth a second look.