Chinese-English Word Counter

Paste mixed Chinese and English text and get four separate counts — CJK characters, English words, numbers and punctuation — instead of one number that means nothing. The counting runs as regular expressions in the page, so the text never leaves your device.

Counting Two Writing Systems at Once

A word processor counts words, which is fine until you paste a paragraph that is half English and half Chinese. Then the figure it returns is either meaningless or actively misleading, because Chinese does not put spaces between the units being counted. This page stops trying to force one number out of two writing systems and reports them side by side.

• CJK characters are counted across five Unicode blocks: Unified Ideographs, Extension A, Compatibility Ideographs, Hiragana and Hangul syllables. That means Japanese kana and Korean syllables are picked up by the same counter as Chinese hanzi rather than being silently dropped.
• English words are counted by matching runs of the letters A to Z, so 'the' and 'through' each count once regardless of length.
• Numbers are counted as separate runs of digits, and punctuation as every character that is neither a letter, a digit, an underscore nor a space.
• A total token count is shown as the sum of those four figures, which is a conversational measure of length rather than a substitute for any single one of them.

What Counts as a Word, a Character, and a Token

1. Paste or type the text into the single input field. There is no file upload — the box is the only way in.
2. Press Count. Results replace whatever was shown before; there is no history and no accumulation between presses.
3. Read the four lines separately. CJK Characters is a character count, not a word count, because segmenting Chinese into words requires a dictionary and a policy about ambiguity that a browser tool cannot decide for you.
4. Use Total tokens only when you want a rough sense of volume. It adds a character, a word, a number and a punctuation mark into the same sum, so it is a length indicator, not a linguistic unit.

Where the Numbers Come From

Knowing the exact rules matters, because the counter is literal about them. Every figure traces back to one regular expression, and the cases below are the ones people run into first.

• Half-width Latin letters only. A word is a run of A–Z, so 'café' counts as one word (the é is not matched but it does not split the run), while 'naïve' counts as two, because the ï sits between two letter runs. Accented Latin text, Vietnamese and romanised transliterations can therefore look inflated.
• Chinese punctuation is counted as punctuation, not as characters. Full-width commas, corner brackets and ideographic full stops all land in the punctuation figure rather than the CJK figure.
• Numbers split on any non-digit. '3.14' registers as two numbers, and '1,000,000' as three — a comma or a decimal point is punctuation, not part of the number.
• Mixed strings split at the boundary. 'PDF转换' contributes one English word and two CJK characters, which is usually the answer someone working on a bilingual document actually wants.
• Chinese characters count one each with no segmentation, so 词语 counts as two characters. This is the standard way translation work is measured and the reason character counts and word counts are quoted separately.

Using It on a Translation Job

The split count is most useful when two numbers are being compared rather than one being quoted.

• Quoting a Chinese-to-English job: price the source by character and the target by word, and check the expansion ratio the two figures imply before committing to a rate.
• Checking subtitle and voice-over limits, where the relevant constraint is characters per line in the target language rather than words per minute.
• Spot-checking a machine translation against its source to see whether content has been dropped or duplicated — a large shift in the character-to-word ratio usually means it has.
• Billing for proofreading mixed documents, where you need a defensible breakdown of what was actually reviewed.
• Sanity-checking an abstract or metadata field that has a limit expressed in characters and was written in a mix of scripts.

Known Blind Spots

The counter is a regular-expression tool, and regular expressions cannot do everything a linguist would ask of them. These are the limitation cases worth knowing before you rely on a figure.

• No Chinese word segmentation. Each hanzi is one unit. If your contract is priced per Chinese word rather than per character, this counter gives you the wrong basis and you need a segmenting tool instead.
• No character count including spaces and no character count excluding spaces for Latin text. Only word runs are measured, so 'character count' for English content is not available here.
• No paragraph, sentence or line counts.
• Punctuation in the Latin sense includes symbols such as @ and /, so URLs and email addresses tend to inflate the punctuation figure.
• Emoji and other non-BMP characters are not counted in any of the four categories, so they will not appear in the total either.
• Full-width Latin letters and digits (Ａ, １) sit outside the A–Z and 0–9 ranges and are treated as punctuation.

Mixed-Language Counting Questions

Why is my Chinese character count different from my word processor's?

Word processors usually count Chinese by segmenting it into words, which produces roughly two thirds of the character count. This page counts characters, one hanzi at a time, because that is what translation pricing normally uses. Both numbers are defensible; they answer different questions.

Does it count Japanese and Korean as Chinese characters?

It counts them in the same category. Kana and Hangul syllables are included in the CJK figure alongside hanzi, because the shared field is 'text that is not space-delimited Latin'. If you need Japanese and Chinese separated, this is not the right tool.

Why does one accented word sometimes count as two?

Because the word pattern matches A–Z only. An accent falling between two letter runs splits them, so 'naïve' registers as two words while 'café' registers as one, since the accent comes after the final letter. For heavily accented Latin text, treat the word count as an upper bound.

Is the text I paste stored?

No. The counting happens in the page as it runs, and there is no server call, no account and no saved history. When you close the tab the text is gone, which is often the point — contracts and unpublished manuscripts are common inputs here.

Last verified: October 3, 2026

Launch Chinese-English Word Counter

All Tools · All Tools · About · How We Test · Contact · Privacy

Maintained by Huang Bin · Last updated October 3, 2026