Mixed Chinese-English Word Count
Word count and character count side by side · Chinese, English, digits and punctuation broken down · nothing is uploaded
📖 How to use it
- Paste mixed text and four headline numbers appear: word count, character count, Chinese characters, English words.
- Use the breakdown to decide which number applies: submission systems usually mean characters, coursework usually means words.
- The gap formula below lists, item by item, why the two numbers differ — with your own values already filled in.
- If a target asks for "no more than N words", read the English word row; if it says "N characters", read the character row.
❓ Frequently asked questions
Q: Why is the word count so much lower than the character count?
A: They count different units. Under the word-count measure a Chinese character is 1 and an English word is 1; under the character-count measure every non-space character is 1. An eight-letter English word is 1 in one and 8 in the other — a gap of 7. The more English your text contains, the wider the gap grows.
Q: How are Japanese, Korean or Russian texts handled?
A: Chinese characters count as Chinese and Latin letters count as English words; kana, Hangul and Cyrillic are neither, so they are grouped as "other scripts" and counted character by character in their own row. They are never folded into the Chinese character count, which would otherwise inflate it.
Q: Is my text uploaded?
A: No. The counting logic runs in the page, the input never leaves your device, and there is no upload endpoint at all. Open the Network panel in your dev tools to confirm.
🧮 The two formulae, and where the difference lives
Let a text contain H Chinese characters, P CJK punctuation marks, W English words, L Latin letters, N digit runs, D digit characters, S ASCII punctuation marks, E emoji and O other-script characters. Then:
word count = H + P + W + N + E + O
character count (no spaces) = H + P + L + D + S + E + O
characters − words = (L − W) + (D − N) + S
That last line is the important one: the difference comes only from three places — counting English letter by letter instead of word by word, counting digits one by one instead of run by run, and whether ASCII punctuation is included. Chinese characters, CJK punctuation, emoji and other scripts contribute equally to both measures, so they never affect the gap. The formula on this page plugs your current text into exactly those terms.
This also explains something you have probably seen: write "性能提升 20%" and then "performance improved by 20 percent" and the character counts come out similar, but under the word-count measure the English version splits into many more units — and the two measures diverge sharply. Before submitting a length-limited document, confirm which measure the recipient means.
🧾 Worked example
Step by step, using the sample text:
Python 3.12 的 GIL 移除后，CPU 密集型任务的性能提升约 20%（官方数据）。
| Step | Value | Derivation |
|---|---|---|
| Chinese characters H | 19 | Excluding Python / GIL / CPU / 3.12 / 20 |
| CJK punctuation P | 4 | ，。（）— the brackets are full-width |
| English words W / letters L | 3 / 12 | Python, GIL, CPU — 12 letters in total |
| Digit runs N / digits D | 2 / 5 | 3.12 (3 characters) and 20 (2 characters) |
| ASCII punctuation S | 2 | The decimal point and the percent sign |
| Spaces | 6 | Between Chinese and English, and between words — counted only by the "with spaces" figure |
| Word count = 19 + 4 + 3 + 2 | 28 | H + P + W + N |
| Character count = 19 + 4 + 12 + 5 + 2 | 42 | H + P + L + D + S (excluding the 6 spaces) |
| Gap = (12 − 3) + (5 − 2) + 2 | 14 | 9 + 3 + 2 = 14, matching 42 − 28 — the formula is self-consistent |
| Total characters (with spaces) | 48 | 42 + 6 spaces, which is also the grapheme count here |
So the same sentence is "28 words, 42 characters, 48 with spaces". If a system allows "no more than 40 characters", 42 is over the line while 28 is comfortably inside — which is exactly why mixed text needs the measure stated up front. Extrapolating at a 20% English share, a 3,000-word mixed document can differ by over a thousand between the two measures.
🎯 When this is useful
- Word-count checks for study-abroad essays, English abstracts and bilingual CVs (read the English word row when the requirement says "words").
- Checking a document against a submission or registration system's character limit.
- Translation and localisation quotes: confirm the other side's measure before quoting per thousand characters or words.
- Subtitle and voice-over work, where line length and total length both matter.
- A rough reference when estimating tokens for a prompt — more characters means more tokens, though not linearly.
- Not suited to: predicting a specific model's token count exactly. Tokenisation depends on the model's vocabulary and algorithm; character counts are only a rough guide.