Chinese text cleaner. 8 toggleable rules: collapse whitespace, remove inter-CJK spaces, normalize line endings, collapse blank lines, remove zero-width chars, normalize smart quotes, strip emoji, strip URLs.
Chinese Text Cleaner (中文文本清洗)
How to use
Toggle rules
8 rules, each toggleable. The first 5 are on by default; the latter 3 (quotes, emoji, URLs) are off (enable as needed).
Paste text
Cleaning is live. The stats bar shows before/after char counts and how many you saved.
Copy the result
One-click copy of the cleaned text.
Chain with other tools
This tool + simp/trad converter + full/half-width converter = complete Chinese text preprocessing pipeline.
Chinese Text Cleaning: Common Issues from OCR, Clipboard, and Scrapers
Chinese text often arrives with "dirty data" when crossing tool boundaries — extra spaces, zero-width characters (commonly injected by Word and PDF copy/paste), smart quotes, mixed full/half-width punctuation, and more. These problems are visually invisible but format-breaking, especially in:
Common sources + cleaning strategies
(1) OCR output: spaces erroneously inserted between Chinese characters (since engines recognise character-by-character). Rule: enable "Remove inter-CJK spaces".
(2) PDF / Word clipboard: often carries zero-width characters (U+200B, U+FEFF, etc.) making documents "look identical but encode differently". Rule: enable "Remove zero-width chars".
(3) Smart quotes: Word auto-converts "..." to "..." (curly quotes). Breaks JSON, SQL, HTML. Rule: enable "Normalize smart quotes".
(4) Web scrapers: often pick up extra whitespace, URLs, emoji. Rule: enable the relevant filter.
Order of operations matters
The tool processes in this order: normalize line endings → strip zero-width → strip URLs → strip emoji → normalize quotes → remove inter-CJK spaces → collapse whitespace → collapse blank lines. The sequence is designed so each rule receives a "clean intermediate state" — standard practice in rules-based text processing.
Privacy
All processing happens in your browser. No text is ever sent to our servers. Safe for confidential content.
10 Facts about Chinese Text Cleaning
Zero-width characters (U+200B, U+200C, U+200D, U+FEFF) are visually invisible. Word, PDF, and Notion frequently inject them on copy, but you can't see them by eye.
Smart quotes (curly "" '') are Word's default — pressing the straight " key gets auto-converted. They trigger syntax errors in JSON, SQL, and command-line tools.
OCR engines (like Tesseract) split by character — that's why scanned Chinese often comes out as "汉 字 之 间 有 空 格". The tool's "Remove inter-CJK spaces" rule directly addresses this.
Regex is critical for Chinese text processing. The CJK Unified range [一-鿿] (U+4E00 – U+9FFF) + Compatibility Extension [㐀-䶿] (U+3400 – U+4DBF) cover 27,000+ characters. The tool's "inter-CJK spaces" rule uses both ranges.
U+FEFF (BOM, Byte Order Mark) marks the byte order of UTF-8 files. Windows Notepad inserts it on save; Linux/Mac tools may error on it. The tool strips it along with other zero-width characters.
Emoji are "double-byte characters" in CJK text (4-8 UTF-8 bytes each). They cause truncation errors if a database column isn't wide enough. The "Strip emoji" option bulk-removes them.
Windows / Mac / Linux use different line endings: Windows = CRLF (\r\n), Mac/Linux = LF (\n), legacy Mac OS = CR (\r). The tool normalises everything to LF — the cross-platform standard.
"Line break" and "paragraph break" are different. A paragraph can have multiple soft line breaks (\n). A blank line (\n\n) marks paragraph separation. The "Collapse blank lines" rule preserves one blank line for paragraphs while removing excess.
"Regular space" and "full-width space" (U+3000) are different characters. The "Collapse whitespace" rule handles both — merging any whitespace (regular spaces, tabs, full-width spaces, non-breaking spaces) into a single regular space.
Pairs with RT-CHN-037 (Simp/Trad), RT-CHN-038 (Full/Half-Width), and RT-CHN-039 (Vertical Layout) — the complete Chinese text-processing toolset.
Frequently Asked Questions
-
Fully local. All processing is in-browser; no text is uploaded. Even confidential content is 100% safe.
-
The first 5: collapse whitespace, remove inter-CJK spaces, normalize line endings, collapse blank lines, remove zero-width chars. These are side-effect-free "safe cleanup". The latter 3 (quotes, emoji, URLs) are opt-in — they may change meaning, so off by default.
-
No. The rule is strictly limited to spaces between two CJK characters — so "汉 字" becomes "汉字", but "I love 中文" is preserved.
-
Preserves one blank line (paragraph separator). Removes excess (3+ blank lines become 1). Preserves paragraph structure while removing redundancy.
-
Curly "" → straight " "; curly '' → straight ' '. Word's "smart quotes" are a common source of JSON and SQL errors — this rule fixes that.
-
Covers major emoji ranges: U+1F300-U+1FAFF (main emoji block), U+2600-U+27BF (symbols + dingbats), U+1F600-U+1F64F (faces). A few edge-case emoji may slip through.
-
Strongly recommended. Zero-width chars, mixed spaces, and smart quotes all cause mysterious "search doesn't match" bugs. Production text input should always be sanitised this way.
-
The original is preserved in the left input box (never overwritten). You can edit, toggle rules, and compare before/after at any time. The result (right) is read-only.
-
Enable everything. PDF clipboard typically carries: zero-width chars, misplaced spaces, smart quotes, possibly emoji/URLs. The tool's "all on" config is ideal for PDF cleanup.
-
Only the first 5 "side-effect-free" rules. Leave "Normalize quotes" off (might alter string literals); leave "Strip emoji/URLs" off (may contain code-relevant content).
Pick up where you left off
Stored only in this browser — never sent to our servers.