Rare Chinese character reader. Enter one or more Hanzi to see pinyin, Zhuyin, Cantonese Jyutping, stroke count and meaning. Where the reading libraries have nothing, it shows the Unicode Han Database (Unihan 18.0) value and says so, and where nobody records a reading it says that too. Runs in your browser.
Rare Chinese Character Reader
Data: Unihan 18.0, © Unicode, Inc., used under the Unicode License v3. English definitions: CC-CEDICT (MDBG), CC BY-SA 4.0.
How to use
Type or paste the character
Put one or more Hanzi in the box. If you cannot type one, copy it from a web page, PDF or document. Every CJK Unified Ideographs block, Extension A through J, is recognised.
Click "Look Up"
Each character gets its own card; repeats are shown once. Ctrl/⌘ + Enter also works.
Read the readings and where they came from
The card lists pinyin, Zhuyin, Jyutping and a stroke count. A value tagged "Unihan" comes from the Unicode Han Database, not the everyday reading libraries. "No reading recorded in Unihan" means none of the sources has one.
Check the code point and meaning
Under the glyph is its code point and block (for example U+2EBF0 · Ext I). The meaning comes from CC-CEDICT, or from Unihan's English definition where CC-CEDICT has no entry.
Why a rare character so often has no reading on file
Unicode 18.0, dated 16 September 2026, encodes 101,997 CJK unified ideographs across the original block and Extensions A to J, and 74,401 of them sit outside the Basic Multilingual Plane, where older software often breaks. A character that displays is not a character anyone has recorded a reading for. The pinyin library this tool uses, pinyin-pro, reads 79 of the 6,592 characters in Extension A, and when it has no data it hands the character straight back — which is how an earlier version of this page came to print the glyph itself in the Pinyin row.
The tool now asks pinyin-pro (Mandarin) and to-jyutping (Cantonese) first. Where either has nothing, it shows the value recorded in Unicode's own Han Database, Unihan 18.0 — kMandarin, kTGHZ2013, kXHC1983 or kHanyuPinyin for Mandarin, kCantonese for Cantonese — tagged "Unihan" on the card. Where neither has anything, it says "No reading recorded in Unihan" rather than guess. Mandarin coverage by block:
| Block | Characters | pinyin-pro reads | With Unihan |
|---|---|---|---|
| URO (4E00–9FFF) | 20,992 | 20,856 | 20,924 |
| Ext A (3400–4DBF) | 6,592 | 79 | 5,788 |
| Ext B (20000–2A6DF) | 42,720 | 36 | 14,618 |
| Ext I (2EBF0–2EE5F) | 622 | 0 | 99 |
| Ext J (323B0–3347F) | 4,298 | 0 | 0 |
| All twelve ranges | 101,997 | 21,131 | 44,362 |
Extension J, added in Unicode 17.0 with 4,298 characters, has no reading at all in Unihan. Extension I's 622 characters were added in 2023 as "urgently needed", and Unihan tags every one of them with the source GIDC23, which UAX #38 defines as "ID system of the Ministry of Public Security of China, 2023". Unihan records a Mandarin reading for 99 of them. So a character can be encoded and display perfectly and still have no reading in any source this tool checks — Unihan, or its three libraries, pinyin-pro, to-jyutping and CC-CEDICT. For a name, ask the person.
Where each row comes from
Pinyin comes from pinyin-pro, and the Zhuyin is spelled, syllable by syllable, from the pinyin shown. Jyutping comes from to-jyutping, whose repository does not say where its pronunciation data comes from. The gaps are filled from Unihan 18.0. Unicode's own documentation for it, UAX #38, describes kMandarin as the most customary reading, intended for collation and transliteration, which may not match a given dictionary; where kMandarin has two values the first is the mainland preference and the second Taiwan's, and the card shows both. pinyin-pro gives mainland readings, so Zhuyin spelled from it is not necessarily the Taiwan dictionary reading. Stroke counts come from hanzi-writer's stroke-order data — a pen count, available for 9,565 characters — and otherwise from Unihan's kTotalStrokes, which counts by IRG convention and can differ from a pen count by a stroke or two, so the card labels it "IRG count". Meanings come from CC-CEDICT, or Unihan's kDefinition where CC-CEDICT has no entry.
The lookup runs in your browser. The page downloads its data from this site: one dictionary file, and Unihan shards of 2,048 code points each — only the ones covering the characters you enter. Our server logs can therefore show which range of code points was looked up, never the characters you typed.
What Unicode records about rare characters, and what it leaves blank
Of Unicode 18.0's 101,997 unified ideographs, 74,401 (72.9%) sit outside the Basic Multilingual Plane: each takes a surrogate pair in UTF-16 and four bytes in UTF-8.
All 622 characters of Extension I carry the Unihan source tag GIDC23, which UAX #38 defines as "ID system of the Ministry of Public Security of China, 2023". Unihan records a Mandarin reading for 99 of them and an English definition for none.
龘 is three 龍 (dragon) stacked, 48 strokes by Unihan's count. The two databases describe it differently: CC-CEDICT calls it an old variant of 龖 (dá), which it glosses "a vista of a dragon in flight"; Unihan's kDefinition reads "the appearance of a dragon walking".
嫑 is 不 written over 要. CC-CEDICT reads it biáo, "(dialect) don't", a contraction of 不要; Unihan's kMandarin gives báo. One character, two sources, two readings — which is why each card says where its reading came from.
Cantonese is the stronger side here: of Extension A's 6,592 characters, to-jyutping has a Jyutping reading for 6,582, while pinyin-pro has Mandarin for 79.
焱, 淼 and 矗 are each one component written three times. CC-CEDICT lists 焱 (yàn) as a variant of 焰 (flame), glosses 淼 (miǎo) "a flood; infinity" and 矗 (chù) "lofty; upright".
China's 通用规范汉字表, published by the State Council in 2013, lists 8,105 characters in three levels: 3,500 in level 1, and 1,605 in level 3, which per Unicode's description are used for personal names, place names and science and technology.
There are two stroke counts. hanzi-writer counts pen strokes; Unihan's kTotalStrokes follows the IRG's conventions. Of the 9,565 characters both cover, they agree on 9,104 — 之 is 3 strokes by pen and 4 by the IRG.
Frequently Asked Questions
-
Every CJK Unified Ideographs block in Unicode 18.0 — the original block and Extensions A to J, 101,997 characters — plus compatibility ideographs, which are looked up as the unified character they normalise to. Unassigned code points inside those blocks, punctuation and other scripts are skipped, and the page says so under the results.
-
Because neither pinyin-pro nor Unihan has a Mandarin reading for it. Of Unicode 18.0's 101,997 unified ideographs, 44,362 have one in either; none of Extension J's 4,298 does. It does not mean the character has no pronunciation — only that the sources this tool checks do not record one: pinyin-pro and Unihan for Mandarin, to-jyutping and Unihan for Cantonese. Where CC-CEDICT has an entry, its reading appears with the meaning. For a name, ask the person whose name it is.
-
It is Unicode's own record, but UAX #38 says plainly that kMandarin is the most customary reading chosen for collation and transliteration, and may not match a given dictionary. Where kMandarin is empty the tool uses, in turn, kTGHZ2013 (Tongyong Guifan Hanzi Zidian), kXHC1983 (Xiandai Hanyu Cidian) or kHanyuPinyin (Hanyu Da Zidian); hover the tag to see which field it came from. For names and places, an authoritative dictionary or the person concerned is the final word.
-
Yes. Paste several characters into the box and the tool builds one card per character; repeats are shown once.
-
The Pinyin row shows pinyin-pro's main reading first and then the others it records (行: xíng, also háng · hàng · héng), with a Zhuyin spelling for each. The meaning area groups CC-CEDICT's senses by reading so you can pick by context. The Jyutping row shows to-jyutping's single reading.
-
hanzi-writer's stroke-order data covers 9,565 characters; for the rest the tool shows Unihan's kTotalStrokes, which counts by IRG convention and can differ from a pen count by a stroke or two (之 is 3 by pen, 4 by the IRG). Neither is the Kangxi count that name-numerology calculations use.
-
A form is shown only when CC-CEDICT and Unihan both treat the two as a traditional–simplified pair. A variant is not one: 俊 no longer lists its old form 㑺 as "Traditional" — an error in this tool's earlier dictionary, corrected on 3 Nov 2026. Look up 㑺 itself and the meaning says it is an old variant of 俊.
-
No. The lookup runs in your browser and what you type, names included, is not sent to our servers. The page does download data files from this site; the Unihan ones are split into shards of 2,048 code points and only the shards covering your characters are fetched — so our logs can show which range of code points was looked up, never the characters themselves. Your browser may cache these files like any other.
-
Yes, as long as you can copy it — select it in a web page, PDF or document and paste it in. If it shows as an empty box, that is your font lacking the glyph; the character itself can still be looked up.
Method & sources
How it computes
Characters are accepted from every CJK Unified Ideographs block in Unicode 18.0, Extension A to J (Blocks.txt × PropList Unified_Ideograph; 101,997). Pinyin: pinyin-pro; Jyutping: to-jyutping; strokes: hanzi-writer; meaning: CC-CEDICT. Where one has nothing, the Unihan 18.0 value is shown and labelled; where none has a reading the card says so. Nothing is generated.
What this tool implements
- Four sources, reported separately rather than merged. A character can have a well-attested pinyin and no Cantonese reading on file, and showing that honestly is more useful than filling the gap.
- Pinyin is shown with tone marks and 注音 alongside, because the two notations serve different readerships and converting between them loses nothing.
- Jyutping follows the Linguistic Society of Hong Kong scheme — alphanumeric with a tone digit, no diacritics.
- A character with more than one reading has more than one shown. Rare characters are exactly where a single "the" pronunciation is most likely to be wrong.
Sources
- pinyin-pro v3.28.1 (MIT): https://github.com/zh-lx/pinyin-pro — the Mandarin pinyin and 注音 conversion, vendored locally with no runtime fetch.
- to-jyutping by CanCLID, BSD-2-Clause: https://github.com/CanCLID/to-jyutping — the Cantonese romanisation. ⚠️ Its repository does not document which dictionary its pronunciation data comes from, so that source cannot be named here without guessing.
- Jyutping scheme: Linguistic Society of Hong Kong, 1993, https://lshk.org/jyutping-scheme/
- Stroke counts come from the Hanzi Writer data set, which derives from Make Me A Hanzi and ultimately from fonts released by Arphic Technology in 1999 under the Arphic Public License.
- The Unicode Consortium. The Unicode Standard, Version 18.0.0 (16 September 2026). https://www.unicode.org/versions/Unicode18.0.0/ — Blocks.txt and PropList.txt (Unified_Ideograph) define which code points the page accepts.
- Unicode Han Database (Unihan), Unicode 18.0.0: Unihan_Readings.txt and Unihan_IRGSources.txt (2026-07-31), documented in UAX #38, tr38-41, ed. Ken Lunde. https://www.unicode.org/reports/tr38/ — © 1991-2026 Unicode, Inc., used under the Unicode License v3, https://www.unicode.org/license.txt
- CC-CEDICT, release of 2026-05-24, community-maintained and published by MDBG, https://www.mdbg.net/chinese/dictionary?page=cc-cedict — the English definitions, Creative Commons Attribution-ShareAlike 4.0 International, https://creativecommons.org/licenses/by-sa/4.0/. The tool's derived dictionary file is CC BY-SA 4.0 as well.
What can make this go out of date
- The shards hold only what pinyin-pro 3.28.1, to-jyutping and hanzi-writer-data lacked when they were built. Upgrading a library means rebuilding; until then the library is still asked first, so a stale shard can miss a gap but never override a library reading.
- Unicode adds ideographs with each release (Extension J arrived in 17.0, one more Extension D character in 18.0). A character from a later version is reported as not assigned until the table is regenerated.
Pick up where you left off
Stored only in this browser — never sent to our servers.