Chinese stroke order
Paste Chinese text and get per-character stroke counts, stroke names in order, stroke shapes in order and the turn count
要拆分笔顺的汉字
一次可拆一整段（单次上限 5 万字符）；夹带的英文、数字、标点按「非汉字」处理。
每一笔怎么显示
拆分结果
等待输入
逐字对照表固定列出全部字段，与上面的显示模式无关
| 汉字 | 笔画数 | 逐笔笔画名称 | 逐笔笔画形状 | 折笔数 |
|---|---|---|---|---|
| 还没有内容 | ||||
对照表最多显示前 2000 个汉字，完整的用「下载 CSV」。
笔画符号对照表
下表由笔顺库自带的笔画字典现场生成（共 — 类），不是手工抄录。禁用 JavaScript 时这张表不显示，拆分结果不受影响。
| 笔画名称 | 形状 | 平笔 / 折笔 | 折几次 | 说明 |
|---|---|---|---|---|
| 笔画字典未能加载 | ||||
「共用符号」那一列就是本页唯一的含糊处：这 5 组笔画在库内用同一个符号记录，所以拆到这些笔画时会并列写成「斜钩 / 卧钩」，本工具不做猜测取舍。
数据来源、规则与已知边界
数据来源：笔顺数据来自开源库 cnchar 3.2.6 的笔画顺序插件 cnchar-order 3.2.6（MIT 许可证，作者 theajack），随本站自托管，不引用任何 CDN。库内含 6939 个规范汉字的笔顺（数字由库内数据现场核对得出）。
逐笔名称与形状：每个汉字拆成一串笔画，每笔给出名称（如「横折钩」）与形状符号（如「𠃌」）。「折笔数」统计这一字里有几笔被库记作折笔，判断依据是库里逐笔记下的平笔 / 折笔标记与折数，不是本页自己定的规则——注意库里把竖钩、弯钩这类钩画也记作折笔，所以「我」的折笔数是 2（竖钩 + 斜钩），不是 1。
「点2」按「点」显示：库里把点分成了两类，其中叫作「点2」的那一类（形状记作 ㇀）在 6939 字里有 998 字用到，心、必、办、写、军、协的第一笔都是它。日常没人管它叫「点2」，所以拆分结果里统一显示成「点」；符号对照表里两种记法都原样列出并注明对应关系，笔画数与形状一个都不改。
五组合并笔画：斜钩与卧钩、横撇与横钩、横折折与横折弯、横折折折钩与横撇弯钩、竖折撇与竖折折，在库里共用同一个符号，无法区分。拆到这些位置时本页并列写成「斜钩 / 卧钩」，宁可能见两个候选，也不替你猜一个。
不提供「五类笔画 / 笔顺编号」数字：横竖撇点折五类归并在竖钩、弯钩这类钩画上存在两种通行口径（一种把竖钩计入竖类，一种按「有转折就算折」计入折类），而本页的数据用的是后一种标记。与其输出一个看着像国标、实际口径不明的编号串，不如不输出——需要五类或编号的场合请按你所在场景（教材、GB 13000.1、输入法）的口径自行归类。
查不到的字：繁体字（國、說）、CJK 扩展区生僻字（𠀀）以及库里没收录的字，会在结果里明确标成「库内无笔顺数据」，绝不按形近字凑一个结果。繁体字请先用下面的「简繁转换」转成简体再拆。
不做笔顺动画与写字练习：动画需要每一笔的矢量轮廓数据（体积是本页笔顺库的数十倍，且多数开源字形的授权是 CC BY-SA 系，随商用页面分发要连带开源整站）。本页只做「拆得清楚、可以复制与导出」这一件事。
教学与考试场景请再核一遍：笔顺规范在不同教材、不同年份的字表之间偶有差异，开源词库也难免有错漏。给孩子辅导作业、参加汉字听写大会这类场合，请以教材与《现代汉语通用字笔顺规范》为准，本页结果作为快速参考。
FAQ
From the cnchar-order plugin of the open-source library cnchar 3.2.6 (MIT license, author theajack), self-hosted by this site with no CDN and no external requests anywhere on the page. The dictionary covers stroke order for 6,939 standard hanzi, each stroke carrying its name (e.g. horizontal-turn-hook), its shape glyph (𠃌) and a flat / turning marker. It is a community-maintained wordlist: curated by the author yet still possibly flawed. For homework help or dictation contests, defer to textbooks and the official 'Specification of Standard Stroke Order for Modern Chinese Common Characters' - treat this page as a quick reference.
Because five stroke pairs share one symbol in the dictionary and cannot be distinguished: slanting hook vs lying hook, horizontal-turn vs horizontal-hook, horizontal-turn-turn vs horizontal-turn-bend, horizontal-turn-turn-hook vs horizontal-sweep-bend-hook, vertical-turn-sweep vs vertical-turn-turn. At those positions this page lists both candidates side by side (shape likewise written as ㇂ / ㇃) without guessing - the stroke-symbol table above lists every shared mapping. The vast majority of common strokes (horizontal, vertical, left-falling, dot, horizontal-turn, vertical hook...) are unique and unaffected.
Deliberately not done. Even the five-class merge has two accepted readings on hooks: one counts vertical-hook under vertical, another follows 'anything with a turn counts as turning' - and this page's data uses the latter marking. Likewise the numeric 'stroke codes' used by different systems differ (some merge horizontal strokes, some keep sweep and dot separate). Outputting a code string that looks official but whose rule cannot be explained misleads more than saying nothing, so this page gives only per-stroke names, shapes and turning counts; do the five-class grouping and coding yourself per the convention of your context.
No. The dictionary holds only simplified standard hanzi; traditional forms (國, 說, 筆), rare characters in CJK extension blocks (𠀀, 龘, 靐) and anything not in the dictionary are marked 'no stroke data in dictionary'. For traditional forms, you can first convert them to simplified with our 'simplified/traditional converter' and then decompose (note simplified-traditional is one-to-many, the converted character may not be the original one); for extension-block rarities, printed references like the Great Chinese Dictionary remain the only option for now.
No upload. The stroke dictionary is a self-hosted script (cnchar core 77KB + order plugin 143KB, loaded once on first use, then served by browser cache); decomposition runs entirely in your browser - text never leaves your machine, refreshing clears it, the page makes no network requests and there is no third-party CDN. Up to 50,000 characters per run; measured, decomposing over two thousand characters takes tens of milliseconds. Results can be copied or downloaded as .txt, and the per-character table also exports to CSV (with UTF-8 BOM, so Excel opens it with a double click without mojibake).