✓ 本地优先 · 数据不出本机Local-first · your data never leaves

论文级动态图表
一次规格,处处复现
Publication-Grade
Motion Charts

把 CSV / TSV / XLSX / JSON 变成可直接投稿、答辩与汇报的动态图表和文档。60 种图表、8 种产物、6 套出版规格——全部在本机完成。 Turn CSV / TSV / XLSX / JSON into publication-ready animated charts and documents. 60 chart types, 8 output formats, 6 publication presets — all rendered on your machine.

60种图表chart types
11大图表族families
8种产物formats
923自动化测试tests
动态排名条形图示例 / Bar race demo
零上传Zero upload全部本地渲染rendered locally
SVG · PNG · GIF · MP4SVG · PNG · GIF · MP4PDF · PPTX · DOCX · bundlePDF · PPTX · DOCX · bundle
为什么是本地优先Why local-first

未发表的数据,不该离开你的电脑Unpublished data should not leave your desk

读取、清洗、统计、渲染、导出,全在当前设备完成。没有云端排队,没有账号,没有遥测。Parsing, cleaning, statistics, rendering and export all happen on-device. No cloud queue, no account, no telemetry.

未发表数据不出本机Unpublished data stays put

环境监测、医疗、政务数据常有硬性要求。整条链路本地执行,只有你主动提供数据直链时才会联网取数,并记录来源与哈希。Monitoring, medical and government data often can't leave the machine. The whole pipeline runs locally; the only network fetch is a data URL you explicitly provide, logged with source and hash.

复现是确定性的Reproducible by design

同一份规格 + 数据哈希 + 渲染命令,产出逐字节一致的结果。每次渲染都写入 manifest,可双机比对、可归档、可交给合作者复现。The same spec, data hash and command yield byte-identical output. Every render writes a manifest — compare across machines, archive it, hand it to a collaborator.

断网也能干活Works fully offline

装机完成后,从数据到成稿整条链路离线可用,不依赖任何云端服务,也不会因为服务端升级而悄悄改变出图结果。After setup, the entire data-to-deliverable pipeline works offline. No cloud service, and no silent server-side upgrade quietly changing your figures.

能力全景Capabilities

从数据到成稿,一套工具走完From raw data to final figure, in one tool

不是又一个网页图表库,而是把科研出图真正需要的表达、规格与复现,做进了同一条命令行链路。Not another web chart library — the expression, specs and reproducibility that research figures actually need, wired into one CLI pipeline.

60 种图表 · 11 大族60 charts · 11 families

排名、趋势、占比、流向、分布、层级、极坐标、地图、文本——每种都带 demo 数据、配色与风格预设。Ranking, trend, composition, flow, distribution, hierarchy, polar, maps, text — each with demo data, palettes and style presets.

8 种产物形态8 output formats

SVG、PNG、GIF、MP4、PDF、PPTX、DOCX,外加可归档的复现包 bundle——一次渲染任意组合。SVG, PNG, GIF, MP4, PDF, PPTX, DOCX, plus an archival bundle — combine any of them in one render.

四档清晰度Four resolutions

4K / 2K / 1080p / 720p 只改变像素密度,不动排版、字号、线宽与留白——切换档位无需重排。4K / 2K / 1080p / 720p change pixel density only — never layout, type, line weight or margins. Switch without re-laying out.

统一动效内核One motion engine

入场错峰 + 时间轴推进,按进度分帧;GIF、MP4、关键帧横排共用同一帧序列,动效与静帧永远一致。Staggered entrances plus timeline progression, framed by progress. GIF, MP4 and storyboards share one frame sequence — motion and stills always agree.

Σ
统计表达Statistical expression误差棒(对称 / 非对称 / CI)、置信带、显著性括号与星号、样本量 n、拟合公式与 R²、阈值带。Error bars (symmetric / asymmetric / CI), confidence bands, significance brackets, n=xx, fit equations with R², threshold bands.
A4
出版规格Publication presetsNature 单/双栏、Science、学位论文、PPT 16:9、社媒方图,共 6 套;等宽缩放后套用字号线宽下限。6 presets: Nature single/double column, Science, thesis, PPT 16:9, social square — scaled to width, with type and line minimums.
◎
可访问性审计Accessibility audit四种色觉模拟 + 灰度可分性 + WCAG 对比度;60 图默认配置零致命项,可切换灰度输出。Four color-vision simulations, grayscale separability and WCAG contrast; zero fatal issues across 60 charts, with a grayscale mode.
↺
溯源与复现Provenance & replaymanifest 记录输入哈希、渲染参数、每个产物字节与 SHA-256,图脚可打来源与规格指纹。Manifests log input hashes, render parameters and per-file SHA-256; footers can carry source and spec fingerprint.
图表家族Chart families

60 种图表,按族选用不迷路60 charts, organised into families

先决定“要说明什么关系”,再选图。每种图表都附带 demo 数据与配色,可先验证效果再替换为自有数据。Decide what relationship you want to show, then pick a chart. Every type ships with demo data and palettes — validate the look, then swap in your data.

01

排名与比较Ranking & Comparison

动态排名条形图、条形图、柱状图、对比条形、折线排名、棒棒糖、子弹、瀑布、评分环、液面图Bar race, bar, column, comparison bar, line ranking, lollipop, bullet, waterfall, score ring, liquid fill

02

趋势与时间Trend & Time

动态折线、双轴折线、堆叠面积、河流图、K 线、时间线(横/竖)、甘特图Animated line, dual-axis line, stacked area, streamgraph, candlestick, timeline (h/v), Gantt

03

结构占比Composition

饼图、环形、玫瑰图 6 变体、堆叠柱、漏斗、矩形树图、圆堆、旭日、极坐标柱、沃罗诺伊Pie, donut, 6 rose variants, stacked column, funnel, treemap, circle packing, sunburst, polar bar, Voronoi

04

关系与流向Relation & Flow

气泡图、散点图、桑基图、弦图、平行坐标、维恩图、力导向图Bubble, scatter, Sankey, chord, parallel coordinates, Venn, force graph

05

前后对比Before–After

斜率图、蝴蝶图、人口金字塔、哑铃图、发散条形Slope, butterfly, population pyramid, dumbbell, diverging bar

06

量率同屏Dual-axis

折线柱组合图,双轴刻度严格对齐、网格不错位Combo line + column with strictly aligned dual axes and no grid drift

07

分布形态Distribution

直方图、箱型图、热力矩阵Histogram, box plot, heatmap matrix

08

层级结构Hierarchy

树状图,间距按实测标签宽度反算,避免文字交叠Tree diagram with spacing derived from measured label widths to avoid collisions

09

极坐标 / 径向Polar & Radial

雷达图、极坐标折线、径向折线、径向柱(玉珏)Radar, polar line, radial line, radial bar

10

地图Maps

热力地图、气泡地图、流向地图,均含全国 / 分省版Heat, bubble and flow maps, in national and provincial variants

11

文本Text

词云,字号与布局按词频与实测宽度排布Word cloud with size and layout driven by frequency and measured widths

→

想直接看效果?Want to see them all?

打开画廊,按族筛选、点击放大看细节。Open the gallery to filter by family and zoom in.

进入画廊Open gallery
科研级的严谨Research-grade rigor

审稿人要的那几样,图里本来就有The things reviewers ask for are built in

误差与显著性、期刊尺寸、色盲友好、灰度打印——不必再导出到别的软件里补画。Uncertainty and significance, journal sizes, color-vision safety and grayscale printing — without exporting to another tool to patch them in.

  • 误差与不确定性:对称 / 非对称误差棒、置信带、样本量标注Uncertainty: symmetric / asymmetric error bars, confidence bands, sample-size labels
  • 显著性标注:括号与星号自动分层,峰值组占顶层、同层区间不重叠Significance: brackets and stars auto-tiered, peaks on top, no overlap within a tier
  • 拟合与阈值:1–4 次多项式,图内给出方程与 R²;参考线、阈值带(如 AQI 分级)Fits & thresholds: polynomial fits with equation and R²; reference lines and threshold bands
  • 出版尺寸:6 套预设,按画布宽度等比缩放并套字号线宽下限Publication sizes: 6 presets, scaled to canvas width with type and line minimums
  • 可访问性:色觉模拟、灰度可分性、WCAG 对比度;--grayscale 灰度输出Accessibility: color-vision simulation, grayscale separability, WCAG contrast; --grayscale output
带误差与分布的箱型图示例 / Box plot with distribution

示例 · 箱型图:四分位区间、离群值与样本量 nExample · box plot: quartile ranges, outliers and sample size n

领域场景包Domain packs

五个领域,一整套口径Five domains, one consistent vocabulary

术语、单位、精度、阈值、推荐图组、可直接渲染的典型规格,以及该领域常见审稿意见——一条命令套用。Terminology, units, precision, thresholds, recommended chart sets, ready-to-render specs and common reviewer comments — applied with one command.

eco

生态环境Ecology

颗粒物源解析、空气质量、达标评价PM source apportionment, air quality, compliance

查看样例图View samples→
pubhealth

公共卫生Public health

发病率、监测、干预效果Incidence, surveillance, intervention effects

查看样例图View samples→
econ

经济金融Economy

增速、结构、区域对比Growth, structure, regional comparison

查看样例图View samples→
energy

能源双碳Energy

能耗、排放、结构转型Consumption, emissions, transition

查看样例图View samples→
edu

教育科研Education

招生、就业、学科评估Admissions, employment, evaluation

查看样例图View samples→

套用示例:Example: motionchart init --domain eco --chart heatmap --data x.csv

选图与上手Choosing & getting started

不知道用什么图?让它先给建议Not sure which chart? Ask it first

三个入口,从“这堆数据该画什么”到“图已经出好了”。Three entry points, from "what should I even chart?" to "the figure is done".

try

数据体检推荐Data triage

motionchart try 数据.csv 识别列类型与时间字段,推荐 Top 3 图表,直接出并排对比图与三份可渲染 spec。motionchart try data.csv detects column types and time fields, recommends top 3 charts, renders a side-by-side comparison and three ready specs.

wizard

问答式生成Guided wizard

motionchart wizard 一步步问出意图,每步回车即走默认值——不必记任何参数。motionchart wizard walks through intent step by step; press Enter to accept defaults — no parameters to memorise.

gallery

单页选择器Single-page picker

motionchart gallery 生成按族分组的动态预览、选图决策树与能力对比表,本地打开即用。motionchart gallery builds a family-grouped preview with a decision tree and capability table — open it locally.

承诺清单The promise, itemised

  • 数据读取 / 清洗 / 统计 — 本机,不联网Parse / clean / stats — on-device, offline
  • 渲染 / 导出 / 保存 — 本机,写入你选定的目录Render / export / save — on-device, into the folder you pick
  • 遥测 / 行为采集 — 不做Telemetry / behaviour tracking — none
  • 唯一外发:你主动提供的 http(s) 数据直链Only outbound: an http(s) data URL you explicitly provide
i 不做三件事:不上传数据、不采集遥测、不记录使用行为。装机只从公开镜像获取 Node / ffmpeg 等软件包,不涉及用户数据。Three things we don't do: no data upload, no telemetry, no behaviour logging. Setup only fetches public packages (Node / ffmpeg) from mirrors — never your data.
数据安全Data safety

可以不信文档,自己验Don't take the docs' word — verify it

把“本地优先”做成可检查的事实:环境清单、体检报告、产物 manifest,逐项都能核对。"Local-first" is made checkable: environment inventory, health report and per-render manifest — all verifiable line by line.

verify — terminal
# 环境清单(含工具目录位置)
motionchart doctor

# 六项体检报告(机器可读)
motionchart onboard --check --json
923自动化测试用例automated tests
52测试文件test files
7层测试金字塔test layers
0可访问性致命项a11y fatal issues

单元 × 属性 × 视觉回归 × 端到端 × 性能 × CLI 契约 × 跨平台 CI;覆盖率门禁写入配置,低于阈值构建即失败。Unit × property × visual regression × e2e × performance × CLI contract × cross-platform CI; coverage gates fail the build below threshold.

快速开始Quick start

三步,从零到第一张图Three steps to your first figure

需要 Node.js ≥ 20 与一款中文字体。安装约一分钟,只装一次;之后每次画图都不用再装。Requires Node.js ≥ 20 and a CJK font. Install takes about a minute, once — after that, every chart needs nothing more.

1
安装渲染引擎Install the engine一条全局安装命令,走国内镜像;只装这一个程序,放在独立目录,不修改系统文件。One global install via a China mirror — a single program in its own directory, no system files touched.
2
体检并记住作品目录Onboard & pick a folder六项检查(Node、字体、ffmpeg、写权限、空间、镜像),点选一次保存位置即记住。Six checks (Node, fonts, ffmpeg, write access, space, mirror) and a one-time folder pick that's remembered.
3
渲染验证图Render a demo确认链路真的通,再把 demo 数据换成本机文件。Confirm the pipeline works, then swap the demo data for your own file.
i 首次使用由 AI 代跑:先说明装什么、多久、装在哪(不碰你的文件与数据),征得同意后自动安装,你只需回答三个问题——存哪、静态还是动画、多清晰。First run can be handled by an AI assistant: it explains what, how long and where, asks for consent, then installs — you just answer three questions: where, static or animated, how sharp.
查看完整安装与文档Read the full install guide
motionchart — terminal
# 1 · 安装(约 1 分钟,只装一次 / install once)
npm install -g motionchart-cli --registry https://registry.npmmirror.com

# 2 · 六项体检 → 选目录 → 验证图 / onboard
motionchart onboard

# 3 · 查看效果,替换为自有数据 / render demo
motionchart demo --formats png --open

免全局安装:npx -y motionchart-cli doctorNo global install: npx -y motionchart-cli doctor

60 种图表,先看再决定See all 60 charts before you commit

画廊按 11 个族分组,点击任意图表放大查看细节;全部由本项目用合成数据本地渲染。The gallery groups all 60 by family — click any chart to zoom in. Every one rendered locally from synthetic data.

打开图表画廊Open the gallery
常见问题FAQ

先把顾虑讲清楚The concerns, answered up front

安装会不会弄坏电脑?装在哪?Will installing mess up my computer? Where does it go?
只安装一个命令行程序到独立目录,不修改系统文件,也不需要管理员权限;图表只写入你点选的目录。卸载只需 npm uninstall -g motionchart-cli。It installs a single CLI program into its own directory — no system files modified, no admin rights. Charts are written only to the folder you choose. Uninstall with npm uninstall -g motionchart-cli.
会不会偷偷联网?Does it phone home?
默认不联网。唯一的外发请求,是你主动给出 http(s) 数据直链时抓取该地址,并记录来源与 SHA-256。装机阶段只从公开镜像下载 Node / ffmpeg 等软件包。No. The only outbound request is fetching an http(s) data URL you explicitly provide, logged with source and SHA-256. Setup only downloads public packages (Node / ffmpeg) from mirrors.
我的数据会被上传吗?Is my data uploaded?
不会。读取、清洗、统计、渲染、导出全部在本机完成;不上传数据、不采集遥测、不记录使用行为。Never. Parsing, cleaning, statistics, rendering and export all run on-device; no uploads, no telemetry, no behaviour logging.
为什么别人家点开就能用,这个要安装?Why does this need an install when others just open in a browser?
因为别人在云端渲染,你的数据要传上去;本工具在本地渲染,所以数据不出本机、断网可用、结果可复现。安装是一次性的,之后每次画图都不用再装。Because they render in the cloud, which means uploading your data. This renders locally — data never leaves, it works offline, and results are reproducible. The install is one-time; later charts need nothing.
支持哪些数据格式?What data formats are supported?
CSV / TSV / XLSX / JSON,可读本地文件或 http(s) 数据直链。数据体检会识别列类型、时间字段与统计列(误差、置信区间、样本量等),据此决定图表的可用能力。CSV / TSV / XLSX / JSON, from local files or http(s) URLs. Triage detects column types, time fields and statistical columns (error, CI, n) to decide which chart capabilities apply.

把下一张图,交给可复现的链路Give your next figure a reproducible pipeline

本地渲染、规格驱动、逐字节可复现——从论文配图到答辩演示,一条命令走完。Local rendering, spec-driven, byte-reproducible — from paper figures to defence decks, in one command.

开始使用Get started