文件格式任意转换助手Transfer Any File

离线 · 无上传 · 浏览器本地完成Offline · no uploads · runs in the browser

转换一览Conversion index

这些页面不是「我们支持 14 种格式」的另一种说法。格式数只是图的顶点数：14 种格式、48 条注册路径，图上任意一种源格式都能走到全部 11 种可写格式，共 143 个组合，其中界面对用户实际提供 116 个。剩下的 27 个不是没做，而是语义上无效，所以置灰并说明原因——每个配对页面里的「不做什么」写的就是这一类事。These pages are not another way of saying “14 formats supported”. The format count is just the number of vertices: 14 formats and 48 registered direct routes, and every source format reaches all 11 writable ones, which makes 143 combinations reachable and 116 selectable. The other 27 are not missing work — they are semantically invalid, so the picker greys them out and says why. Each pair page spells that class of limits out in its own “what it does not do”.

查一条转换路径Find a route

选一个源格式和一个目标格式，这里给出扩展实际执行的链路、步数，以及这一对的说明入口。工作台里的下拉会按你拖进来的那批文件实时收窄，这里给的是图上的完整答案。Pick a source and a target to see the chain the extension actually runs, how many steps it takes, and where that pair is written up. The workbench narrows its picker to the batch you dropped; this is the full answer on the graph.

这个查找器需要脚本运行；下面的格式矩阵始终列出每一条链路。The finder needs its script to run; the format matrix below always lists every route.

选两个格式，看看它走几步。Choose two formats to see how many steps it takes.

工作台空状态：拖放区与格式选择器
拖入、选格式，在你自己电脑上转换Drop, pick a format, convert on your own machine

按来源格式浏览Browse by source format

每个来源一张页：它的每个开放目标、完整链路，以及那条路保留什么、丢掉什么。矩阵里没有专页的格子，落点也在这些页上。One page per source format: every open target, the full route it walks, and what that route keeps or drops. The matrix cells without a page of their own land on these pages too.

格式矩阵The format matrix

这张表是把上面那段话摊开写：行是 14 种格式，列是 11 种可写格式，一共 154 个格子，其中 11 格是同一个格式对着自己。剩下的 143 个可达组合里，27 个语义无效，所以界面对用户实际提供 116 个。格子里写的是这条链路实际走的步数——它不是「我们支持 N 种转换」的另一种说法，而是那张「没有手工维护的转换矩阵」被现推出来给你看：48 条注册路径，闭包自己算出这些格子。The grid is that sentence spread out: 14 formats down the side, the 11 writable ones across the top, 154 cells in all, of which 11 are a format meeting itself. Of the remaining 143 reachable combinations, 27 of them are blocked by the policy as semantically invalid, so the picker offers 116. What a cell holds is the number of steps that route actually runs — not another way of saying “we support N conversions”, but the matrix nobody maintains worked out in front of you: 48 registered routes, and these cells are their closure.

行 = 源格式，列 = 目标格式；数字是步数，✕ 是界面置灰的组合，– 是同格式。每个开放格子都能打开它自己的说明页或来源格式页。Rows are source formats, columns the writable ones; a digit counts steps, ✕ is a pair the picker greys out, – is the same format. Every open cell opens either its own page or the source format's page.
源 ↓ · 目标 →Source ↓ · Target → MarkdownMarkdown HTMLHTML Word (.docx)Word (.docx) PDFPDF Excel (.xlsx)Excel (.xlsx) CSVCSV 纯文本 (TXT)plain text (TXT) JSONJSON PNGPNG JPEGJPEG WebPWebP
MarkdownMarkdown 同一格式same format 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 4 步链路，有专页4 steps, with its own page 3 步链路，有专页3 steps, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 3 步链路，有专页3 steps, with its own page 3 步链路，有专页3 steps, with its own page
HTMLHTML 1 步链路，有专页1 step, with its own page 同一格式same format 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 3 步链路，有专页3 steps, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page
Word (.docx)Word (.docx) 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 同一格式same format 2 步链路，有专页2 steps, with its own page 4 步链路，有专页4 steps, with its own page 3 步链路，有专页3 steps, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 3 步链路，有专页3 steps, with its own page 3 步链路，有专页3 steps, with its own page
PDFPDF 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 同一格式same format 界面置灰：PDF 没有表格结构greyed out: no table structure 界面置灰：PDF 没有表格结构greyed out: no table structure 2 步链路，有专页2 steps, with its own page 界面置灰：PDF 没有表格结构greyed out: no table structure 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page
Excel (.xlsx)Excel (.xlsx) 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 同一格式same format 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 3 步链路，有专页3 steps, with its own page 3 步链路，有专页3 steps, with its own page
CSVCSV 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 同一格式same format 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 3 步链路，有专页3 steps, with its own page 3 步链路，有专页3 steps, with its own page
纯文本 (TXT)plain text (TXT) 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 4 步链路，有专页4 steps, with its own page 3 步链路，有专页3 steps, with its own page 同一格式same format 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 3 步链路，有专页3 steps, with its own page 3 步链路，有专页3 steps, with its own page
JSONJSON 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 同一格式same format 2 步链路，有专页2 steps, with its own page 3 步链路，有专页3 steps, with its own page 3 步链路，有专页3 steps, with its own page
PNGPNG 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 同一格式same format 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page
JPEGJPEG 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 1 步链路，有专页1 step, with its own page 同一格式same format 1 步链路，有专页1 step, with its own page
WebPWebP 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 同一格式same format
BMPBMP 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page
GIFGIF 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page
SVGSVG 2 步链路，有专页2 steps, with its own page 1 步链路，有专页1 step, with its own page 2 步链路，有专页2 steps, with its own page 2 步链路，有专页2 steps, with its own page 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 界面置灰：图片没有文字层greyed out: no text layer 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page 1 步链路，有专页1 step, with its own page

图例：数字 = 这条链路走的步数（4 是今天最长的一条，中间产物不落盘）；✕ = 图上可达但语义无效，工作台会置灰并给出原因 —— 「图片无法转换为文本或表格数据（需 OCR，暂不支持）」「PDF 无法可靠转换为表格/结构化数据」；– = 源与目标是同一个格式。带下划线的数字都能点：有专页的进那一页看保留项与丢弃项，没有专页的进来源格式页看这一条路的说明。Legend: a digit is how many steps that route runs (4 is the longest one today, and no intermediate ever touches the disk); ✕ is reachable in the graph but semantically invalid, so the workbench greys it out and says why — “Images cannot become text or tabular data (needs OCR, not supported)” and “PDF cannot be reliably converted to tabular data”; – is a format meeting itself. Every underlined digit opens something: a pair page with that route's kept and dropped details, or its source format's page where the route is explained.

把指针移到任一格，或用 Tab 聚焦矩阵后按方向键逐格走，这里会读出这条链路的完整走法。Hover a cell, or press Tab and walk the grid with the arrow keys, and the full route is read out here.

文档Documents

Markdown、Word、HTML、PDF 之间互转；文档类产物的差别写在各自页面上，尤其 PDF 是图像还是可编辑文档。Markdown, Word, HTML and PDF. The artifacts differ more than the format names suggest, which is why each page says whether it yields an image or something editable.

  • Markdown → Word (.docx)Markdown → Word (.docx) 在浏览器本地把 .md 转成 .docx：标题层级、列表、表格、代码块写成真实 Word 段落，不是截图。批量、可预览、零网络请求。Convert .md to .docx in your browser: headings, lists, tables and code become real Word paragraphs, not a screenshot. Batch, previewable, zero requests.
  • Word (.docx) → MarkdownWord (.docx) → Markdown 浏览器本地把 .docx 转成 Markdown：mammoth 读语义结构，内嵌图片以 data URI 保留，不联网、不排队。支持批量与 ZIP。Turn .docx into Markdown in your browser: mammoth reads the semantic structure, embedded images come along as data URIs. Offline, batch capable, ZIP output.
  • PDF → 纯文本 (TXT)PDF → plain text (TXT) 浏览器本地用 pdf.js 提取 PDF 文字层，输出纯文本或 Markdown。扫描件没有文字层，所以这条对图片型 PDF 无解——项目不含 OCR。pdf.js in your browser pulls the PDF text layer out as .txt or Markdown. A scan has no text layer, so image-only PDFs are a dead end here: the app ships no OCR.
  • Markdown → PDFMarkdown → PDF Markdown → HTML → PDF，本地按 A4 逐页切片成图像型 PDF。好处是哪台机器打开都一样，代价是没有文字层、不可选中检索。Markdown → HTML → PDF, sliced page by page onto A4 locally. It looks identical everywhere; the price is no text layer, so nothing is selectable or searchable.
  • Word (.docx) → PDFWord (.docx) → PDF Word → HTML → PDF：mammoth 读语义结构后整页栅格化，按 A4 切片。文字层没有，版式与 Word 打印不尽相同，先预览再下载。Word → HTML → PDF: mammoth reads the structure, the page is rasterised onto A4. No text layer, and the layout is not Word’s own print output — preview first.
  • Markdown → HTMLMarkdown → HTML 浏览器本地把 .md 渲染成可直接双击打开的完整 HTML：GFM 表格与任务列表可用，样式内联，脚本与事件属性在净化阶段移除。Render .md into a complete, double-clickable HTML document locally: GFM tables and task lists work, styles are inlined, scripts and event attributes removed.
  • HTML → PDFHTML → PDF 浏览器本地把 .html 渲染成多页 A4 PDF：800 px 版心按整页高度切片，切片是无损 PNG；脚本被移除，远程资源不会被取回。Render .html into a multi-page A4 PDF locally: an 800 px layout sliced down its full height, stored as lossless PNG. Scripts are removed; nothing is fetched.
  • PDF → HTMLPDF → HTML 浏览器本地用 pdf.js 按阅读顺序提取文字层，链接标注按协议白名单保留，每页之间插入分页线，产出一个自包含的 .html。全程不上传，也不含 OCR：扫描件没有文字层。pdf.js extracts the text layer in reading order, keeps link annotations through a scheme allowlist and marks page breaks in one self-contained .html. No OCR.
  • PDF → Word (.docx)PDF → Word (.docx) 这条是多步链：pdf.js 先按阅读顺序把文字层提成 HTML，再由它打包成 .docx，中间的 HTML 只在内存里、不会写到你的下载目录。全程离线，无 OCR，也不生成原生 OOXML 段落结构。A two-step chain: pdf.js lifts the text layer into HTML in reading order, that HTML is packed into .docx, and the intermediate never leaves memory. No OCR.
  • Word (.docx) → HTMLWord (.docx) → HTML 浏览器本地用 mammoth 读 .docx 的语义结构（标题、列表、粗斜体、超链接、表格），文档里的图片内联成 data URI，最后经 DOMPurify 净化输出单个 .html。不上传、不装 Office，也不读旧版 .doc。mammoth reads a .docx’s structure — headings, lists, emphasis, links, tables — inline images become data URIs, DOMPurify sanitises the single .html. No .doc.
  • Excel (.xlsx) → PDFExcel (.xlsx) → PDF 这条是多步链：SheetJS 先把每个工作表写成带表名的 HTML 表格，再由它生成 A4 竖版 PDF；中间产物只在内存里。日期不会被读成一串数字，公式值取上次保存时缓存的结果。全程离线。SheetJS writes each worksheet as an HTML table under its name, which becomes a portrait A4 PDF, intermediate held in memory. Formula cells keep cached values.
  • 纯文本 (TXT) → Markdownplain text (TXT) → Markdown 浏览器本地把 .txt 包成带围栏的 Markdown 代码块：内容原样保留，围栏长度会按文本里出现的最长反引号串加长，所以你的三反引号不会把结构撑破。它不猜标题、不猜列表——那是"值保真"与"替你排版"的区别。Wrap a .txt in a fenced Markdown code block verbatim: the fence grows past the longest run of backticks in your text, and it guesses no headings or lists.
  • HTML → Word (.docx)HTML → Word (.docx) 在浏览器本地把 .html 打包成 .docx：标题、列表、表格、链接与内联图示成为 Word 里的真实内容；远程子资源在打包前剥掉，Word 打开时不会替你联网。Pack an HTML page into a real .docx in your browser: headings, lists, tables and diagrams survive, and remote subresources are stripped before Word opens it.
  • HTML → MarkdownHTML → Markdown 浏览器本地把 HTML 转成 GFM：标题、列表、表格与任务复选框保留，内联 SVG 先栅格化成 PNG，图片仍写成 Markdown 链接。Turn an HTML page into GFM in your browser: headings, lists, tables and task checkboxes survive, inline SVG becomes a PNG, images stay as links.
  • HTML → 纯文本 (TXT)HTML → plain text (TXT) 在浏览器里从 HTML 抽取可读文本：块级元素换行，列表保留编号，链接保留 URL，表格单元格用制表符分隔。Extract readable text from an HTML page in the browser: blocks become line breaks, lists keep their numbers, links keep their URLs.
  • Excel (.xlsx) → HTMLExcel (.xlsx) → HTML 在浏览器里把 .xlsx 变成独立 HTML：每个工作表一张表并带表名标题，日期落成文本，产物不含脚本。Turn .xlsx into a standalone HTML page in your browser: one table per sheet with a heading, dates as text, and no scripts in the output.
  • CSV → PDFCSV → PDF 在浏览器里把 .csv 排成 PDF 表格：先解析成 HTML 表格，再按 A4 切成多页。数字与日期按单元格原样呈现，全程不上传。Turn a .csv into a PDF in your browser: parsed into an HTML table, then sliced into A4 pages. A rendered image, with no text layer and no upload.
  • Excel (.xlsx) → Word (.docx)Excel (.xlsx) → Word (.docx) 在工作台里把 .xlsx 转成 .docx：每个工作表写成一张 Word 表格，多表时带上表名标题。值保真、可编辑，全程离线。Each worksheet becomes an editable Word table, named when there is more than one. Values are kept; styling is not. Offline, in your browser.
  • PDF → MarkdownPDF → Markdown 在浏览器里把 .pdf 的文字读出来写成 Markdown：文本项按位置归行，书签命中的行成为标题，链接注释还原成真链接。不 OCR、不上传。Extract a PDF’s text into Markdown in your browser, with link annotations restored as links. No OCR, no layout kept, nothing uploaded.
  • 纯文本 (TXT) → PDFplain text (TXT) → PDF 在浏览器里把 .txt 排成 PDF：整篇进一个 <pre> 块，缩进与对齐照原样保留，再按 A4 切页。不上传、不联网。The whole file goes into one <pre> block, so indentation survives, then it is cut into A4 pages. Offline, in your browser.
  • Word (.docx) → 纯文本 (TXT)Word (.docx) → plain text (TXT) 在浏览器里把 .docx 读成纯文本：有序列表按位置重新编号、表格写成制表符分隔的行、链接以 [url] 附在文字后。全程离线。Headings become lines, ordered lists are renumbered by position, tables come out tab-separated and links follow as [url]. Offline, in the tab.
  • CSV → HTMLCSV → HTML 在浏览器里把 .csv 变成一张真的 HTML 表格：带引号的逗号与字段内换行按一个单元格处理，值原样呈现。批量、可预览、零网络请求。A real HTML table from your .csv: quoted commas and newlines inside a field stay one cell, and values print as written. Batch, previewable, zero requests.
  • Excel (.xlsx) → 纯文本 (TXT)Excel (.xlsx) → plain text (TXT) 在浏览器里把 .xlsx 读成纯文本：按工作表顺序逐行取单元格文本，同行以制表符分隔；日期归一，空单元格留空。Cell text in worksheet order, tabs between cells in a row and a newline between rows. Dates normalised, empty cells left empty.
  • CSV → MarkdownCSV → Markdown 把 .csv 变成 GitHub 风格的管道表：引号里的逗号与字段内换行按 CSV 语法解析成一个单元格，格子里的竖线自动转义。批量、可预览、零网络请求。A CSV becomes a pipe table: quoted commas and newlines inside a field parse as one cell, and a pipe in a cell is escaped. Batch, offline, no requests.
  • Markdown → CSVMarkdown → CSV 把文档里的管道表变成能再计算的 CSV：首行当列名，其余行当记录，数字按值写出不被显示格式改写，文件带 BOM 让 Excel 直接双击不乱码。A pipe table becomes a CSV you can compute on: first row as column names, values written as values, and a BOM so Excel opens the file without mojibake.
  • HTML → CSVHTML → CSV 把网页保存下来的表格变成能再计算的 CSV：首行当列名，`<th>` 与 `<td>` 都算一格，链接只取文字，页面里的远程引用一个字节都不去取。A table from a saved page becomes a CSV you can compute on: first row headers, `th` and `td` both count as cells, and no remote reference is ever fetched.
  • Excel (.xlsx) → MarkdownExcel (.xlsx) → Markdown 把工作簿写成管道表：每个工作表一张表并冠一行小标题，日期按 ISO 文本落地，合并区与样式按纯文本的规矩取舍。批量、可预览、零网络请求。A workbook becomes pipe tables — one per worksheet under its own heading, dates as ISO text, merges handled as plain text allows. Batch and fully offline.
  • 纯文本 (TXT) → Word (.docx)plain text (TXT) → Word (.docx) 把纯文本变成 Word 打得开的 .docx：换行、缩进与制表符原样保留，排版交给 Word。批量、可预览、零网络请求，不经过任何在线转换服务。A plain-text file becomes a .docx Word opens: line breaks, indentation and tabs preserved, layout left to Word. Fully offline, never sent to a service.
  • JSON → PDFJSON → PDF 先把 JSON 渲染成可折叠的语法高亮查看器，再把这份页面按 A4 分页写成 PDF。产物是位图、没有文字层，但结构与配色都在。JSON becomes a collapsible, colour-coded viewer, and that page is captured to A4 pages inside a PDF — an image with no text layer, structure and colours intact.
  • JSON → 纯文本 (TXT)JSON → plain text (TXT) 把 JSON 变成结构查看器那份可读文字：每个键值各占一行、引号保留，配色与缩进不带过去。全程离线，文件不出本机。JSON becomes the viewer’s readable text: one key and value per line, quotes kept, colour and indentation lost. Fully offline — nothing is uploaded.
  • CSV → 纯文本 (TXT)CSV → plain text (TXT) 把 CSV 变成制表符分隔的纯文本：一行一条记录、值原样保留，贴进聊天记录或纯文本编辑器都不散列。全程离线，不经过服务器。A CSV becomes tab-separated plain text: one record per line, values as written, so it pastes into chat or a text editor without breaking columns.
  • JSON → Word (.docx)JSON → Word (.docx) 把 JSON 的结构交给 .docx：产物里带的是那份可折叠查看器的 HTML，Word 打开时才排版。全程离线，文件不出本机。JSON becomes a .docx carrying the structure viewer’s own HTML: Word lays it out when the file opens. Fully offline — nothing is uploaded.
  • Markdown → 纯文本 (TXT)Markdown → plain text (TXT) 去掉记号、留下阅读顺序：标题变成一行文字，链接把地址附在文字后，列表按位置重新编号。全程离线，不经过服务器。Strip the markup: headings become lines, links keep their URL after the text, and ordered lists are renumbered. Offline — nothing is uploaded.
  • 纯文本 (TXT) → HTMLplain text (TXT) → HTML 把纯文本包成一个能直接打开的 HTML 页面：字符逐一转义、换行保留为 `<br>`、整段落在 `<pre>` 里，不把任何内容当标记解释。Plain text becomes an HTML page you can open: characters escaped, line breaks kept, the body wrapped in `<pre>` with nothing interpreted as markup.
  • JSON → HTMLJSON → HTML 把 JSON 变成一份能双击打开的自包含网页：可折叠的层级树、按类型着色、数组与对象带计数，值全部作为转义文字呈现。JSON becomes a self-contained page you can double-click: a collapsible tree, colour by type, array and object counts, every value escaped as text.
  • HTML → JSONHTML → JSON 把网页里的 `<table>` 变成对象数组：首行当键名，`<th>` 与 `<td>` 都算一格；多张表按小标题分组。全程离线，页面里不请求任何东西。Turn a page’s `<table>` into objects: the first row supplies keys, `th` and `td` both count, and several tables group by heading. Offline — nothing is fetched.
  • Markdown → JSONMarkdown → JSON 把 Markdown 表格变成对象数组：表头行给出键名，两张以上按 `##` 小标题分组，没有表格时读代码块。全程离线，不请求图片与脚本。Turn Markdown tables into objects: the GFM header row supplies keys, and several tables group under their `##` headings. Offline, no uploads.
  • Markdown → Excel (.xlsx)Markdown → Excel (.xlsx) Markdown 表格经 HTML、JSON、CSV 三步落进一张工作表：15 位以内的数字变可求和，`00424`、日期串和 16 位以上的账号保持原字符，`=` 开头加护栏。全程离线。A GFM table lands in a worksheet via JSON and CSV: numbers become summable, 00424 and 16-digit IDs stay text, and a leading = is disarmed. Offline.
  • 纯文本 (TXT) → JSONplain text (TXT) → JSON `.txt` 里其实是一段 JSON：解析后按两级缩进重排，数字与布尔保持类型，不是 JSON 就报错而不是包成字符串。全程离线。JSON saved as .txt comes back indented two spaces, with numbers and booleans still themselves. Prose fails instead of being wrapped into a string. Offline.
  • 纯文本 (TXT) → CSVplain text (TXT) → CSV `.txt` 里是一段 JSON 数组时：键名取并集当表头、每条记录一行，没有键的元素写进合成的 `_value` 列而不是丢掉。全程离线。A JSON array saved as .txt becomes a table: union keys as headers, one row per record, and bare values caught by a `_value` column instead of dropped. Offline.
  • JSON → MarkdownJSON → Markdown JSON 先渲染成语法着色的查看器，再逐行落成 Markdown：键、值与 `Object(4)` 摘要都作为文字保留，没有标题也没有列表。全程离线。Each JSON line becomes its own Markdown paragraph: keys, values and the Object(4) summaries stay as words. No headings, no lists, nothing uploaded.
  • HTML → Excel (.xlsx)HTML → Excel (.xlsx) 网页里的 `<table>` 经 JSON 与 CSV 落进一张工作表：15 位以内的数字可求和，`00424` 与 16 位账号保持字符，`=` 开头加护栏。全程离线。A page table goes through JSON and CSV into one worksheet: 15-digit numbers become summable, 00424 and 16-digit IDs stay text, and = is disarmed. Offline.
  • CSV → Word (.docx)CSV → Word (.docx) CSV 先按 `raw: true` 读成一张 HTML 表，再打包成 Word 文档：`00424`、`2024-01-05` 保持写下的字符，排版由 Word 决定。全程离线。The CSV is read with raw: true into an HTML table, then packed as a Word file whose layout Word itself decides: 00424 and 2024-01-05 stay as written. Offline.
  • Word (.docx) → CSVWord (.docx) → CSV mammoth 先把文档变成 HTML，表格里第一行给出列名、其余每行一条记录；值按写下的字符保留。两份表格会撞闸。全程离线。Mammoth turns the document into HTML, the first table row supplies keys and values stay exactly as typed. Two tables in one file stop it. Offline.
  • Word (.docx) → Excel (.xlsx)Word (.docx) → Excel (.xlsx) Word 表格经 HTML、JSON、CSV 三步落进一张工作表：数字只在字节能逐字回来时才是数字，16 位编号保持文本。两份表格会撞闸。全程离线。A Word table travels HTML → JSON → CSV into one worksheet, where a number returns only if its bytes do and 16-digit ids stay text. Offline, no uploads.
  • 纯文本 (TXT) → Excel (.xlsx)plain text (TXT) → Excel (.xlsx) 四步链：`.txt` 整段包进 `<pre>`，按 JSON 解析、摊成 CSV、再写工作簿。数字只在字节能逐字回来时才成数，16 位以上留在文本。全程离线。Four steps: the text is wrapped in a pre block, parsed as JSON, spread into CSV and written as a workbook. Only round-trip values become numbers.
  • Word (.docx) → JSONWord (.docx) → JSON mammoth 出 HTML 再取表格：一张表变对象数组，多张表按位置与文档小标题配对分组（配不上时回落成 SheetN），没表时依次试代码块与整份正文。全程离线。Mammoth to HTML, then the tables: one becomes an array of objects, several group by heading position (or SheetN), else the first text found is parsed.

数据Data

Excel、CSV、JSON 三方互转，按值而非显示文本序列化；PDF 与图片不提供到表格结构的路径，界面上会置灰并说明原因。Excel, CSV and JSON, serialised from stored values rather than display text. PDF and images have no route into a table structure, and the picker greys those out with a reason.

  • Excel (.xlsx) → CSVExcel (.xlsx) → CSV 浏览器本地把 .xlsx 转成 CSV：按单元格值序列化，日期统一格式，带 UTF-8 BOM，公式注入被转义。多 sheet 每表一份 CSV 装进 ZIP。Convert .xlsx to CSV in your browser from cell values, not display text: dates normalised, UTF-8 BOM, formula injection escaped, one CSV per sheet in a ZIP.
  • CSV → Excel (.xlsx)CSV → Excel (.xlsx) 浏览器本地把 .csv 转成 .xlsx：只有能精确回写的字段才当数字存，长编号与前导零原样保留；GBK/GB18030 编码自动兜底。Convert .csv to .xlsx in your browser: only fields that round-trip exactly become numbers, so long IDs and leading zeros survive. GBK / GB18030 decodes itself.
  • JSON → CSVJSON → CSV 浏览器本地把 JSON 对象数组转成 CSV：表头取所有键的并集，嵌套值以 JSON 文本入格，公式注入被转义，带 UTF-8 BOM。Convert a JSON array of objects to CSV locally: headers are the union of every key, nested values land as JSON text, formula injection is escaped, UTF-8 BOM.
  • Excel (.xlsx) → JSONExcel (.xlsx) → JSON 浏览器本地把 .xlsx 转成 JSON：单表给行对象数组，多表按工作表名分组；数字与布尔是真类型，日期是两段式 ISO 文本。Convert .xlsx to JSON locally: one sheet becomes an array of row objects, several are keyed by sheet name. Numbers and booleans stay typed, dates are ISO text.
  • CSV → JSONCSV → JSON 浏览器本地把 .csv 转成 JSON 行对象：字段先按原文读入，只在「解析→回写完全一致」时才恢复数字类型；GBK / GB18030 自动兜底。Convert .csv to JSON row objects locally: every field starts as text and regains a numeric type only when parse-then-reprint reproduces it exactly.
  • JSON → Excel (.xlsx)JSON → Excel (.xlsx) 浏览器本地把 JSON 数组转成 .xlsx：表头取所有对象键的并集，嵌套对象与数组以 JSON 文本进单元格。链路是 JSON → CSV → Excel，中间不落盘。Turn a JSON array into .xlsx locally: the header is the union of every object’s keys and nested values go into the cell as JSON text, routed via CSV.

图片Images

可写格式为 PNG / JPEG / WebP 三种（浏览器不提供 BMP、GIF、SVG 编码器），这三种只能作为输入。PNG, JPEG and WebP are the writable image formats — a browser ships no encoder for BMP, GIF or SVG, so those three are input-only.

  • PNG → WebPPNG → WebP 浏览器里把 PNG 编成 WebP，透明通道保留，质量档与目标体积先走质量阶梯再缩像素。最多 200 张一批，结果可打 ZIP。Encode PNG to WebP in the browser with alpha preserved. The target-size lever walks the quality ladder before it throws pixels away; 200 images per batch.
  • SVG → PNGSVG → PNG 浏览器本地把 SVG 栅格化成 PNG（也可出 JPEG / WebP）。矢量按当前尺寸渲染，脚本与事件属性先行剥离，外链资源不会去取。Rasterise SVG to PNG in your browser at the size you set: scripts and event attributes are stripped first, and remote resources are never fetched.
  • JPEG → PNGJPEG → PNG 浏览器本地把 .jpg / .jpeg 解码成像素再编为 PNG：尺寸原样，可选最长边，一批最多 200 张可打成一个 ZIP。EXIF 与 GPS 不会跟着走。Decode .jpg to pixels in your browser and re-encode as PNG: dimensions kept, optional longest edge, 200 images in one ZIP. EXIF and GPS do not travel.
  • PNG → JPEGPNG → JPEG 浏览器本地把 PNG 编成 JPEG：透明区域铺纯白展平，质量档 40%–90%、目标体积 20 KB–2 MB，先降质量再缩像素。最多 200 张一批。Encode PNG to JPEG in your browser: alpha flattens onto pure white, with a 40%–90% quality step and a 20 KB–2 MB target that spends quality before pixels.
  • JPEG → WebPJPEG → WebP 浏览器本地把 .jpg 编成 WebP：质量档 40%–90%、目标体积 20 KB–2 MB，先走质量阶梯再缩像素。最多 200 张一批，结果可打 ZIP。Encode .jpg to WebP in your browser with a 40%–90% quality step and a 20 KB–2 MB target — the quality ladder first, pixels second. 200 images per batch.
  • WebP → JPEGWebP → JPEG 浏览器本地把 .webp 解码重编为 JPEG：透明铺纯白，质量档与目标体积照旧，一批最多 200 张。动图 WebP 只会留下一帧。Decode .webp and re-encode as JPEG locally: alpha flattens onto white, quality and target size as usual, 200 per batch. Animated WebP yields one frame.
  • PNG → PDFPNG → PDF 浏览器本地把 PNG 装进 PDF：常规路径不重画像素，页面尺寸由图像像素换算，横竖自动。一批最多 200 张，默认各自一个 PDF，也可勾选合并成一份多页 PDF。Put PNGs into PDFs locally: pixels are embedded, not redrawn, and each page takes its image’s size and orientation. One PDF each, or one merged multi-page PDF.
  • JPEG → PDFJPEG → PDF 浏览器本地把 .jpg 装进 PDF：页面按像素尺寸换算、横竖自动。这条路解码后再编码一次，PNG 那条则不重编码，两者是有损与无损的对照；默认一张一份，也可勾选合并成一份多页 PDF。Put .jpg files into PDFs locally, sized from their pixels. This route re-encodes once; the PNG route does not. One PDF each, or one merged multi-page PDF.
  • PDF → JPEGPDF → JPEG 浏览器本地用 pdf.js 把 PDF 逐页渲染成 JPEG：渲染密度四档 96 / 144 / 200 / 300 DPI（默认 144）、质量档 40%–90%、可指定最长边，多页自动一页一张装进 ZIP。文件不出本机，也不含 OCR。Render each PDF page to JPEG at 96, 144, 200 or 300 DPI (144 by default), quality 40%–90%, optional longest edge; multi-page batches arrive as a ZIP.
  • PDF → PNGPDF → PNG 浏览器本地用 pdf.js 把 PDF 逐页渲染成 PNG：渲染密度四档 96 / 144 / 200 / 300 DPI（默认 144）、可指定最长边，多页自动一页一张装进 ZIP。PNG 没有质量档也不吃目标体积——无损的意思是不替你决定丢哪些像素。文件不出本机，也不含 OCR。Render each PDF page to PNG at 96, 144, 200 or 300 DPI with an optional longest edge, one image per page in a ZIP. PNG takes no quality step: it is lossless.
  • PDF → WebPPDF → WebP 浏览器本地把 PDF 逐页渲染成 WebP：密度四档 96 / 144 / 200 / 300 DPI、质量档 40%–90%、可指定最长边，多页一页一张装进 ZIP。它与 JPEG 的差别只有体积与容器兼容性——渲染前一样先铺白底，所以这条路没有透明输出。文件不出本机。Render each PDF page to WebP at 96–300 DPI, quality 40%–90%, one image per page in a ZIP. Like the JPEG route it paints the page white first: no transparency.
  • HTML → PNGHTML → PNG 浏览器本地把 .html 渲染到画布再编码为 PNG：800 px 宽、按内容实际高度出整页长图，远程引用与脚本在渲染前就被处理掉。全程不出本机，没有截图软件、也没有服务端渲染。Render a .html file to a canvas and encode it as PNG: 800 px wide, one tall image as long as the content is, with remote references never fetched.
  • GIF → PNGGIF → PNG 浏览器本地把 GIF 解码后编成 PNG：静态图无损换容器，动图取首帧并在结果卡片上写明丢了帧。BMP / GIF / SVG 只能作为输入——浏览器不提供它们的编码器，所以这条的方向不可逆。Decode a GIF and encode it as PNG: a still gets a lossless container swap, an animation keeps its first frame and says so on the result card. One-way only.
  • BMP → PNGBMP → PNG 浏览器本地把 .bmp 解码后编成 PNG：像素尺寸与颜色原样带走，不再叠有损压缩。BMP 只能作为输入（浏览器不提供它的编码器），所以这条方向不可逆；一批最多 200 张，结果可打 ZIP。Decode a .bmp and encode it as PNG in your browser: pixel size and colour carry over with no lossy pass. BMP is input-only, so the route does not run backwards.
  • BMP → JPEGBMP → JPEG 浏览器本地把 .bmp 解码后编成 JPEG：质量档 40%–90%、目标体积 20 KB–2 MB、可指定最长边，够不着目标时先降质量再缩像素。BMP 只能作为输入，方向不可逆。Decode a .bmp and encode it as JPEG with a 40%–90% quality step, a 20 KB–2 MB target and an optional longest edge — quality gives way before pixels do.
  • SVG → JPEGSVG → JPEG 浏览器本地把 .svg 栅格化后编成 JPEG：先净化掉脚本与事件属性、按 width / viewBox 定尺寸（都没有时按 1024 px 方形），再画到画布编码。SVG 只能作为输入，方向不可逆；需要透明请改选 PNG 或 WebP。Rasterise a .svg to JPEG: scripts and event attributes are sanitised out first, size comes from width / viewBox (else a 1024 px square). SVG is input-only here.
  • SVG → WebPSVG → WebP 浏览器本地把 .svg 栅格化后编成 WebP：带 Alpha，所以透明图形的背景能留住；质量档 40%–90%、目标体积 20 KB–2 MB、最长边可调。尺寸按 width / viewBox 推断，外部引用一律不请求。Rasterise a .svg to WebP locally — the SVG route that keeps transparency, with quality, target size and longest edge adjustable and nothing external fetched.
  • WebP → PNGWebP → PNG 在浏览器里把 .webp 解码并重编码为 PNG，透明通道原样保留；可批量、可限制最长边，文件不上传。Decode a .webp in your browser and re-encode it as a PNG with transparency intact. Batch it, cap the long edge, no upload.
  • GIF → JPEGGIF → JPEG 把 .gif 解成静态 .jpg：浏览器取第一帧，铺白底后重编码；多帧 GIF 会在结果里如实标记丢帧，文件不上传。Turn a .gif into a still .jpg in the browser: the first frame is flattened onto white and re-encoded, and extra frames are reported as lost.
  • WebP → PDFWebP → PDF 在浏览器里把 .webp 铺成一页 PDF：页面尺寸按图片像素换算，透明铺白，整页只有图像、没有文字层。Put a .webp on a PDF page sized to its pixels, in your browser. Transparency is flattened onto white and no text layer is added.
  • PNG → HTMLPNG → HTML 把 .png 包成单图 HTML：原始字节以 data URI 内嵌，不重编码，透明与像素原样保留，文件可离线打开。Wrap a .png in a one-image HTML page with its original bytes as a data URI: no re-encode, transparency intact, and it opens anywhere.
  • SVG → PDFSVG → PDF 在浏览器里把 .svg 转成 .pdf：SVG 先归一化并栅格化成 PNG，再按像素尺寸铺成一页。不联网、不上传，脚本内容先剥除。The SVG is normalised, stripped of active content and rasterised to a bitmap page sized to its pixels. No vector, no text layer.
  • Word (.docx) → PNGWord (.docx) → PNG 在浏览器里把 .docx 渲染成一张 PNG：先由 mammoth 生成语义化 HTML，再整页画到画布上导出。可指定最长边，全程不上传。A .docx renders through HTML to one long PNG, sized by the longest edge you set. Pixels, not selectable text; nothing is uploaded.
  • WebP → Word (.docx)WebP → Word (.docx) 在浏览器里把 .webp 放进一份 .docx：图片字节原样内嵌，不另附文件、不发请求。图不是文字，Word 显示与否取决于它的 WebP 支持。The WebP goes inside the .docx as raw bytes — no file alongside, nothing fetched. Whether Word paints a WebP depends on its own support.
  • GIF → PDFGIF → PDF 在浏览器里把 .gif 写成 PDF：取第一帧重编码到白底、页面按像素尺寸生成，动图会在结果里说明丢了帧。不上传。The first frame is re-encoded onto white and laid on a page sized to its pixels; an animated source says so in the result. Nothing uploaded.
  • SVG → HTMLSVG → HTML 在浏览器里把 .svg 包成一份能直接打开的 HTML：图形仍是矢量，缩放不失真；脚本与事件属性在净化时被剥掉。不上传。The SVG is wrapped in a page you can open, and it stays vector — zoomable, with scripts and event attributes removed by sanitising. Nothing uploaded.
  • JSON → PNGJSON → PNG 先把 JSON 渲染成可折叠的查看器，再把整页无损截成一张 PNG：换行由排版决定，高度超过 32768 px 就报错而不是截断。全程离线。The viewer is captured whole into one lossless PNG: wrapping follows the layout, and past 32768 px of height it fails instead of truncating. Offline.
  • JSON → JPEGJSON → JPEG JSON 先渲染成查看器，再按 A4 分页写成无损页面、逐页编码成 JPEG：多页会打包成 ZIP，长文件一页一张。全程离线。JSON renders as the viewer, is written as lossless A4 pages, then each page is encoded to JPEG: several pages arrive in a ZIP. Fully offline.
  • Markdown → PNGMarkdown → PNG Markdown 先按 GFM 渲染成带样式的 HTML，再无损截成一张长图：表格有条纹、行内 SVG 图形保留。高度超过 32768 px 会报错而不是截断。GFM renders to styled HTML, then the page is captured losslessly: striped tables and inline SVG diagrams survive the trip. Past 32768 px it fails, not trims.
  • CSV → PNGCSV → PNG CSV 先按 `raw: true` 读成 HTML 表，再整页无损截成一张 PNG：1px 边框与行条纹由自带样式表决定。高度超过 32768 px 会报错而不是截断。Read with raw: true into HTML, then captured losslessly as one PNG: 1px rules and row striping come from the bundled stylesheet. Past 32768 px it fails.
  • CSV → JPEGCSV → JPEG 表格先写成无损的 A4 PDF 页面，再逐页编码成 JPEG：质量与目标体积在这里有档位，长表会跨页并打包成 ZIP。全程离线。The table becomes lossless A4 PDF pages, then one JPEG per page: quality and target size have dials here, and long tables arrive in a ZIP. Offline.
  • Excel (.xlsx) → PNGExcel (.xlsx) → PNG 每个工作表渲染成带表名小标题的表格，整叠堆成一张无损 PNG；高度超过 32768 px 报错而不是截断，长簿子的像素密度会降档。Each worksheet renders as a titled table and the stack is captured losslessly in one PNG; past 32768 px it fails, and long workbooks drop a density step.
  • HTML → JPEGHTML → JPEG 文档整篇渲染一次、按 A4 切成无损页面，再逐页编码成 JPEG：两页以上打包成 ZIP，透明区域铺白。全程离线，脚本不执行、远程资源不取字节。The page is rendered once, sliced into A4 pages stored losslessly, then encoded JPEG by JPEG; several pages arrive in a ZIP. Fully offline.
  • Markdown → JPEGMarkdown → JPEG Markdown 先按 GFM 渲染成带样式的 HTML，再整篇渲染一次、按 A4 切片存无损页面，逐页编码成 JPEG；多页打包成 ZIP。全程离线。GFM renders to styled HTML first, then the document is captured once, sliced into lossless A4 pages and encoded JPEG by JPEG, zipped when there are several.
  • HTML → WebPHTML → WebP 文档整篇渲染一次、按 A4 切成无损片段存进 PDF，再逐页编码成 WebP：两页以上打包 ZIP，质量与目标体积可调。全程离线，脚本不执行。The document is rendered once, sliced into lossless A4 pages inside a PDF, then encoded WebP page by page; several arrive in a ZIP. Fully offline.
  • Markdown → WebPMarkdown → WebP Markdown 先按 GFM 渲染成带样式的 HTML，再整篇渲染一次、按 A4 切成无损片段，逐页编码成 WebP；多页打包 ZIP。全程离线。GFM renders to styled HTML first, then the document is captured once, sliced into lossless A4 pages and encoded WebP page by page, a ZIP when several.
  • Word (.docx) → JPEGWord (.docx) → JPEG 文档先由 mammoth 变成带样式的 HTML（内嵌图片会读成 data URI 留在页面里），再按 A4 切成无损片段、逐页编码成 JPEG；多页出 ZIP。全程离线。Mammoth turns the document into HTML — embedded pictures arrive as data URIs — then it is sliced into A4 pages and JPEG-encoded one page at a time.
  • JPEG → HTMLJPEG → HTML 文件以 data URI 内嵌进一份能直接打开的网页，`alt` 与 `title` 用你的文件名主干；不解码也不重编码，像素就是你原来那一份。全程离线。The file is embedded as a data URI with your file name as alt and title; nothing is decoded or re-encoded, so the pixels are the ones you had.
  • JPEG → MarkdownJPEG → Markdown 产物是一行自包含的图片语法：`![名字](data:image/…;base64,… "名字")`，字节就是原文件；不识别图里的文字，也不产生描述。全程离线。You get one self-contained image tag whose bytes are the original file. No OCR, so no text is invented and nothing is uploaded.
  • PNG → Word (.docx)PNG → Word (.docx) 图片作为包内媒体部件进 .docx，画它不需要任何一次取数；不重编码，透明与像素原样保留。全程离线，文件不出本机。The picture travels as a media part inside the .docx, so nothing has to be fetched to draw it. No re-encode: alpha and pixels stay as they were.
  • JPEG → Word (.docx)JPEG → Word (.docx) 照片作为包内媒体部件进 .docx，画它不需要任何一次取数；像素原样、不重编码，但也不会识别图里的文字。全程离线。The photo rides into a .docx as a media part, so nothing has to be fetched to draw it. Pixels are untouched, but nothing is read out of the picture.
  • GIF → HTMLGIF → HTML 整份文件以 data URI 内嵌进一份能直接打开的网页：不解码、不重编码，所以一帧都不丢；`alt` 与 `title` 用你的文件名主干。全程离线。The whole file is embedded as a data URI: no decode, no re-encode, so no frame is dropped. Opens offline, nothing uploaded.
  • GIF → MarkdownGIF → Markdown 一行自包含的图片语法，负载是整份动图的 base64：不取首帧、不重编码；能否播放由阅读器决定。全程离线，不上传。One self-contained image tag whose base64 is the whole animation. No first-frame grab, no re-encode; playback is up to the reader.
  • BMP → HTMLBMP → HTML 整份 .bmp 以 data URI 原样内嵌进一份能直接打开的网页：不解码、不重编码、不降尺寸；`alt` 与 `title` 用你的文件名主干。全程离线。The whole .bmp is embedded verbatim as a data URI in a page you can open: no decode, no re-encode, no downscale. Fully offline.
  • BMP → PDFBMP → PDF 解码后铺白底、按无损 PNG 流写入 PDF：页面尺寸由图片像素推（96 dpi），一张一页，不是 A4。全程离线，不上传。Decoded onto a white ground and written as a lossless PNG stream: the page is sized from the picture’s pixels, one per page — not A4.
  • SVG → Word (.docx)SVG → Word (.docx) 两步进 Word：图形在打包前被栅格化成 PNG 并提成包内图片部件，界面会写明这次发生了栅格化。全程离线，不上传。Two steps into Word: the artwork is rasterised to PNG before packing and lifted into an image part, and the result card says so. Fully offline.
  • WebP → HTMLWebP → HTML 整份 .webp 以 data URI 原样内嵌进一份能直接打开的网页：不解码、不重编码，透明与动画数据都在。全程离线，文件不出本机。The whole .webp goes into a page you can open as a data URI, byte for byte: no decode, no re-encode, so alpha and the animation data stay put.
  • WebP → MarkdownWebP → Markdown 一行自包含的图片语法，负载是整份 .webp 的 base64：不减帧、不重编码，透明照旧；能否显示与是否播放都由阅读器决定。全程离线。One self-contained image tag whose base64 is the whole .webp: no frame loss, no re-encode, alpha untouched. Display and playback are the reader’s call.
  • BMP → MarkdownBMP → Markdown 一行自包含的图片语法，负载是整份 .bmp 的 base64：不重编码、不降尺寸。但 BMP 的 data URI 不一定被阅读器显示，那种场合请用外链。One inline image tag whose base64 is the whole .bmp — no re-encode. Readers do not always show a BMP data URI, so an external file is the safer shape.
  • PNG → MarkdownPNG → Markdown 一行自包含的图片语法，负载是整份 .png 的字节：不重编码，透明与像素原样。显示与否由阅读器决定，全程离线。One inline image tag whose payload is the whole .png: no re-encode, alpha and pixels untouched. Whether it renders is the reader’s call.
  • Word (.docx) → WebPWord (.docx) → WebP 三步：mammoth 出 HTML，浏览器按 A4 渲染成无损 PNG 切片写进 PDF，最后逐页画白底编码成 WebP。只有最后一步有损。全程离线。Three steps: mammoth to HTML, A4 render as lossless PNG slices in a PDF, then each page onto white and encoded to WebP. Only the last step is lossy.
  • Excel (.xlsx) → JPEGExcel (.xlsx) → JPEG 三步：工作表出 HTML，浏览器按 A4 渲染成 PNG 切片写进 PDF，再逐页画白底编码成 JPEG。读的是算完的值，不重算公式。全程离线。Three steps: sheets to HTML, an A4 render into PNG slices inside a PDF, then each slice onto white and encoded as JPEG. Computed values, never re-evaluated.
  • GIF → Word (.docx)GIF → Word (.docx) 动图原样进 .docx：第一步把整份文件编成 data URI，第二步把它提成包里独立的图片部件。帧一段不少，播不播由打开它的程序决定。The animation rides into a .docx: step one embeds the whole file, step two lifts it into its own image part. Every frame stays; playback is the reader’s call.
  • 纯文本 (TXT) → PNGplain text (TXT) → PNG 把 .txt 排成等宽网页后整页渲染成一张 PNG：两步到图片，中间不经 PDF，空白与换行按源码排版保留。全程离线。A .txt laid out as a monospace page and rendered whole into one PNG: no PDF in the middle, whitespace and line breaks kept as written.
  • GIF → WebPGIF → WebP 取首帧重编码成 WebP：这一路会丢动画并在结果上固定提示；与 JPG 不同，它不铺白底，透明像素保留。全程离线。The first frame, re-encoded as WebP: the animation goes, with a fixed notice, and unlike JPEG no white ground is painted, so alpha survives. Fully offline.
  • 纯文本 (TXT) → JPEGplain text (TXT) → JPEG 三步：等宽排版成网页、按 A4 渲染成无损 PNG 切片写进 PDF，再逐页画到白底编码成 JPEG。只有终点那一次有损。全程离线。Three steps: monospace HTML, A4 render into lossless PNG slices inside a PDF, then each page onto white and encoded once as JPEG.
  • 纯文本 (TXT) → WebPplain text (TXT) → WebP 与 TXT→JPG 同一条三步链，终点换成 WebP：中间是 PNG 无损切片，只有终点量化一次；质量与目标尺寸都在这一步。全程离线。The same three-step chain as TXT to JPG, ending in WebP: lossless PNG slices inside, one quantisation at the endpoint, where quality and target size apply.
  • BMP → Word (.docx)BMP → Word (.docx) 整份 .bmp 经 HTML 进 Word，那条 data URI 被提成包里独立的图片部件：不重编码、不降尺寸。体积随源文件，通常不减。The whole .bmp travels through HTML into Word and becomes its own image part: no re-encode, no downscale. Size follows the source and rarely shrinks.
  • CSV → WebPCSV → WebP 三步：表格出 HTML、按 A4 渲染成无损 PNG 切片写进 PDF，再逐页画到白底编码成 WebP。只有终点量化一次。全程离线。Three steps: the table becomes HTML, is laid out on A4 into lossless PNG slices inside a PDF, then drawn page by page onto white and encoded as WebP.
  • Excel (.xlsx) → WebPExcel (.xlsx) → WebP 三步：工作表出 HTML、按 A4 渲染成无损 PNG 切片写进 PDF，再逐页画白底编码 WebP。多张表各带标题，值不重算。全程离线。Three steps via a PDF: sheets to HTML, A4 pages as lossless PNG slices, then one WebP encode per page. Sheet titles survive; values are not re-evaluated.
  • JSON → WebPJSON → WebP 三步：先做可折叠查看器那份 HTML，再按 A4 渲染成无损 PNG 切片写进 PDF，最后逐页画白底编码成 WebP。缩进是样式、预算提示会一起进图。Three steps via a PDF: the viewer’s HTML, A4 pages as lossless PNG slices, then one WebP encode per page — budget notices included.
  • BMP → WebPBMP → WebP 浏览器把 .bmp 解码后重编码成 WebP：这是 BMP 出口里体积真会下来的一条，代价是像素重画一遍、不再逐字节相同。全程离线。The browser decodes the .bmp and re-encodes it as WebP — the BMP exit where size really drops, paid for with a full redraw and no byte-for-byte sameness.
  • SVG → MarkdownSVG → Markdown 两步：先包成能打开的网页，再序列化时把内联 SVG 栅格化成 PNG，产物是一行内联图片语法。矢量结构到此为止。Two steps: wrap it as a page, then on the way out each inline SVG is rasterised to PNG, leaving one inline image line. The vector ends here.

需要把一张图压到指定体积？用Need a target size? Use the 独立的图片压缩工具standalone image compressor。.

这里列出的是全部 116 个配对；工作台里的目标格式下拉按批次实时给出可用项。This index lists all 116 pairs; the workbench shows what is available for the exact batch you dropped in.

从 Chrome 应用商店安装Install from the Chrome Web Store 安装步骤Install steps GitHub 源码Source on GitHub