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

BMP 转 Word——离线、字节进包成图片部件、不重编码BMP to Word offline — raw bytes as an image part

两步，而且**这一步不碰像素**：`bmp>html` 用 `FileReader.readAsDataURL` 把文件的原始字节编成 data URI 放进一个 `<img>`，`html>docx` 再把这份文档交给打包器；打包器会把 `<img>` 上那条 `data:` 从标记里提出来，写成包里一个**独立的 base64 媒体部件**并重写引用（提升逻辑与 MIME 无关，位图族都是同一趟）。所以三件事同时成立：像素与你给的字节逐字节相同（不解码、不上画布、不再压一次——这与 BMP→PNG / JPG / WebP 那几条必然重画完全不同）；`.docx` 里不残留 data URI；渲染这张图不需要从网络取任何东西。要说清的代价有两条。**一是体积**：BMP 常常完全不做压缩，base64 又每三个字节写成四个字符，这一路是搬运不是优化，所以产物不会变小——真要压体积得走会重编码的那几条。**二是类型前提**：只有当浏览器认出文件类型时前缀才是 `image/bmp`，而那是白名单内的引用类型；从 ZIP 解包出来的文件不带 MIME 类型，前缀会变成 `application/octet-stream`，它不在白名单里，会在写出前被剥掉——**那种场合连图片部件都不会生成**，Word 里显示成什么由打开它的程序决定，本页不替它担保。文档体是 altChunk：逐字排版、分页与图片的显示尺寸都由 Word 打开时决定，本扩展不产出 Word 原生段落、也不给缩放参数。还有一条边界照旧写明：不 OCR、不读画面内容，产物里唯一来自理解的字符串没有——只有你的文件名（进 `alt` 与 `title`，去扩展名、按属性位置转义）。BMP 在这套界面里**只能作输入**，浏览器不提供 BMP 编码器。全程只有 storage 权限，不联网。Two steps, and **neither one touches the pixels**: `bmp>html` encodes the file’s raw bytes with `FileReader.readAsDataURL` into a data URI inside an `<img>`, and `html>docx` hands that document to the packer, which lifts the `<img>`’s `data:` out of the markup and writes it as **its own base64 media part**, rewriting the reference (the lift is independent of MIME — every bitmap in this family takes the same trip). So three things hold at once: the pixels are byte-identical to what you handed over, because nothing decodes, resamples or recompresses — unlike BMP to PNG, JPG or WebP, which must redraw a canvas; no data URI survives in the `.docx`; and drawing the picture needs nothing from the network. Two costs, plainly. **Size first**: BMPs are usually uncompressed and base64 writes four characters per three bytes, so this route carries rather than optimises and the artifact does not shrink — if you need smaller, take a route that re-encodes. **The type premise second**: the prefix is `image/bmp`, a reference type inside the allow-list, only when the browser recognised the file’s type; a file unpacked from a ZIP carries no MIME type, so its prefix comes out as `application/octet-stream`, which is outside that list and is stripped before writing — **in that case no image part is created at all**, and what Word then shows is that program’s decision, which this page does not vouch for. The document body is an altChunk, so typesetting, paging and the picture’s displayed size belong to Word when it opens; no native Word paragraphs and no scaling knobs come from here. The usual boundary holds: no OCR, no reading of the picture, and the only string the route contributes is your file name (into `alt` and `title`, extension dropped, escaped for attribute position). And BMP stays an **input-only** format in this interface — browsers do not encode it. All of it on the storage permission, with no network.