发布于 2026-10-31 · CATAITO

PDF 压缩完全指南:文件为什么会大、怎么压才有效

一份扫描版合同 40MB 发不出去、一份幻灯片导出的 PDF 比原稿还大——先弄清体积都在哪,再决定怎么压。本文讲清 PDF 的体积构成(扫描图、嵌入字体、冗余结构)、三档压缩强度的取舍、哪些文件压不动,以及云端压缩与本地压缩的隐私差别。附免费在线 PDF 压缩工具,纯浏览器本地处理。

PDF 的体积都花在哪了

一个 PDF 动辄几十 MB,字节通常花在三个地方:

  • 扫描图 / 内嵌照片(大头):扫描仪默认 300 DPI 全彩输出,一页 A4 就是一张 2500×3500 像素的照片;100 页的扫描件等于 100 张高清照片,几十 MB 很正常。这是压缩空间最大的部分。
  • 嵌入字体:为了让没有安装该字体的电脑也能正确显示,PDF 会把用到的字体文件整个嵌进去,中文字体一套就能占 5–20MB。这部分通常不该动(动了会乱码)。
  • 冗余结构:编辑器反复保存会留下未引用的对象、旧版本页面数据。重新组织文件结构(对象流重存)能白捡一些体积。

所以压缩 PDF 的正确思路是:结构重存打底,重点重编码内嵌图片,不动文字与字体。

什么压缩是有效的

有效的压缩几乎都围绕「图片重编码」:

  1. 降采样:屏幕阅读根本用不着 300 DPI——把长边 3500px 的扫描图降到 1600–2400px,肉眼在屏幕上几乎看不出差别,像素数量却少了 60–80%。
  2. JPEG 重编码:质量 60–88 的 JPEG 对扫描图和照片来说损失极小,体积却能再砍一半以上。
  3. 结构重存:把对象重新组织成紧凑的跨引用流,对反复保存过的老文件有额外收益。

本站的 压缩 PDF 工具把这三步做在一起:重建文件结构 + 找出内嵌 JPEG 大图按所选强度降采样重编码。一条原则是宁可不压也不弄坏——带透明通道(SMask)的图、非 JPEG 格式的图、过小的图一律不碰;只有真的变小了才替换回文件。

三档强度怎么选

档位参数适合
强力长边 1600px · 质量 60邮件附件、聊天工具等有体积限制的传输
均衡(默认)长边 2400px · 质量 75大多数场景:屏幕阅读清晰、体积明显下降
轻微不缩放 · 质量 88打印、存档等对画质有要求的场景

三档都只动图片:文字与矢量内容按原始数据保留,可选中、可检索、可复制,任何档位下都不受影响。

哪些文件压不动

压缩不是万能的,四类文件收益很小:

  • 纯文字 PDF:没有大图可压,文字数据本身已经很小。
  • 已经优化过的 PDF:其他工具压过一遍的文件,二次压缩空间有限。
  • 图片是非 JPEG 格式(PNG / Flate 无损存储)的文件:本工具为保安全只重编码 JPEG 图片,无损格式的图不碰。
  • 体积主要来自嵌入字体的文件:动字体会乱码,属于「不该压的部分」。

所以正常的工具会如实标注:压完没变小的文件显示「收益有限」而不是假装成功。遇到这类文件,出路是换格式或重新生成(比如从源头重新导出更小的 PDF),而不是反复硬压。

云端压缩 vs 本地压缩

主流在线压缩器(iLovePDF、Smallpdf 等)把文件上传到服务器压缩后再给下载链接。压缩的对象往往是扫描合同、档案、医疗单据这类敏感文件,值得三思:文件离开设备,你无法控制服务器端副本的留存周期,上传大文件也慢。

另一条路线是纯浏览器本地压缩:解析、重编码、重组全部在浏览器里完成,文件不经过网络,断网也能用。本站的 压缩 PDF 工具就是这一路线,处理多少文件都不上传、不排队、不限次。

常见问题

Q:为什么我的 PDF 压缩后没变小?

A:压缩主要针对内嵌图片。纯文字、已优化过、或图片是无损格式的文件本来就没什么可压的,工具会如实标注。

Q:压缩会影响文字清晰度吗?

A:不会。文字与矢量按原始数据保留,只重编码位图图片;文字始终可选中、可检索。

Q:压缩强度选哪一档?

A:网传选强力,犹豫选默认均衡,打印存档选轻微。三档都只动图片。

Q:压缩后的文件还能恢复吗?

A:不能,降采样丢弃的像素找不回来。请保留原始文件,确认压缩结果没问题后再处理原文件。

Q:带密码的加密 PDF 能压缩吗?

A:不能,会被标记跳过。先在阅读器里另存为无密码副本再压缩。

Q:把文件上传到在线压缩工具安全吗?

A:主流工具会上传到服务器处理,敏感扫描件不建议。纯浏览器本地压缩文件不离开设备,更稳妥。

免费在线压缩 PDF

三档强度 · 前后体积对比 · 文字不受影响 · 纯浏览器本地处理

打开压缩 PDF 工具 →

Where the Bytes Live

When a PDF weighs tens of megabytes, the bytes usually live in three places:

  • Scans and embedded photos (the big one): scanners default to full-color 300 DPI — one A4 page is a 2500×3500-pixel photo; a 100-page scan is 100 of them. This is where compression wins the most.
  • Embedded fonts: so text renders on machines without the font installed, PDFs embed the font files — a CJK font alone can weigh 5–20MB. This part usually shouldn't be touched (touching it breaks text).
  • Redundant structure: repeated editor saves leave unreferenced objects and stale page data behind; reorganizing the file (object-stream rewrite) picks up some free savings.

So the right mental model: structure rewrite as the baseline, re-encode embedded images as the lever, never touch text or fonts.

What Actually Shrinks Files

Effective compression is almost all about re-encoding images:

  1. Downsampling: on-screen reading never needs 300 DPI — dropping a 3500px-long-edge scan to 1600–2400px is visually indistinguishable while cutting 60–80% of pixels.
  2. JPEG re-encoding: quality 60–88 costs scans and photos almost nothing visually while halving size again.
  3. Structure rewrite: re-packing objects into compact cross-reference streams gives older, repeatedly-saved files an extra shave.

Our Compress PDF tool does all three: rebuild the file structure, find large embedded JPEGs, and downsample + re-encode them at the chosen strength. One principle governs it: never compress at the cost of breaking the file — images with transparency (SMask), non-JPEG images, and tiny images are left alone, and a replacement only lands if it's actually smaller.

Choosing a Strength Level

LevelParametersBest for
Strong1600px long edge · quality 60Size-limited transfers: email attachments, chat apps
Balanced (default)2400px · quality 75Most cases: crisp on screen, clearly smaller
LightNo resampling · quality 88Print and archive — quality-sensitive work

All three levels only touch images: text and vectors are kept as original data — selectable, searchable, copyable — unaffected at any strength.

Files That Barely Compress

Compression isn't magic. Four kinds of files have little to gain:

  • Text-only PDFs: no big images to squeeze; text data is already tiny.
  • Already-optimized PDFs: files another tool has compressed once have limited room left.
  • Files whose images are non-JPEG (PNG / Flate lossless): to stay safe this tool only re-encodes JPEG images and leaves lossless ones alone.
  • Files where fonts dominate: tampering with embedded fonts corrupts text — that's the "shouldn't compress" part.

A trustworthy tool reports honestly: files that didn't shrink are marked "little to gain" instead of pretending success. For those, the way out is a different format or re-exporting from the source — not brute-forcing another pass.

Cloud vs Local Compression

Mainstream online compressors (iLovePDF, Smallpdf, etc.) upload your file to their servers and hand back a download link. What gets compressed is often exactly the sensitive kind — scanned contracts, archives, medical bills — so think twice: the file leaves your device, you don't control the server-side copy's lifetime, and big uploads are slow.

The alternative is in-browser local compression: parsing, re-encoding, and restructuring all happen in your browser with no network involved, offline included. Our Compress PDF tool works this way — no uploads, no queues, no limits, however many files you run.

FAQ

Q: Why didn't my PDF get smaller?

A: Compression targets embedded images. Text-only, already-optimized, or lossless-image files have little room — the tool marks them honestly.

Q: Does compression affect text sharpness?

A: No. Text and vectors are kept as original data; only bitmaps are re-encoded. Text stays selectable and searchable.

Q: Which strength level?

A: Strong for sharing, Balanced (default) when unsure, Light for print and archive. Images only, at every level.

Q: Can a compressed file be restored?

A: No — downsampled pixels are gone. Keep the original; only replace it after checking the result.

Q: Can I compress password-protected PDFs?

A: No — flagged and skipped. Save an unprotected copy from your PDF reader first.

Q: Is it safe to upload files to an online compressor?

A: Mainstream tools upload to their servers. For sensitive scans, prefer a tool that compresses entirely in your browser.

Free Online PDF Compressor

Three strength levels · Before/after sizes · Text untouched · 100% browser-local

Open Compress PDF Tool →