How to Make a PDF Smaller (and What You Give Up)

Last updated: 2026-10-22 · Checked and published by Leo, who runs Hanji

Why PDFs get big, how to shrink one while keeping the text, measured results on six test files, and how small your page count can get.

Most advice on shrinking a PDF stops at “use a compressor”. That works until a form says 200 KB and your file is 9 MB, or until the result comes back blurry and you cannot tell why. This guide explains where a PDF’s size actually comes from, what each compression method does to your pages, and how to plan around a hard limit before you start.

All figures below were measured with Hanji’s own tools in Chrome on Windows on 2026-10-22, on test PDFs we generated to behave like everyday documents: photos of paper at 12 megapixels, a typed report, a form, screenshots and a vector drawing. The tool has changed since our first measurements, so every number was measured again. 1 KB here means 1,000 bytes and 1 MB means 1,000,000 bytes.

Why some PDFs are huge and others are tiny

A PDF is a container. It can hold text, fonts, vector drawings (lines, tables, logos) and images, and each of these costs very different amounts of space.

  • Text is stored as characters plus positions. A whole page of it is a few kilobytes. A 5-page PDF of plain text we tested was 6.7 KB in total.
  • Fonts are embedded so the file looks the same everywhere. A Latin font is usually small, but Korean, Japanese and Chinese fonts contain thousands of characters. When a program embeds the whole font instead of only the characters used, a one-page letter can carry several megabytes of font data. No compressor can remove it without breaking the text.
  • Images are where almost all large PDFs get their size. A photo from a modern phone is 12 megapixels (4032×3024). Put four of those on four A4 pages and you have the 9.4 MB “phone scan” we use as our main test file below.

The image problem is a mismatch of resolution. A 4032-pixel photo stretched across the 11.7-inch height of an A4 page works out to about 345 pixels per inch. Printers need around 300 for sharp output, and a screen shows a whole page at well under 150. Most of those pixels are never seen.

Scanners make the same mistake in a different way. An A4 page scanned at 600 DPI in colour is about 4960×7016 pixels, roughly 35 million pixels per page, nearly three times a phone photo. For a signed form that will be read on a screen, around 150 DPI is plenty (Epson’s scanner guide suggests 96–150 dpi for email and on-screen viewing, and 200–300 dpi for printing or text recognition), and greyscale is enough for black ink.

The two methods, and what each one does

Hanji’s Compress PDF offers two methods because they solve different problems.

Compress images (keep text) finds each photo or scan inside the PDF, shrinks it so its longest side is at most a set number of pixels, and saves it again as JPEG. Text, fonts, links and vector drawings are not touched, so text stays sharp and searchable. The three strengths are:

Strength Longest side of each image JPEG quality Roughly, for a full-page A4 image
Light 2500 px 80 about 210 DPI
Recommended 1800 px 68 about 150 DPI
Strong 1200 px 50 about 100 DPI

An image is only replaced when the new version is at least 10% smaller, so a PDF that is already well compressed comes back nearly unchanged. Images in unusual formats (for example 1-bit black-and-white scans, CMYK images, or images with a stencil or colour-key mask) are left as they are. Images with ordinary soft transparency are recompressed and keep their transparency.

Fit a target size (keeps the text when it can) runs that same image compression for you. It tries five strengths, the three above plus two harder ones (longest side 900 px at quality 40, then 650 px at quality 30), and stops at the first file that fits your limit, so it never squeezes harder than it must and does not pad a small file up to the target. Text stays text. Only if even the hardest strength is too big, or the PDF has no pictures it can shrink, does it fall back to redrawing every page as a picture, starting at 150 DPI with JPEG quality 75 and stepping down to 60 DPI at quality 35 until the file fits. The result tells you which of the two happened.

The cost of that last step is real: the pages become images, so text can no longer be selected, searched or copied, and a screen reader cannot read it. For a scanned document that was an image to begin with, you lose nothing extra.

Measured: real-style files, every setting

We ran the real tool in Chrome on 2026-10-22. The test PDFs: a 4-page phone scan (four full 12 MP photos of paper on A4 pages, 9.4 MB) and its first page (2.3 MB); a 2-page application form (typed fields, a photographed ID card and a photographed signed page, 4.3 MB); a 14-page typed report with three photos (1.7 MB); an 8-page report with three PNG screenshots (0.6 MB); and a 3-page site plan made only of vector lines, with no pictures (3.6 MB).

The 4-page phone scan under every setting:

Method Result Change
Compress images, Light 2.4 MB 75% smaller
Compress images, Recommended 1.02 MB 89% smaller
Compress images, Strong 477 KB 95% smaller
Fit target 1 MB 477 KB images only, text kept
Fit target 500 KB 477 KB images only, text kept
Fit target 200 KB 131 KB images only, text kept
Fit target 100 KB does not fit the tool says so

All the files at each target:

File Original 100 KB 200 KB 500 KB 1 MB
1-page phone scan 2.3 MB 64 KB 119 KB 254 KB 593 KB
4-page phone scan 9.4 MB does not fit 131 KB 477 KB 477 KB
2-page form 4.3 MB 74 KB 134 KB 358 KB 358 KB
14-page typed report 1.7 MB 46 KB 100 KB 483 KB 483 KB
8-page report, PNG screenshots 0.6 MB 61 KB 188 KB 423 KB already under
3-page vector site plan 3.6 MB does not fit does not fit 389 KB, pages redrawn at 80 DPI 902 KB, pages redrawn at 120 DPI

Everything except the site plan kept its text selectable: in the form and the two reports we extracted the text from the result and found it all there. The site plan had no pictures for the image step to shrink, so its pages were redrawn as pictures, and the result note said so with the DPI used.

Two things stand out. First, the image step alone got almost every file under the limit, so text rarely has to go. If your limit is 1 MB or more, you will usually keep it. Second, results often land well below the limit: with five strengths and no padding, the 1 MB target gave 477 KB for the 4-page scan, because the next gentler strength was 1.02 MB, just over.

How small is realistic for your page count

From the 4-page phone scan, a photographed page costs roughly:

  • About 255 KB per page at Recommended (1.02 MB for four pages): sharp on screen, fine for reading a form. This fits a 1 MB limit for about four pages.
  • About 120 KB per page at Strong (477 KB for four pages): normal print is readable, fine print and stamps are softer. This is the 500 KB target for about four pages.
  • About 33 KB per page at the hardest strengths (131 KB for four pages): headings and normal print stay legible on screen, small print and handwriting go soft.
  • Under about 25 KB per page did not fit: four pages could not get under 100 KB even with the pages redrawn down to 60 DPI.

Pages that are mostly text weigh almost nothing, so for a typed report only the photos count: our 14-page report fitted 100 KB with all its text intact. If a file cannot be made small enough, the tool says so instead of handing you something unreadable.

Split first, then compress

The simplest way to make a PDF smaller is to send fewer pages. Forms often ask for one document but people attach the whole bundle: the cover letter, the blank back sides, the instructions page that came with the form.

Open the file in Split PDF, keep only the pages that were asked for, and compress what remains. Dropping from 6 pages to 3 doubles the space each page gets under the same limit, which is often the difference between a sharp page and a soft one. If a portal accepts several files, sending each document separately under its own limit is better than squeezing them into one.

What you lose at each step

Think of it as a ladder and stop at the first rung that fits:

  1. Compress images, Light or Recommended. Text stays text. Photos lose detail you would only notice zoomed in.
  2. Compress images, Strong. Text still stays text. Photos and scanned pages look soft at full screen, so check small print and stamps.
  3. Fit a target, generous limit (1 MB or 500 KB for a few pages). Text stays selectable; the pictures look like a good scan.
  4. Fit a target, tight limit (200 KB, 100 KB). Text usually stays text, but the pictures look like a fax or a low-resolution scan. If even that is too big, the pages are redrawn as pictures and the text can no longer be selected. Check that every number is still readable before you submit.

Each step is a new file; your original is never changed, so you can always go back up the ladder.

Common mistakes

  • Building a PDF from full-size photos and then fighting the size. If you are making the PDF yourself, choose a lower image quality at that point. See turning phone photos into a clean PDF for measured sizes.
  • Printing photos to PDF. “Save as PDF” or “Microsoft Print to PDF” in a print dialog gives you no control over image quality, and the result depends on what the app hands to the printer. A tool that lets you pick quality is more predictable.
  • Scanning at 600 DPI in colour “to be safe”. An A4 page at 600 DPI has nearly three times the pixels of a phone photo. Use about 150 DPI for documents read on screen and 200–300 DPI for ones that will be printed, in greyscale for black-and-white documents.
  • Compressing an already compressed file again and again. Each JPEG round loses a little more. Go back to the original and pick a stronger setting instead.
  • Assuming the target method always flattens your text. It does so only as a last resort, and the result note says whether it happened. If you need the text and pages were redrawn, try a larger limit, or remove pages with Split PDF first. If Compress images finds nothing to compress, the size is coming from embedded fonts or vector drawings, and removing pages is the cleaner fix.
  • Not checking the result. Open the compressed file and zoom to 100% on the smallest text, ID numbers and signatures before uploading.

If the file is going to an upload form, the guide to upload size limits explains how forms count KB and MB and what to do when a limit seems impossibly small.

Guides