# How to Reduce PDF File Size Without Ruining It: What Makes PDFs Big and What Each Method Actually Does

Shrink the images, not the text: downsample photos and scans to 150–200 DPI and keep text as text. We measured every method; a 12 MB scan fell by 90%.

---

- **Canonical URL:** https://dothecalculation.com/blog/files/how-to-reduce-pdf-file-size
- **Category:** Files & PDFs
- **Author:** Do The Calculation Team
- **Published:** 2026-09-29
- **Reading time:** 26 min read
- **Publisher:** Do The Calculation (https://dothecalculation.com)
- **Methodology:** https://dothecalculation.com/methodology

---

## How to reduce PDF file size: the short answer

To make a PDF smaller without ruining it, shrink what is actually heavy, which is almost always the images. Recompress photos and scanned pages at about 150 to 200 DPI while the text stays real text, let a lossless clean-up remove duplicates and leftovers, and if you still have the original document, export it again with a smaller setting instead of printing it to PDF. Do not turn text pages into pictures: in our test that made a 215 KB report almost seven times bigger.

The rest of this guide explains where the bytes in a PDF live, what each method really changes, and what happened when we made the same ten-page document in six different ways and ran every version through every compression level. The numbers are byte counts from files we made and measured on 29 September 2026, not estimates.

Tool: [Compress a PDF in your browser](https://dothecalculation.com/tools/compress-pdf) — Clean-up, image recompression and a target-size mode, with a text check before and after. Your file is processed on your device and never uploaded.

## What makes a PDF big: the anatomy of a file

A PDF is a container. Inside it are pages, and each page is a set of drawing instructions that refer to resources stored elsewhere in the file: fonts, images, colour profiles, form fields, annotations. The PDF standard, ISO 32000, describes images as image XObjects that a page paints with a single instruction, and it defines a list of compression filters those objects can use, including Flate (the zlib/deflate method ZIP uses), DCT (JPEG), JPX (JPEG 2000), JBIG2 and CCITT fax for black-and-white images (https://archive.org/details/pdf320002008_201911). Which resources a file holds, and how each is compressed, decides almost all of its size.

Text on its own is cheap. In our tests a 200-page PDF of plain text weighed 234 KB, a little over 1 KB a page. When a PDF is large, the reason is nearly always on the list below, and usually near the top of it.

### Images and their resolution

A photo is stored as a grid of pixels, and the size of that grid has nothing to do with how big the photo looks on the page. A 4000 × 3000 pixel photo placed 14 cm wide in a report is printed at about 726 pixels per inch, more than three times the roughly 200 that Word itself kept when it exported our report. Unless something reduces it, the PDF keeps every one of those pixels. The number that matters is the effective resolution: the image’s pixels divided by the inches it occupies on the page. If you want to see how pixels, inches and density relate, the [screen resolution and PPI calculator](/calculators/screen-resolution-ppi-calculator) does the same arithmetic for displays.

**Pixels in an image, and its size before compression**

```
pixels = (width in inches × DPI) × (height in inches × DPI); raw bytes = pixels × bytes per pixel
```
- Bytes per pixel: 3 for colour (24-bit RGB), 1 for greyscale (8-bit), 1/8 for black and white (1-bit).
- A US Letter page at 300 DPI is 2550 × 3300 = 8,415,000 pixels, about 24 MB of raw colour data before compression.
- Doubling the DPI multiplies the pixel count by four, because both sides double.

### Colour depth

A colour pixel carries three values, a greyscale pixel one, a black-and-white pixel a single bit. That is why a colour scan of a plain typed letter carries far more data than the page needs, and why scanner apps offer greyscale and black-and-white modes. Compression narrows the gap, but it starts from very different amounts of data.

### Scan settings

A scanned page is one picture of the whole sheet, including the white paper, so a scan’s size is set almost entirely by its DPI, its colour mode and its JPEG quality. The scanner or app decides these when you press the button; whatever it produces, the PDF simply wraps. Our 17-page letter scanned at 300 DPI weighed 40.7 MB, about 2.4 MB a page. That is roughly two thousand times the 1 KB a page of the text-only PDF above, for the same kind of content.

### Embedded fonts, and why subsetting matters

So that a PDF looks the same on every device, the fonts it uses are usually embedded in it. Adobe’s documentation describes the two choices: embedding the entire font, or embedding a subset containing only the characters used in the file, and notes that embedding slightly increases file size (https://helpx.adobe.com/acrobat/desktop/create-documents/explore-advanced-conversion-settings/font-handling-distiller.html). For a Latin font the difference is modest. For a font covering Chinese, Japanese or Korean, a full embed can be very large, which is why a short letter in one of those languages is sometimes several megabytes. Word’s own exports subset: both of our Word PDFs carried Calibri and Calibri Bold as subsets.

### Duplicated resources

Some programs store the same logo, background or font again for every page instead of once for the whole file, and merging PDFs brings in each source file’s own copies. In one of our tests a 40-page PDF with the same logo stored on every page weighed 402 KB; merging the identical copies, with nothing else changed, took it to 28 KB.

### Attachments, metadata and thumbnails

A PDF can carry whole files as attachments, and some electronic invoice formats deliberately travel that way, with machine-readable data attached to the visible invoice. Metadata (the document information and XMP packet with title, author and editing history) and embedded page thumbnails are usually small, a few kilobytes, but they add up across a batch. Adobe’s PDF Optimizer lists discarding embedded page thumbnails among its options (https://helpx.adobe.com/acrobat/desktop/create-documents/optimize-pdfs/pdf-optimizer-settings.html).

### Incremental updates

When some programs save changes to a PDF, they do not rewrite the file. The standard allows an incremental update, in which, in the words of ISO 32000, “changes shall be appended to the end of the file”, leaving the original contents intact. That is quick, and it is how signed PDFs can be added to without breaking earlier signatures, but it means a file edited many times can still contain every earlier version of a replaced image. Rewriting the file drops objects that nothing refers to any more.

### Forms and annotations

Fillable fields, comments, stamps and highlights each store their own appearance, a small drawing of how they look, in addition to their data. One comment weighs almost nothing; a form with hundreds of fields, or a review copy with a stamp image pasted on every page, can add noticeably. Our 2-page form test went from 48 KB to 35 KB with a lossless clean-up alone.

_[Figure: The same ten pages, six different file sizes — One 10-page report with four photos, a chart and a table, turned into a PDF six ways. Nothing about the content changed; only how the PDF was made.]_

## The experiment: one document, six PDFs, every compression level

Advice about PDF size is easy to repeat and hard to check, so we built a test document and measured what each method does to it. Here is exactly how the test was made, including its limits.

- The document: a 10-page US Letter survey report written as a Word .docx in Calibri 11 pt, with headings and body text on every page, four photos, one bar chart (an 1800 × 1000 pixel PNG) and one table. The .docx itself was 6,001,238 bytes (5.7 MB), because Word keeps the original photos inside it.
- The photos: each a 4000 × 3000 pixel JPEG of about 1.5 MB, generated for the test as synthetic scenes with added grain. They are smoother than real camera photos, so they compress better than yours probably will. Treat our percentages as a best case for photos.
- The exports: made with Microsoft Word 16.0 (Microsoft 365) on a Windows 10 PC, using Word’s own Save as PDF with each of its two quality options, and separately printed with Microsoft Print to PDF.
- The scans: simulated rather than made on a physical scanner. Each page of the Standard export was drawn at 150, 300 and 600 DPI, given paper grain and a slight blur, and saved as a JPEG at quality 85, one image per page with no text layer. That is the structure of a typical scanner’s PDF.
- The compression: every file was passed through our [Compress PDF tool](/tools/compress-pdf) at each level, run in headless Chromium on the same PC. For each run the text of every page was extracted before and after and compared.
- Not tested: macOS Preview, because we had no Mac for this test. Where this guide describes Preview, it relies on Apple’s documentation.

### Step 1: how the PDF was made

**The same 10-page report, made into a PDF six ways (sizes in bytes; KB and MB are binary, 1 MB = 1,048,576 bytes)**
| How it was made | Bytes | Size | What was inside |
| --- | --- | --- | --- |
| Word, Save as PDF, Standard (publishing online and printing) | 219,897 | 215 KB | Each photo downsampled to 1102 × 826 px (about 200 ppi at 14 cm wide), about 23 KB each; Calibri fonts subset; 14,808 characters of text |
| Word, Save as PDF, Minimum size (publishing online) | 151,690 | 148 KB | Photos 529 × 396 px (about 96 ppi), about 7 KB each; same fonts and text |
| Microsoft Print to PDF | 3,023,943 | 2.9 MB | Photos kept at their full 4000 px width, cut into strips of 250 to 500 px (33 image objects); text still text (14,790 characters) |
| Simulated scan, 150 DPI | 1,987,135 | 1.9 MB | One 1275 × 1650 px JPEG per page; no text layer |
| Simulated scan, 300 DPI | 5,186,691 | 4.9 MB | One 2550 × 3300 px JPEG per page; no text layer |
| Simulated scan, 600 DPI | 12,712,759 | 12.1 MB | One 5100 × 6600 px JPEG per page; no text layer |

Two things stand out before any compression. First, Word’s own export already did most of the work: it reduced each 4000-pixel photo to about 200 pixels per inch at the size it was placed, and stored it as a small JPEG. Second, printing the same document to PDF produced a file about 14 times the size of Word’s export, because the print route passed the photos through at full resolution. The text survived in both, so neither file looks wrong. Only the size gives it away.

### Step 2: each file through each compression level

Our compressor has a lossless Clean-up level and three image presets. Light re-encodes images as JPEG at quality 85 and caps them at 300 DPI at their printed size (long edge at most 2400 px); Balanced uses quality 75, 200 DPI and 2000 px; Strong uses quality 60, 150 DPI and 1600 px. Text, fonts and drawings are left alone at every level.

**Result in bytes, with the change from the original (measured 29 September 2026)**
| File | Original | Clean-up (lossless) | Light | Balanced | Strong |
| --- | --- | --- | --- | --- | --- |
| Word Standard export | 219,897 | 216,267 (−1.7%) | 190,933 (−13.2%) | 184,076 (−16.3%) | 157,630 (−28.3%) |
| Word Minimum size export | 151,690 | 148,061 (−2.4%) | 146,742 (−3.3%) | 144,061 (−5.0%) | 141,797 (−6.5%) |
| Microsoft Print to PDF | 3,023,943 | 3,019,701 (−0.1%) | 757,401 (−75.0%) | 558,766 (−81.5%) | 515,526 (−83.0%) |
| Scan, 150 DPI | 1,987,135 | 1,985,673 (−0.1%) | 1,827,001 (−8.1%) | 1,616,068 (−18.7%) | 1,193,324 (−39.9%) |
| Scan, 300 DPI | 5,186,691 | 5,185,231 (0.0%) | 2,926,065 (−43.6%) | 1,901,822 (−63.3%) | 1,213,732 (−76.6%) |
| Scan, 600 DPI | 12,712,759 | 12,711,297 (0.0%) | 2,920,302 (−77.0%) | 1,924,216 (−84.9%) | 1,227,828 (−90.3%) |

Every run passed the text check. For the Word files, the text extracted after compression was identical to the text before it; the scans had no text layer, so there was nothing to change. Each file took between 11 milliseconds and 2 seconds.

_[Figure: Three scans of the same pages end up the same size — The 150, 300 and 600 DPI scans through each compression level. The extra pixels of the high-resolution scans are the first thing compression removes.]_

### Step 3: a target size, and rasterising

**Target-size mode and the rasterise level**
| Test | Before | After | What happened |
| --- | --- | --- | --- |
| 600 DPI scan, target “smaller than 2 MB” | 12,712,759 | 1,924,216 | The tool searched its settings and chose the Balanced-equivalent step (quality 75, 200 DPI, 2000 px) in 6.8 seconds |
| 300 DPI scan, rasterised at 150 DPI | 5,186,691 | 1,342,085 (−74%) | Smaller, but no better than Strong (1,213,732), and the scan had no text to lose |
| Word Standard export, rasterised at 150 DPI | 219,897 | 1,516,678 (+590%) | Almost seven times larger, and the text was gone: every page became a picture |

## Six lessons from the numbers

The test was small, but the pattern in it follows directly from how PDFs store images, so it is worth spelling out.

- The way a PDF is made decides its size more than any compressor. If you have the original document, exporting it again with the right setting beats compressing the PDF afterwards: Word’s Minimum size export (148 KB) was smaller than every compressed version of the Standard export, even Strong (154 KB), and it got there without a second lossy step.
- Printing to PDF can be far bigger than saving as PDF. In our test it was about 14 times bigger. Use the application’s own Save as PDF or Export command when there is one.
- Scanning above what you need buys you nothing after compression. The 600 DPI scan was 6.4 times the size of the 150 DPI one, yet after Balanced compression the 300 and 600 DPI scans both ended at about 1.8 MB, and after Strong all three scans landed between 1,193,324 and 1,227,828 bytes. The extra detail was captured, stored, sent and then thrown away.
- File size does not grow as fast as pixel count. Going from 150 to 600 DPI multiplied the pixels by 16 but the file by 6.4, because JPEG compresses smooth, enlarged detail efficiently. It still grew six-fold.
- Compressing a badly made file is not the same as making it well. The Print to PDF file at Balanced was 558,766 bytes, still about two and a half times Word’s own Standard export, and an already-efficient file has little left to give: the Minimum size export lost only 6.5% even at Strong.
- Clean-up is safe but rarely dramatic on a tidy file (0 to 2.4% here). Its big wins come from duplicates, uncompressed data and leftovers you cannot see from the outside, like the 40-page logo file, so it is worth running first on anything.

> **Why our photo figures are a best case** — Our test photos were generated, with grain added, and they are smoother than photos from a real camera or phone. Real photos carry more fine detail and sensor noise, and JPEG has to spend more bytes on them. Expect the same direction of change on your own files, with somewhat smaller percentages on photo-heavy documents.

## What each compression method really does

Every PDF compressor, whether it is built into an app, sold as software or offered online, uses some mix of the methods below. Knowing which one a tool is using tells you what you can lose.

### Lossless clean-up: object streams, duplicates and leftovers

A lossless pass rewrites the file without changing anything you can see. It can pack the many small objects of a PDF into compressed object streams, a feature ISO 32000 introduced in PDF 1.5 “to allow indirect objects other than streams to be stored more compactly”; merge identical images and fonts into one copy; compress any data stored uncompressed or in a text encoding; and drop objects nothing refers to, such as images replaced in earlier edits. It is the right first step for every file, and the only step for files where every pixel matters.

### Removing metadata, thumbnails and attachments

Stripping the document information, XMP metadata and embedded thumbnails saves a few kilobytes and removes details such as the author name and editing software, which is sometimes the more useful effect. Removing attachments can save a great deal, but check what they are first: an electronic invoice may need its attached data file, and a portfolio may be mostly attachments.

### Image downsampling

Downsampling reduces the number of pixels in an image. Adobe describes it as lowering image resolution “by merging the colors of original pixels into larger pixels” (https://helpx.adobe.com/acrobat/desktop/create-documents/optimize-pdfs/pdf-optimizer-settings.html). Done by effective resolution, it is mostly invisible: an image printed 14 cm wide does not need 4000 pixels to look sharp at 100% zoom or on paper. Around 150 DPI suits reading on screen, 200 DPI is comfortable for zooming and ordinary printing, and 300 DPI is print quality. Below 150 DPI, small text in scans and fine detail in photos begin to soften.

### JPEG re-encoding

Saving an image again as JPEG at a lower quality setting throws away detail the eye is less sensitive to. The Library of Congress describes JPEG’s standard modes as lossy compression built on discrete cosine transforms (https://web.archive.org/web/20260829020120/https://www.loc.gov/preservation/digital/formats/fdd/fdd000017.shtml). Lossy means what is discarded cannot be restored, and each further re-encode discards a little more. At quality 75 to 85 the change is usually hard to see on photos; lower settings start to show blocky edges around text and fine lines. JPEG suits photos and poorly suits charts, screenshots and diagrams, which are better kept lossless; our guide to [PDF, JPG and PNG](/blog/files/pdf-vs-jpg-vs-png) explains why.

### Rasterising: redrawing every page as a picture

Rasterising draws each page, text and all, as one image and replaces the page with that picture. It guarantees a predictable result for any content, which is why some tools use it, and it is the only way to shrink pages whose images a compressor cannot otherwise handle. But text becomes pixels: it can no longer be selected, searched or read aloud by a screen reader, links and form fields disappear, and on a text-heavy page the picture is usually bigger than the text it replaced, as our +590% result showed. Use it only on a scan that has no text layer anyway, and only if gentler methods cannot reach your target.

### Font subsetting and unembedding

If a file embeds whole fonts, replacing them with subsets of only the characters used can save a lot on large or East Asian fonts. Going further, a font can be unembedded altogether. Adobe’s PDF Optimizer allows this, saying that if you do not require an exact match and prefer a smaller file you can skip embedding fonts, which means another computer may substitute a different font. Our compressor does neither: the library it writes PDFs with cannot subset existing fonts, and we would rather say so than pretend. When fonts are the weight, the fix is to export again from the source with subsetting switched on.

### OCR of scans

Optical character recognition reads the text in a scanned image and adds it to the page as an invisible text layer, so the scan becomes searchable and copyable. It adds bytes rather than removing them. Run OCR for searchability, not size, and do it before compressing if your OCR program needs a sharp image to read small type.

### PDF/A caveats

PDF/A is the archival form of PDF. The Library of Congress notes that in PDF/A-1 all fonts must be embedded and encryption is disallowed (https://web.archive.org/web/20221006120332/https://www.loc.gov/preservation/digital/formats/fdd/fdd000125), so unembedding fonts or adding a password breaks conformance. PDF/A-1 is built on PDF 1.4, and object streams only arrived in PDF 1.5, so a compressor that writes object streams produces a file that is no longer valid PDF/A-1. PDF/A-2, built on PDF 1.7, added support for JPEG 2000 compression (https://web.archive.org/web/20260731124625/https://www.loc.gov/preservation/digital/formats/fdd/fdd000319.shtml). If a file must stay PDF/A, compress it with software that can write and check PDF/A, and keep the archival master untouched.

_[Figure: Three levels of compression, and what each can cost you — From safest to most destructive. Most files only need the first two.]_

## Built-in options on Mac, Windows, Office, iPhone and Android

You may already have a way to make a smaller PDF. Here is what each official source documents, and what we measured where we could.

### macOS Preview: the Reduce File Size Quartz filter

In Preview, open the PDF, choose File > Export, pick Reduce File Size from the Quartz Filter menu, and save (https://support.apple.com/guide/preview/reduce-the-size-of-a-pdf-prvw1509/mac). Apple’s page also lists options to optimise images for screen and save images as JPEG, and warns that the compressed PDF may be of lower quality than the original, with the size and image quality depending on the options you choose. We could not run Preview for this test, so we have no measured figures for it. Whatever you use, compare the result with the original at 100% and 200% zoom, especially small text in scans, and save it as a new file so the original survives.

### Windows: Microsoft Print to PDF

Printing to the Microsoft Print to PDF printer is a common way to “save” something as a PDF on Windows, and it is often suggested as a way to shrink one. In our test it did the opposite: the Word report printed this way was 3,023,943 bytes against 219,897 for Word’s own Save as PDF, about 14 times bigger, with the photos kept at full resolution and cut into 33 strips. That is one test with one application, but it is enough to say that printing to PDF is not a reliable way to reduce size. If the application has its own Save as PDF or Export, use that.

### Word, PowerPoint and Excel: Minimum size and Compress Pictures

Microsoft’s Office apps offer two quality options when saving as PDF: Standard (publishing online and printing) for documents that need high print quality, and Minimum size (publishing online) when file size matters more (https://support.microsoft.com/en-us/office/save-or-convert-to-pdf-or-xps-in-office-desktop-apps-d85416c5-7d77-4fd6-a216-6f4bf7c7c110). In Word 16.0, Minimum size took our report from 219,897 to 151,690 bytes, with photos at about 96 ppi instead of 200.

You can also shrink the pictures inside the document before exporting. Microsoft states that the default picture resolution in Microsoft 365 is 220 ppi (https://support.microsoft.com/en-us/office/graphics-visuals/reduce-the-file-size-of-a-picture-in-microsoft-office), and its Word guidance suggests choosing 150 ppi or lower as the default resolution, saying that in most cases you will not need more. The same page advises, if you embed fonts, to embed only the characters used and not to embed common system fonts (https://support.microsoft.com/en-us/word/reduce-the-file-size-of-your-word-documents). Cropped parts of pictures can be deleted too, which Microsoft mentions as a further saving.

### Adobe Acrobat: Reduced Size PDF and PDF Optimizer

Acrobat has a one-step command, Menu > Save as Other > Reduced size PDF, where you choose which Acrobat versions the file must stay compatible with (https://helpx.adobe.com/acrobat/desktop/create-documents/optimize-pdfs/reduce-file-size.html). Its PDF Optimizer gives separate control over image downsampling and compression, font embedding and objects to discard, such as embedded thumbnails and document tags. Be careful with the last: tags carry the structure screen readers rely on, so discarding them trades accessibility for bytes.

### iPhone and iPad: Notes and Files

The Notes and Files apps both have Scan Documents, which saves the pages as a PDF (https://support.apple.com/en-us/108963). Apple’s page describes capturing, adjusting corners and filters, and gives no file-size or resolution option. If a phone scan comes out large, compress it afterwards, or photograph the pages and build the PDF with a converter that lets you choose the image size.

### Android: Google Drive scanning

The Google Drive app on Android scans documents to PDF or JPG, with crop and rotate, a clean-up tool and a filter that includes greyscale (https://support.google.com/drive/answer/3145835?hl=en&co=GENIE.Platform%3DAndroid). Its help page lists no size or resolution setting either. Choosing greyscale for black-and-white paperwork is the one size-relevant choice it documents: one channel of data per pixel instead of three.

## Scan smaller from the start: DPI, colour and format

For scans, the cheapest compression is the one you do at the scanner. Our test shows why: a 600 DPI scan of ten ordinary pages was 12.1 MB, and after compressing it to fit a 2 MB limit it was the same size as the 300 DPI scan compressed the same way. Everything above what the destination needs is wasted.

- For email, portals and reading on screen: 150 to 200 DPI. Our 150 DPI scan of ten pages was 1.9 MB before any compression.
- For printing or documents with small print: 300 DPI, then compress with the Light setting if needed.
- For archives: follow the archive’s own rules. NARA’s 2002 guidance for scanned textual records recommended 400 ppi for greyscale and colour and 600 ppi for black and white, and did not accept lossy compression; NARA now marks that page as superseded, so check its current guidance before relying on those figures (https://www.archives.gov/records-mgmt/initiatives/scanned-textual.html). Archival scans are masters, not the copies you email.
- Colour mode: greyscale for typed or handwritten paperwork, colour only when colour carries meaning (stamps, highlighted passages, photos).
- Format: if you photograph pages with a phone, you can build the PDF with the [JPG to PDF converter](/tools/jpg-to-pdf), which lets you choose image size and quality while the PDF is being made, so it starts small.

## When not to compress a PDF

Compressing always creates a new file. For most documents that is fine. For a few, the exact original is the point, and a smaller copy is a different document.

- Digitally signed files. A digital signature covers the exact bytes that were signed, so any rewrite, even a lossless one, leaves a copy whose signature no longer validates. Send the signed original, and if it is too large, ask the signer for a smaller signed version.
- Legal and evidential originals. A court filing, a deed or a notarised document may need to be exactly as issued. Compress a copy for reading if you must, and keep the original as the record.
- Print-ready artwork. Files prepared for a printer often use CMYK colour and high-resolution images deliberately. Downsampling to screen resolution can make them print soft; send them the way the printer asks, often by a file transfer link.
- Archival PDF/A. As above, many compression steps break PDF/A conformance. Keep the master and compress a separate working copy.
- Files with attachments you have not checked. Removing attached files can remove the machine-readable part of an e-invoice or the spreadsheet behind a report.

## What to do, by type of document

Find the row that matches your file. The expected results are from our own measurements, so they show the scale of change, not a promise for your file.

**Decision table: document type, cause, best method, measured result**
| Document type | What makes it big | Best method | What we measured |
| --- | --- | --- | --- |
| Word, PowerPoint or Excel file, and you have the source | Full-size photos placed small | Export again with Minimum size, or compress pictures to 150 ppi first | 215 KB (Standard) to 148 KB (Minimum size) |
| An exported report, and you only have the PDF | Photos at print resolution | Image recompression, Balanced or Strong | 215 KB to 180 KB (Balanced) or 154 KB (Strong), text identical |
| A PDF made with Print to PDF | Photos at full resolution, cut into strips | Save as PDF from the app again; otherwise Balanced | 2.9 MB to 546 KB (Balanced); Save as PDF gave 215 KB |
| Scanned paperwork (JPEG pages) | Pixel count: DPI and colour | Rescan at 150–200 DPI, or Balanced/Strong; target size for a limit | 12.1 MB (600 DPI) to 1.8 MB (Balanced) or 1.2 MB (Strong) |
| Phone photos made into a PDF | Multi-megapixel photos | Image recompression, or rebuild with smaller images | 3-page photo report: 3.1 MB to 100 KB (Balanced) |
| Merged file, or a logo on every page | The same image or font stored many times | Lossless clean-up | 402 KB to 28 KB, nothing visible changed |
| PDF with uncompressed images | Raw pixel data | Clean-up first (lossless), then images if needed | 11.2 MB to 5.0 MB lossless; 510 KB at Balanced |
| Long text-only document | Text, fonts and drawing instructions | Clean-up; for more, export again with subset fonts | 234 KB to 197 KB; the image level has nothing to act on |
| Signed, legal or PDF/A original | Whatever it is | Do not compress the original; send it, split it or share a link | Keep the original |

If a single file is still over a limit after the right method, splitting it is the next step: the [Split PDF tool](/tools/split-pdf) can divide a file by size so each part fits. For email limits, portal limits and every way to get a large file through, see [How to send a large PDF](/blog/files/how-to-send-a-large-pdf).

## A safe order of work

- Keep the original. Save every compressed result as a new file.
- Check what kind of file you have: try selecting a sentence. If you cannot, the page is a picture, usually a scan.
- If you have the source document, export it again with a smaller setting before reaching for a compressor.
- Run a lossless clean-up. It costs nothing visible.
- Recompress images at Balanced (200 DPI, quality 75). Go to Strong only if you need to.
- If you have a hard limit, use a target-size setting so the tool finds the gentlest step that fits.
- Compare the result with the original at 100% and 200% zoom, on the page with the smallest text.
- Rasterise only a scan with no text layer, and only as a last resort. If the file is confidential, prefer a tool that does not upload it; our guide on whether [online PDF converters are safe](/blog/files/are-online-pdf-converters-safe) explains how to check.

Tool: [Try the Compress PDF tool](https://dothecalculation.com/tools/compress-pdf) — Lossless clean-up, three image presets and a target-size mode. It reports what it changed and checks the text before and after, all in your browser.

More guides on formats, conversion and sending files are collected in [Files and PDFs](/blog/files).

## Frequently asked questions

**Why is a 3-page PDF 20 MB?**

Almost certainly images: phone photos or scans kept at full resolution, sometimes stored uncompressed. Text costs about 1 KB a page; a 300 DPI scanned page cost us about 2.4 MB; a page with two uncompressed photos weighed 11.2 MB. In our tests, image recompression took a 3.1 MB photo report to 100 KB and that 11.2 MB page to 510 KB.

**Does printing to PDF reduce file size?**

Not reliably. In our test, Microsoft Print to PDF made a 10-page Word report 3,023,943 bytes, about 14 times the 219,897 bytes of Word’s own Save as PDF, because it kept the photos at full resolution. Use the application’s own Save as PDF or Export, and choose a smaller quality setting if it has one.

**Is 300 DPI too much for a scan?**

For email, upload portals and reading on screen, usually yes: 150 to 200 DPI is enough for ordinary documents. Our 300 DPI scan was 2.6 times the size of the 150 DPI one. Use 300 DPI for print or very small type, and follow an archive’s own rules for archival scans.

**Can I compress a PDF twice?**

Yes, but the second pass usually gains little and costs quality. JPEG re-encoding is lossy, so each pass discards more detail. Our Word Minimum size export, already compressed, shrank only 6.5% at the strongest setting. If the first result is too big, start again from the original with a stronger setting rather than compressing the compressed copy.

**Why did compressing make my PDF bigger?**

Usually because the tool rasterised the pages: our 215 KB Word report became 1.4 MB when every page was redrawn as an image. It can also happen when already-small images are re-encoded. A careful tool keeps the original image when the new one is not smaller and tells you when it cannot shrink the file, so keep your original.

**How do I check what is inside my PDF?**

Try selecting text: if you cannot, the page is an image. Divide the size by the pages; a few KB a page is text, megabytes a page is images or scans. Zoom to 400%: blurry text means a scan. A compressor that reports what it changed, as ours does, also shows how much of the file is images and how much is text, fonts and drawings.

**How much can a PDF be compressed?**

It depends entirely on what is inside. In our tests, lossless clean-up saved 0% on scans and 93% on a file with a repeated logo; image recompression took a 600 DPI scan down 90%; a 200-page text PDF could not go below about 192 KB, because text, fonts and drawings do not shrink. Images are where the savings are.

**Does compressing a PDF lose quality?**

Lossless clean-up does not: nothing visible changes. Image recompression does lose some detail in photos and scans, which at 200 DPI and quality 75 is hard to see at normal zoom. Text is untouched: in our tests the extracted text was identical before and after. Only rasterising turns text into pixels, and that loss is permanent.

**How do I reduce PDF size on a Mac?**

In Preview, choose File > Export, pick Reduce File Size from the Quartz Filter menu and save as a new file. Apple warns the result may be lower in quality. We did not test Preview, so compare it with the original at 100% and 200% zoom, or use a compressor that lets you choose the DPI and quality.

---

_Source: [Do The Calculation](https://dothecalculation.com/blog/files/how-to-reduce-pdf-file-size). Quote freely with attribution and a link to this page._