Where does a PDF’s file size actually come from?

Overwhelmingly from embedded raster images, not from text. People assume a long document is large because of all the words, but text and vector graphics are extraordinarily compact — a hundred-page text report can be a fraction of a megabyte. The weight comes from photographs, scans, screenshots, and high-resolution graphics embedded in the pages, often saved at print resolution when the document will only ever be viewed on screen. This is the single most useful thing to understand about PDF size: if your file is large, look at its images first, because that is where almost all the bytes live and therefore where almost all the savings are. Compressing the images is compressing the PDF; everything else is rounding error.

What is the difference between lossy and lossless compression here?

Lossless compression makes a file smaller while keeping the data bit-for-bit identical — you can reverse it perfectly, so there is genuinely no quality loss, but the savings are modest. Lossy compression achieves much bigger reductions by permanently discarding information the eye is unlikely to miss (the way JPEG works on photos), which means real quality loss if pushed too far. For PDFs, the practical implication is that you have two levers: lossless steps (deduplicating objects, subsetting fonts, stripping unused data, applying lossless image filters) cost you nothing and should always be taken, while lossy image re-encoding is where you trade quality for size and must be applied with judgement. "Without losing quality" in the strict sense means lossless only; in the everyday sense it means lossy compression kept gentle enough that the loss is invisible at normal viewing.

So can I really make a PDF smaller with no quality loss at all?

Sometimes fully, often partly. If your file is bloated by redundant objects, un-subsetted fonts, oversized metadata, or images stored with an inefficient lossless filter, you can recover real size with zero quality loss — that is pure waste. But if the file is large because it genuinely contains high-resolution photos, truly lossless methods can only do so much, and the big reductions require lossy image compression, which is a trade-off rather than a free lunch. The honest framing is: take every lossless saving first (it is free), and then decide how much lossy image compression you are willing to accept. For on-screen use you can usually compress images substantially before any loss is visible; for print you keep more resolution.

How much can I compress images before quality visibly suffers?

It depends on how the PDF will be viewed, because quality only matters relative to the display. The key insight is that screens need far less resolution than print: an image displayed on a screen needs roughly 100–150 pixels per inch to look crisp, while print wants around 300, so a PDF destined only for screen viewing can have its images downsampled toward screen resolution with no visible loss at all — you are throwing away detail that was never going to be seen. The mistake is keeping print-resolution images in a document that will only be emailed and read on a laptop. Match the image resolution to the viewing medium: compress hard for screen-only files, keep resolution high for anything that will be printed, and always preview the densest images before finalising.

What should I never do in the name of compression?

Never rasterise the whole page — that is, never "flatten" a text-and-vector PDF into a page-sized image to shrink it. It is the most common self-inflicted quality disaster: it converts your crisp, selectable, infinitely-scalable text into a fixed-resolution picture of text, which looks soft, cannot be searched or copied, breaks screen readers, and frequently is not even smaller. Genuine compression reduces the embedded images and leaves text and vectors as text and vectors. If a tool offers to "compress" by printing to image or exporting as a flattened scan, decline it. The whole art of keeping a PDF small without losing quality is to compress only the part where bytes live (images) and never touch the part that is already efficient and lossless (text and vectors).

Is it safe to compress a confidential PDF online?

Only with a tool that runs on your own device. Compression is often applied to invoices, statements, and scanned personal documents, and many online compressors upload your file to a third-party server to process it. Client-side (in-browser) tools compress locally so the file never leaves your computer — ScoutMyTool’s PDF tools work this way. For confidential material, confirm the tool is client-side before uploading, or use offline software. A smaller file is not worth handing a sensitive document to a server you do not control.

7 min read

How to keep a PDF small without losing quality

By ScoutMyTool Editorial Team · Last updated: 2026-05-21

For years I treated PDF compression as a slot machine — drag the file into some tool, pull the lever, and hope it came out smaller without looking terrible. Sometimes it did; sometimes it turned my crisp report into a fuzzy mess. What finally made it predictable was understanding the mechanism: where a PDF’s bytes actually live, and what "lossy" versus "lossless" really means. Once you know that almost all the weight is in embedded images, and that some savings are free while others trade quality for size, you stop guessing. This guide is not another list of buttons to press — it is the understanding underneath, so you can compress hard where it costs nothing and gently where it costs quality, and never make the one mistake that ruins a PDF.

Where the bytes live

Component	Share of size	How compressible	Quality cost
Embedded images / scans	Usually most of the file	A lot	Where quality loss happens — compress carefully
Text & vector graphics	Tiny	Already efficient	None — never rasterise it to "compress"
Embedded fonts	Small–moderate	Subset unused glyphs	None if subset correctly
Duplicate / redundant objects	Varies	Deduplicate	None — pure waste removal
Metadata & thumbnails	Small	Strip if not needed	None to the visible content

Step by step — shrink it the right way

Take every lossless saving first. Deduplicate redundant objects, subset embedded fonts to the glyphs actually used, and strip unneeded metadata and thumbnails — all free, no quality cost.
Find the heavy images. Identify the embedded photos and scans driving the size; that is where the real reduction is, not in the text.
Match image resolution to the viewing medium. For screen-only files, downsample images toward screen resolution (~100–150 PPI); keep ~300 PPI for anything that will be printed.
Apply lossy image compression with judgement. Re-encode photos at a sensible quality level and check the densest images — push only until just before loss becomes visible at normal viewing.
Never rasterise text and vectors. Do not flatten the page to an image; keep text selectable and vectors sharp. Compress images, not the whole page.
Keep an uncompressed master. Save the compressed copy separately so you always have the full-quality original to re-derive from.

The principle in one line

Keeping a PDF small without losing quality comes down to one rule: compress where the bytes are (images), the way that fits the destination (screen or print), and never touch what is already efficient (text and vectors). The free, lossless savings — dedup, font subsetting, stripping waste — should always be taken because they cost nothing. The lossy savings — image re-encoding and downsampling — are a deliberate trade you tune to how the document will be seen, and for screen-only files that trade is so favourable it is effectively free. The only true mistake is the brute-force one: rasterising the whole page, which destroys your text to save bytes you could have saved cleanly. Understand the mechanism and the slot machine becomes a dial you control.

FAQ

Where does a PDF’s file size actually come from?: Overwhelmingly from embedded raster images, not from text. People assume a long document is large because of all the words, but text and vector graphics are extraordinarily compact — a hundred-page text report can be a fraction of a megabyte. The weight comes from photographs, scans, screenshots, and high-resolution graphics embedded in the pages, often saved at print resolution when the document will only ever be viewed on screen. This is the single most useful thing to understand about PDF size: if your file is large, look at its images first, because that is where almost all the bytes live and therefore where almost all the savings are. Compressing the images is compressing the PDF; everything else is rounding error.
What is the difference between lossy and lossless compression here?: Lossless compression makes a file smaller while keeping the data bit-for-bit identical — you can reverse it perfectly, so there is genuinely no quality loss, but the savings are modest. Lossy compression achieves much bigger reductions by permanently discarding information the eye is unlikely to miss (the way JPEG works on photos), which means real quality loss if pushed too far. For PDFs, the practical implication is that you have two levers: lossless steps (deduplicating objects, subsetting fonts, stripping unused data, applying lossless image filters) cost you nothing and should always be taken, while lossy image re-encoding is where you trade quality for size and must be applied with judgement. "Without losing quality" in the strict sense means lossless only; in the everyday sense it means lossy compression kept gentle enough that the loss is invisible at normal viewing.
So can I really make a PDF smaller with no quality loss at all?: Sometimes fully, often partly. If your file is bloated by redundant objects, un-subsetted fonts, oversized metadata, or images stored with an inefficient lossless filter, you can recover real size with zero quality loss — that is pure waste. But if the file is large because it genuinely contains high-resolution photos, truly lossless methods can only do so much, and the big reductions require lossy image compression, which is a trade-off rather than a free lunch. The honest framing is: take every lossless saving first (it is free), and then decide how much lossy image compression you are willing to accept. For on-screen use you can usually compress images substantially before any loss is visible; for print you keep more resolution.
How much can I compress images before quality visibly suffers?: It depends on how the PDF will be viewed, because quality only matters relative to the display. The key insight is that screens need far less resolution than print: an image displayed on a screen needs roughly 100–150 pixels per inch to look crisp, while print wants around 300, so a PDF destined only for screen viewing can have its images downsampled toward screen resolution with no visible loss at all — you are throwing away detail that was never going to be seen. The mistake is keeping print-resolution images in a document that will only be emailed and read on a laptop. Match the image resolution to the viewing medium: compress hard for screen-only files, keep resolution high for anything that will be printed, and always preview the densest images before finalising.
What should I never do in the name of compression?: Never rasterise the whole page — that is, never "flatten" a text-and-vector PDF into a page-sized image to shrink it. It is the most common self-inflicted quality disaster: it converts your crisp, selectable, infinitely-scalable text into a fixed-resolution picture of text, which looks soft, cannot be searched or copied, breaks screen readers, and frequently is not even smaller. Genuine compression reduces the embedded images and leaves text and vectors as text and vectors. If a tool offers to "compress" by printing to image or exporting as a flattened scan, decline it. The whole art of keeping a PDF small without losing quality is to compress only the part where bytes live (images) and never touch the part that is already efficient and lossless (text and vectors).
Is it safe to compress a confidential PDF online?: Only with a tool that runs on your own device. Compression is often applied to invoices, statements, and scanned personal documents, and many online compressors upload your file to a third-party server to process it. Client-side (in-browser) tools compress locally so the file never leaves your computer — ScoutMyTool’s PDF tools work this way. For confidential material, confirm the tool is client-side before uploading, or use offline software. A smaller file is not worth handing a sensitive document to a server you do not control.

Citations

Compress the right way — in your browser

ScoutMyTool Compress-PDF reduces the embedded images and leaves your text and vectors crisp, client-side, so the file never leaves your computer — the mechanism above, applied for you.

Open Compress-PDF tool →