Ask for a PDF page as a JPG and you are asked a question back: what resolution? It has to be asked because a PDF page, unlike a photo, has no pixels to begin with. Its text is a list of characters in a font, its drawings are lines and curves, and only scanned pages carry a grid of pixels at all. Turning the page into an image means drawing all of that onto a grid of a size you choose, and that choice decides whether the text is crisp or blurred, whether the file is 200 KB or 20 MB, and whether the image lands in a Word document at the right size or three times too large.
A PDF Page Is Measured in Inches
Everything on a PDF page is placed in what ISO 32000-2 calls user space, and section 8.3.2.3 of the standard sets its unit at 1/72 of an inch, the typographer's point. The page's size is stored the same way. An A4 page is a box 595.28 points wide and 841.89 tall, which is 210 by 297 millimetres, and US Letter is 612 by 792, or 8.5 by 11 inches. Nothing in that description says how many pixels the page should have, because a PDF is meant to come out the same size on any screen or printer.
So a renderer works it out from the resolution you give it, in dots per inch. The width in pixels is the width in inches times the DPI: an A4 page is 8.27 inches wide, so at 150 DPI it is 1240 pixels across and at 300 DPI it is 2480. Text and vector drawings are redrawn from their outlines at whatever size is asked for, so they gain detail all the way up. A scanned page is different. It is already a picture with a fixed number of pixels, so rendering it above the DPI it was scanned at makes a larger file without adding any detail.
How Many Pixels a Page Becomes
Pixel sizes for common paper at each resolution, with the memory one page takes while it is being drawn (4 bytes per pixel, before compression).
| Page | 72 DPI | 150 DPI | 300 DPI | 600 DPI |
|---|---|---|---|---|
| A4 | 595 × 842 | 1240 × 1754 | 2480 × 3508 | 4961 × 7016 |
| US Letter | 612 × 792 | 1275 × 1650 | 2550 × 3300 | 5100 × 6600 |
| A3 | 842 × 1191 | 1754 × 2480 | 3508 × 4961 | 7016 × 9921 |
| Megapixels (A4) | 0.5 | 2.2 | 8.7 | 34.8 |
| Memory while drawing (A4) | 2 MB | 8 MB | 33 MB | 133 MB |
What Each Resolution Is For
Doubling the DPI quadruples the pixels, and roughly quadruples the file. So the useful question is the lowest resolution that does the job.
72 DPI: one pixel per point
The page gets exactly one pixel for each unit of its own coordinate system, so an A4 page is about 600 pixels wide. That is enough for a thumbnail or a preview in a list, and too little to read small print once you zoom in.
150 DPI: reading on a screen
An A4 page at 1240 pixels wide fills a laptop screen at full width with sharp body text, and a whole page stays around a few hundred kilobytes as a JPG. It is the sensible default for sharing pages by email or chat, or putting one on a web page.
300 DPI: printing and text recognition
Print shops ask for 300 pixels per inch at the final print size, and OCR engines are tuned for the same density. The Tesseract documentation states that it works best on images of at least 300 DPI, since below that the strokes of small letters start to merge. If the image will be printed or fed to OCR, use 300.
600 DPI: fine detail and archiving
Worth it for engineering drawings, maps, sheet music and footnotes in 6-point type, where thin lines must stay separate. For ordinary documents it mostly makes files four times larger than 300 DPI without a visible difference on paper.
JPG or PNG for a Page of Text
JPEG, standardised as ITU-T T.81, compresses an image in blocks of 8 by 8 pixels. Each block is turned into a set of frequencies, and the encoder keeps the smooth, low frequencies and coarsens the sharp, high ones, which is where most of the saving comes from. Photographs are almost all smooth gradients, so they survive this well. The edge of a black letter on white paper is the sharpest thing an image can contain, and coarsening it leaves faint grey speckles around each character, the artefact usually called ringing or mosquito noise. Most encoders also store colour at half the resolution of brightness, a scheme known as 4:2:0 chroma subsampling, so red or blue text picks up a soft coloured fringe.
PNG, defined in ISO/IEC 15948, is lossless: every pixel comes back exactly. It predicts each pixel from its neighbours to the left and above and compresses the difference with deflate, and a page that is mostly white paper with black text compresses well that way. A scanned page or one with photos is the opposite case. Paper texture and sensor noise give the predictor nothing to work with, and the PNG can be five or ten times the size of a good JPG.
The practical rule follows from that. Scans, photos and anything that will be emailed go to JPG at a quality of 85 to 90, where the artefacts are hard to see at normal zoom. Text, charts, diagrams and screenshots that someone will zoom into go to PNG.
The DPI number stored inside the image
An image file can record the resolution it was made for. In a JPG that is the density field of the JFIF header, with a unit byte that is 1 for dots per inch. In a PNG it is the pHYs chunk, which counts pixels per metre, so 300 DPI is stored as 11811. Word, LibreOffice and print dialogs use that value to decide how large to place the picture. Browsers' built-in image encoders, which most online converters rely on, leave it empty or write a 1:1 ratio with no unit, and software then assumes 96 or 72 DPI. A 300 DPI render of an A4 page therefore arrives in Word as a sheet about three times the size of A4. Writing the real DPI into the file is what makes the page come back at its original size.
Save PDF pages as images
Pick JPG or PNG and a resolution, see the pixel size of each page, and get every image in one ZIP.
Why a Browser Sometimes Draws Less Than You Asked For
A converter that works in the browser draws each page onto a canvas element, and browsers put a ceiling on how big a canvas can be. Safari's limit is 16,777,216 pixels, the area of a 4096 by 4096 square, and because every browser on an iPhone or iPad uses Safari's WebKit engine, Chrome and Firefox on those devices share it. An A4 page crosses that line at about 416 DPI and an A3 page at 300 DPI already does. Desktop Chrome and Firefox allow canvases up to 32,767 pixels on a side, so there the practical limit is memory: the table above shows a single A4 page at 600 DPI taking 133 MB before it is compressed.
What happens at the limit depends on the converter. WebKit does not throw an error for an oversized canvas, it hands back a blank one, so a careless tool produces a white image of the right dimensions. A careful one lowers the resolution for that page to the largest size the browser can draw, tells you, and records the resolution the image really has rather than the one you asked for.
What Stays Behind in the PDF
An image is a picture of the page, and several things that make a PDF useful do not survive being photographed. The text stops being text: it cannot be searched, selected or read aloud by a screen reader until OCR turns it back into characters. Links, bookmarks and the document outline disappear, because there is nothing left to click. Vector drawings lose their ability to scale, so zooming in past the chosen DPI shows pixels.
Some things do come across. Text typed into form fields and ticked boxes is drawn as it appears, provided the converter asks the renderer to include form values, which is not the default in every library. Highlights, stamps and other annotations that are visible on the page are drawn too, while pop-up notes are not. Layers that the document hides by default stay hidden. And since a PDF page has no background of its own, a converter has to paint it white first, or text would sit on transparency and vanish against a dark viewer.
Frequently Asked Questions
Related Tools
Keep Reading
What the JPEG Quality Slider Is Really Doing to Your Photo
Quality 80 scales the quantization tables to 40 percent, it does not keep 80 percent of the picture. Where the setting stops being free, and why PNG has no dial.
Why Resized Images Come Out Noisy, and the Fix Nobody Applies
Bilinear sampling reads four pixels no matter how far you shrink, so a 20:1 reduction ignores 396 of every 400. Halving in steps reads all of them.
Photos in a PDF: What Happens to a JPG, a PNG and a HEIC
A JPG can be copied into a PDF byte for byte, a PNG is compressed again and a HEIC is refused. Plus why phone photos land sideways and what sets the page size.
Splitting a PDF: Why the Parts Weigh More Than the Whole
One shared font becomes twenty copies, a contents page can drag the whole document along, and edit history quietly drops out. Where the bytes go in a split.