Choosing a page size for scans and screenshots

Choose a real paper size when the document will be printed, and fit the page to the image when it will only ever be read on a screen. A 1920 × 1080 screenshot placed on A4 portrait occupies a band across the middle and leaves large white margins above and below; the same image on a page sized to itself fills the view at any zoom level.

The decision is about the destination rather than the source. Everything else — margins, orientation, order — follows from it.

What are the page sizes actually?

Two families that do not interoperate. The ISO A series halves cleanly and the US sizes do not.

Size Millimetres Inches
A3 297 × 420 11.69 × 16.54
A4 210 × 297 8.27 × 11.69
A5 148 × 210 5.83 × 8.27
Letter 216 × 279 8.5 × 11
Legal 216 × 356 8.5 × 14

A4 is 6 mm narrower and 18 mm taller than Letter. That is enough that a document laid out for one prints with shifted margins on the other, and enough that a full-bleed design gets cropped.

The A series has one property the US sizes lack: each size is exactly half the one above, because the ratio of the sides is the square root of two. Two A5 sheets make an A4, and that is why scaling between them never distorts anything.

When is fit-to-image the right answer?

Whenever the output is a screen. It sizes each page to the image plus the margin, so there is no wasted field and no scrolling past white space.

  • Screenshots. Almost always fit — they are already the shape of a screen.
  • Receipts and phone photos of documents. Fit, because the shapes vary and forcing them onto A4 leaves each one differently placed.
  • Anything for printing. A real size, because a printer needs a page it recognises.
  • Mixed-shape sets. Fit, unless they will be printed, in which case pick one size and accept the margins.

What margin should you use?

For screen reading, a small one — enough to keep the image off the edge without adding a visible frame. For printing, at least 10 mm, because most consumer printers cannot print to the edge and will crop whatever falls in the unprintable border.

The margin also decides how large the image lands. On A4 portrait with a 24-point margin, the usable area is about 547 × 794 points, so a 1920 × 1080 screenshot scales to the full width and about 308 points tall — a third of the page.

Rotating that page to landscape gives a usable area of about 794 × 547, and the same screenshot fills far more of it. Where every image in a set is wider than it is tall, landscape is the better default and almost nobody changes it.

Does the PDF re-compress the images?

Not the JPEGs. A JPEG is embedded exactly as it arrived, with no decode and no re-encode, so the pages carry precisely the quality of the files you supplied and no more loss is introduced.

PNG and WebP are written as lossless pixel data, which preserves everything including transparency — stored as a soft mask so it renders correctly rather than turning black. The trade is size: a large PNG stays large inside the PDF.

The practical consequence is that a PDF built from photos is roughly the size of the photos. If the result needs to be smaller, resizing the images before building it works far better than compressing the PDF afterwards.

Order is the other thing worth setting before building rather than after. Pages come out in the order the images are listed, and reordering a sixty-page document afterwards means rebuilding it — so a quick pass through the list at the start is cheaper than it looks.

What resolution does a scan need?

Three hundred dots per inch for anything that will be printed or read closely, which on A4 is 2,480 × 3,508 pixels. At 150 dpi the same page is 1,240 × 1,754 — perfectly readable on screen and visibly soft in print.

Going above 300 dpi for text documents rarely helps and multiplies the file size: 600 dpi is four times the data for a difference that matters only for fine line art and archival work.

What about the pages after the first?

Every page is sized independently in fit mode and identically in a fixed-size mode, and that difference shows up when a document is printed. A stack of differently sized pages prints centred on whatever paper is loaded, which is fine for reference and looks wrong bound.

For anything that will be stapled, hole-punched or filed, pick one real size for the whole document even where it wastes space. Consistency is the property a physical document needs and a screen document does not.

Questions people ask

Can I mix portrait and landscape? Yes. Each page is built independently, so a set with both orientations produces a document with both.

How many images can I add? Up to sixty in one document, reordered before building.

Is anything uploaded? No. The PDF is assembled in the browser, and neither the images nor the result leave the device.

Why is my PDF larger than the images? It should not be by much. If it is, the images are probably PNGs of photographs — converting those to JPEG first will shrink the result substantially.

Pick the destination first and the page size answers itself. PNG to PDF covers both families of paper size and the fit-to-image option, the image resizer is the right place to shrink things beforehand, and Compress PDF is for the documents you did not build yourself.