Page sizes when combining PDFs — A4, Letter, mixed

One of the under-discussed facts about PDF combining is that the result does not need a single page size. PDF was designed page-by-page from the start: every Page object carries its own MediaBox, an array of four numbers [x0 y0 x1 y1] in points (1 pt = 1/72 inch). A combined PDF can mix A4, Letter, A5, A3, custom sizes, and rotated pages without any normalization step.

The MediaBox in detail

A page's MediaBox defines the physical sheet. Three more boxes refine where content sits:

• CropBox — the visible area in viewers; defaults to MediaBox.
• BleedBox — for print, includes ink bleed area.
• TrimBox — the final trimmed page in print production.

Combiners copy all four when they exist. If you merge a print-ready PDF (with TrimBox set to A4 inside a slightly larger MediaBox to allow bleed) with a regular A4 PDF, the result correctly preserves the print metadata for the first set of pages and uses MediaBox-equals-CropBox for the second.

A4 vs Letter: the perpetual question

A4 is 595 × 842 points (210 × 297 mm). US Letter is 612 × 792 points (8.5 × 11 inches). They differ by 17 points wide and 50 points tall — about a quarter inch on each axis. Combining the two without normalization produces a PDF whose pages alternate slightly in size; viewers handle this fine, but the result prints awkwardly because most printers default to a single tray.

A surprising number of "looks like A4" PDFs are actually Letter, especially anything originating from US-headquartered SaaS tools. Check before combining if uniformity matters.

Rotation

A page can declare /Rotate 0|90|180|270 independently of its MediaBox. A landscape page can be stored as a portrait MediaBox with /Rotate 90, or as a landscape MediaBox with /Rotate 0. Both render identically; both are legal. When combining, the rotation flag is preserved per-page.

This is why occasionally a PDF combiner produces output where one page seems to be on its side: the input had a rotation flag the combiner copied verbatim, but you expected it to "match" the surrounding orientation.

Should you normalize?

The merge does not normalize page sizes. Every input page comes through with its original MediaBox, CropBox, and rotation. Normalization is destructive — you cannot losslessly stretch A4 content to Letter or vice versa — and the merge tool keeps each input's geometry intact rather than guessing.

If you need a uniform output, the right path is to print all sources to PDF at the target paper size first, then combine. Re-rasterizing or re-laying-out at the target size is a separate operation; the combine itself preserves whatever each input declared.

Image pages from image inputs

When you add JPGs, PNGs, or other images to a combined PDF, each image becomes a single page sized to A4 (595 × 842 pt) by default at 96 DPI. The image is fitted into the page rectangle preserving aspect ratio. A portrait phone photo lands centered on a portrait A4 page with white margins on the sides; a landscape photo gets margins top and bottom, or — if the combiner is set to auto-rotate — sits on a landscape A4 page (842 × 595 pt).

Custom sizes and page labels

PDF allows page numbering to be non-sequential via the /PageLabels tree: you can label the first three pages "i, ii, iii" and the rest "1, 2, 3". PageLabels live at the catalog level and don't survive a typical merge — reconciling two non-overlapping label schemes from two inputs has no single right answer, so the result usually falls back to plain 1–N numbering. If consistent labels matter, set them up in the merged PDF afterwards using a tool that exposes PageLabel editing.