PDF Header & Footer

About adding headers and footers to a PDF

A PDF has no concept of a running header. Word processors and typesetters generate one while laying the document out, and by the time it becomes a PDF the result is just ink on each page. So adding a header after the fact means drawing text onto every page — which is exactly what this tool does, with six independently addressable slots and placeholders that resolve differently on each page.

The six slots

Left, centre and right, in a header band and a footer band. Each band can be switched off entirely, and each slot is optional — leaving one blank simply draws nothing there. The convention most documents follow is a title or document reference in the header and administrative detail in the footer: a revision number bottom left, a confidentiality notice bottom centre, page numbers bottom right. Every slot is a single line of text anchored at one point; there is no wrapping and no collision detection, so a long string will run into its neighbour. The preview shows the real geometry against your actual page, which is the quickest way to catch that.

Placeholders

• {page} — the page's physical position in the document, counting from one.
• {total} — the document's total page count, unaffected by the range you stamp.
• {date} — today's date, in one of five fixed formats chosen in the settings. It is written out by hand rather than through the browser's locale machinery, so the string is identical on every machine.
• {filename} — the uploaded file's name with the .pdf extension removed.

They can be mixed with ordinary text and repeated, so {filename} — {page}/{total} — {date} works as written. The little buttons beside each field append a token at the end; you can also just type the braces.

Why the CropBox, and not the page size

Every position here is measured from the page's CropBox — the rectangle a reader displays — rather than the MediaBox, which is the sheet the page was composed on. On an ordinary office document they are the same and the distinction is academic. On anything prepared for print they are not: the MediaBox is typically 3 to 6 mm larger on each side to carry bleed and registration marks. Measuring a footer from the MediaBox on such a file puts it below the visible page, where it looks broken on screen and is trimmed away on press. The same reasoning applies to a file that has been through a crop: its CropBox is smaller than its sheet, and the header belongs inside the crop.

Two details of the format make that rectangle less obvious than it sounds. A PDF rectangle is four numbers naming two opposite corners, and the order is not guaranteed — plenty of files store them back to front — so the box is normalised before anything is measured from it. And a CropBox only means anything where it overlaps the MediaBox, so the two are intersected; a CropBox larger than the sheet is clipped back to the sheet rather than taken at face value. Both steps match what a reader does, which is also how the preview on this page is computed, so what you see and what you download agree even on a malformed file. Each page's /Rotate value is undone before the coordinates are written, so landscape and sideways-scanned pages are stamped in the orientation you see rather than the one stored on disk; a rotation that is not a multiple of 90 is invalid and readers ignore it, so it is treated as zero here too.

Fonts, size and colour

Three faces are offered — Helvetica, Times Roman and Courier — each with a bold variant. All six are members of the fourteen standard fonts that every conforming PDF reader is required to provide, which means no font program has to be embedded, so the hundred-odd kilobytes one would cost is never spent and the text renders identically everywhere. What the file does gain is the ink: every stamped page picks up a small content stream carrying one text-drawing instruction per filled slot. That is about two hundred bytes for a page with a single footer and roughly four hundred for a page with all six, so a 60-page report with a page number grows by about 12 KB and a 300-page document with all six slots filled by a little over 100 KB. It scales with pages times slots, not with the size of the document, and nothing existing is re-encoded.

The trade-off for skipping embedding is coverage. These fonts encode WinAnsi, roughly Latin-1 plus common typographic characters, so Western European accents are fine — é, ü, ñ, ç, å — but nothing beyond that is. Emoji, Greek, Cyrillic, Arabic, Hebrew and CJK are out, and so are the Central and Eastern European Latin letters that Latin-1 never covered: Polish ł and ą, Czech ř, Hungarian ő, Romanian ș. Invisible characters count too — a tab pasted in from a spreadsheet cannot be drawn either. The text is checked before any ink is placed, and if something cannot be drawn the run stops with nothing written and the message names each offending character with its code point, describing the invisible ones by name. That check covers what the placeholders expand to as well, so a file whose own name contains one of these letters will be refused while {filename} is in a slot. A line break never reaches the check: a slot is one line, and pasting text that spans several turns each break into a space, with a note saying it happened — the browser's own behaviour for a single-line field is to delete the breaks and run the lines together, which is worse and silent. Size is in points and is limited to 4 to 48; anything outside that, or left empty, is replaced by the nearest allowed value, and the number actually in use is shown and drawn in the preview so the setting and the file never disagree. The margin is in millimetres, 0 to 60, and is measured from the page edge to the ink rather than to the baseline: the header's ascenders stop at the margin and the footer's descenders stop at it, so equal margins leave equal white space top and bottom. The colour picker writes an exact RGB value.

Scope, and what cannot be undone

Stamp the whole document, or type a range like 2-9, 12; the separate skip-the-first-page checkbox handles the common case of a cover sheet. Pages left out are untouched, and the preview says so: turn to one of them and it is drawn bare, with a line underneath naming the reason, so the setting can be checked without downloading anything. A page number past the end of the document is reported rather than quietly ignored, and that applies to both ends of a range — on a three-page file, 1-999 is refused exactly as 4 is, because it usually means the wrong file is loaded. Text that is not a page number at all is reported as its own kind of mistake, since being told that a twelve-page PDF cannot use “abc” explains nothing. Settings, the range included, stay put when you swap files, so check the range still suits the new document. What you cannot do is remove the result later: the text joins each page's content stream and is then indistinguishable from the rest of the page, so keep the original if there is any chance you will want to change the wording. Run the tool twice and you get two overlapping headers, not a replacement. For text that needs to sit somewhere other than the six slots, PDF Editor places a box anywhere on a page; for a diagonal DRAFT across the middle, Watermark PDF is the right tool; and for plain sequential numbering with no other text, Page Numbers PDF does it in one click. Everything runs in this browser tab — nothing about the document is transmitted anywhere.