Crop a PDF, or Let It Trim the Scanner Edges for You

How it works

1. Choose the PDF that needs cutting down. It is opened here on your own device and never uploaded, which is the only reason a drawing with a subcontractor's pricing on it can be handled this way at all.
2. Drag a box around the part worth keeping and nudge it by its corners, or press Find the edges and let the white margin be measured for you, page by page. The size the page will end up is shown in inches while you drag.
3. Say whether it applies to every page, to the one on screen, or to the pages you list. Then crop and download. Nothing else in the file is touched.

What you get

• One box drawn on one page, applied to every page in the same place
• Drag the box to move it and its corners to resize it, so a crop gets nudged rather than drawn again
• Finds the white margin by itself, page by page, so a stack of scans that all sit differently still comes out square
• Reads the paper color off the corners first, so an off white or gray scan is trimmed instead of ignored
• Ignores any mark shorter than about five points, so one fleck of scanner dust in a margin does not decide the crop
• A margin control, so the cut stops short of the ink instead of shaving it
• The finished page size in inches, live, while you drag
• Sets both of the boxes a PDF keeps for this, so the crop holds in a reader and on a printer
• Pages left alone stay exactly as they were, byte for byte, and so does everything else in the file

Two inches of gray down the side of the drawing

A wide format scanner does not know where your sheet stops. It runs the roller for a fixed width and gives you everything it saw, which on a 24 by 36 sheet fed slightly crooked means a gray band down one edge, a black bar where the lid did not close, and sometimes a finger's width of the next drawing in the stack. The drawing itself is fine. It is just sitting inside something larger and dirtier than it should be.

The same thing happens at every other size. A phone scanning app squares up a receipt and then saves it onto a full letter page, so a 3 inch slip of paper prints at the size of a postage stamp with nine inches of white around it. A page printed to PDF out of a browser carries a header with the site address and a footer with the date and the page count, neither of which belongs in a submittal. A photo report exported from an app puts a band of its own branding across the bottom of every page.

None of that is a problem with the content. It is a problem with the size of the sheet the content was put on, and the fix is to declare a smaller sheet. That is all a crop is. You are not editing the drawing, you are telling every reader and every printer that the page is this rectangle and not that one.

Which is why this is worth doing before anything else happens to the file. A cropped sheet prints bigger on the same paper, reads better on a phone, and stops a reviewer scrolling past half an inch of scanner grime to find the title block.

Draw it once, or let it find the edges

The first way is the one you would expect. The page comes up, you drag a box around what matters, and the part outside the box goes dim so you can see exactly what you are about to lose. Drag inside the box to move it, drag a corner to resize it. The width and height of the finished page are shown in inches under the picture and they change as you drag, so a crop meant to land on 11 by 17 can be landed on 11 by 17 rather than guessed at.

That box is held as a fraction of the page and not as a measurement, which is what lets one box apply to a whole file. Page through the document and the box stays where you put it. On a file where every page is the same size, which is nearly every file, that is exactly right: the header band is in the same place on page forty as it is on page one.

The second way is for the case where it is not. Press Find the edges and each page is looked at on its own: the page is drawn, the paper color is read off its four corners, and the outermost row and column that actually carry ink are found. What comes back is the smallest rectangle that holds the content, plus whatever margin you asked for. Page through and you can see the box it worked out for each one before you commit to any of it.

That is the mode for a stack of scans. Twenty waivers fed through a desk scanner sit slightly differently every time, and one box drawn on the first one clips the second. Measuring each page separately is slower by a second or two and it is the only thing that gets all twenty right.

How the edge finding actually decides, and where it gives up

It does not look for white. It looks for whatever color the corners of your page are, which on a real scan is rarely white and is often a warm gray with a cast from the lid. Four corner patches are sampled, the most common value among them becomes the paper, and anything far enough from that value counts as content. A page scanned onto cream stock is trimmed correctly for the same reason a page scanned onto white is.

Then it ignores the specks. A row or a column has to carry about five points of ink before it counts as content, because a scanner bed picks up dust, a stapled corner leaves a shadow, and a single dark fleck out in the margin would otherwise hold the crop open across the entire sheet and make the whole feature useless. Five points is deliberately a length and not a share of the page, because how much mark means a real mark has nothing to do with how big the sheet is. A speck of dust measures two or three, the stem of a nine point page number measures six or seven, a line of text measures tens, and a rule down the side of a drawing measures hundreds, so the only thing that threshold ever throws away is the dust.

The margin is yours to set: none, an eighth of an inch, a quarter, or a half. A quarter inch is the default and it is the right default, because a crop that stops exactly on the outermost pixel of ink looks like a mistake, and because the outermost pixel of ink is sometimes a speck that got through. Set it to none when you are cropping a photograph, where any margin at all is a white line down the side.

And it gives up honestly. If a page is edge to edge content, a full bleed photograph for instance, there is nothing to trim and that page is left exactly as it was rather than cut by an arbitrary amount. Same if a page is blank: a blank page has no content box, so it keeps the size it had.

A crop hides. It does not delete

This is the one thing on this page worth reading twice. Cropping a PDF changes which part of the page a reader is told to show. Everything outside the new box is still in the file, in full, and it can be brought back by anybody who opens that file in an editor and widens the box again. It takes about four seconds and no special software.

So a crop is the right tool for scanner grime, for white space, for a browser's print header, for trimming a detail out of a big sheet to send to one sub. It is the wrong tool, completely, for anything that actually has to go. If the band at the bottom of the page is another contractor's pricing, or an account number, or somebody's full social security number on a W-9, cropping it out puts you in the worst position available: it looks handled and it is not, and you will send it believing the opposite.

For those, black it out instead. That tool draws over the area and then replaces the page with a flattened picture of itself, so the words underneath are gone from the file rather than merely outside the window. It costs you selectable text on the pages it touches, which is exactly the trade you want when the alternative is sending the thing you meant to remove.

Nothing here will ever describe a crop as removing anything, and the page says so at the point where it matters rather than in a footnote. It is the same reason this site will not flatten a document to make it small and call that compression.

It will not make the file any smaller either

The second thing people expect from a crop is a smaller file, and it follows from the first: nothing was thrown away, so nothing was saved. Crop nine inches of white off a 40 MB photo report and you still have a 40 MB photo report, now with a smaller window onto it. Occasionally the file gets a few hundred bytes larger, because a page that had no crop box before now has one.

If size is the actual problem, make it smaller first and crop afterwards. That tool re-encodes the pictures inside the file and leaves the text as text, which is where the megabytes in a scan or a photo report really are. The order does not matter much, but doing the crop second means you are looking at the final page size while you choose the box.

What a crop does change is the paper. A cropped page fills more of the sheet when it prints, which on a drawing being read at arm's length in a truck is the whole point, and it fills more of the screen on a phone, which is where most of this paperwork is now actually read.

Why both of the boxes get set

A PDF page carries more than one rectangle. The MediaBox is the sheet the page was made on. The CropBox is the part of that sheet a reader is meant to display, and when a file has no CropBox, readers fall back to the MediaBox. Most crop tools set only the CropBox, which is correct by the specification and is also how a crop comes undone on somebody else's desk.

The reason is printing. Plenty of print drivers, and a fair number of older readers, lay the page out from the MediaBox and ignore the CropBox entirely, so a sheet that looked cropped on screen comes out of the plotter with all the gray still on it. Some page level tools do the same thing when they rebuild a document. Both boxes are set here, to the same rectangle, so there is no second opinion left in the file about what the page is.

The other boxes a page may carry, the bleed, trim and art boxes, are prepress instructions from a print shop. If your file has them and they sit outside the new crop, they are pulled in to match rather than left pointing at paper that is no longer there.

All of this happens without rebuilding the page. The content, the fonts, any form fields, bookmarks and links are the ones that were in the file when you opened it, untouched. A crop here writes two numbers pairs per page and changes nothing else, which is also why it is instant on a four hundred page permit set.

Where this will not help

It will not crop different parts of different pages in one pass by hand. One drawn box goes on every page you pick, and the automatic mode measures each page on its own. If page three genuinely needs its own hand drawn box, run the tool twice: split the file, crop each part, and put it back together. That is three minutes and it is honest about what it is doing.

It will not straighten a crooked scan. If the sheet went through the scanner at an angle, the content is at an angle, and cutting a square box around an angled document leaves the angle in place. Rescan it, or accept a slightly larger crop that holds the whole tilted page. If the page is a whole quarter turn out, that is a different and easier problem: turn it the right way up first, then crop.

It will not change what paper the page prints on. A cropped 24 by 36 is a smaller sheet, not a letter sheet, and sending it to an office printer still means scaling. To land a drawing on paper you actually have, change the paper size, which redraws the page onto the sheet you name and keeps the scale honest while it does it.

And it cannot open a password protected file. No browser can. Open it in a reader, save an unprotected copy, and start from that.

What to do around it

Crop late. A crop is a decision about how the finished document is presented, so it belongs after the document is finished: after the pages are in order, after the photographs are in, after anything that had to be blacked out is gone. Cropping first and then merging means a packet whose pages are different sizes, which prints badly on every printer there is.

Check what is underneath before you send it. This is the crop specific version of a rule that applies to any paperwork leaving your hands, and it matters more here than usual precisely because a crop looks like a removal. If the strip you just put outside the window has anything on it that should not travel, black it out before the file leaves.

Number the pages afterwards rather than before. Page numbers are placed relative to the edge of the page, so numbering first and cropping second either cuts the numbers off or leaves them floating in the wrong place. Do it in that order and every number lands the same distance from the new edge.

Then put it in the packet and send it. If it is still too heavy for their mail server, shrink it, which is the part of this that actually removes weight.

Questions people ask

Related