Scans with inch-thick white borders, slides exported onto Letter paper, spreadsheets printed at 60% size — these are all the same problem: the page is bigger than the content on it. PDFs have a dedicated mechanism for exactly this, and using it properly beats screenshotting or re-exporting every time. Here's how cropping actually works inside the file, when to auto-trim versus set fixed margins, and how to do it without uploading the document anywhere.
Every PDF page carries a set of boundary boxes in its dictionary. The MediaBox is the physical page; the CropBox is the region viewers agree to display, print, and export. When a CropBox is present, it wins — which means cropping is the act of writing a smaller rectangle there. No pixels get painted, no text gets touched: you are updating geometry, and the viewer does the rest.
That geometry-first design has two consequences worth knowing. Cropping is reversible, because the page content outside the box is still in the file. And cropping is not secrecy, for exactly the same reason — anyone who lifts the CropBox sees what was outside it. Hiding is not deleting.
The pattern behind all four: content smaller than the page. Cropping closes the gap.
| Mode | How it works | Best for |
|---|---|---|
| Auto-trim (per page) | Each page is rendered and scanned for its content bounds; the box hugs what the detector finds, plus a small padding | Scans, mixed documents, receipts, inconsistent borders |
| Auto-trim (page 1 for all) | One detection, applied uniformly | Slide decks and uniform documents, faster on big files |
| Fixed margins | One inset cut on every page — e.g. 0.5 in off each side | Print prep, binding allowances, uniform trimming |
The tell is consistency. If every page has the same problem, fixed margins are predictable and instant. If borders vary — page 3 has a receipt, page 9 has a full-bleed photo — per-page auto-trim is the only thing that comes out right.
Take a Letter scan, 612 × 792 points (PDF units are points, and 72 points make an inch). Rendered at 2× for detection, the scan measures 1224 × 1584 pixels, and the ink turns out to run from pixel (150, 310) to (1049, 1275). Mapping those pixel bounds back to PDF space — x = 150 ÷ 2 = 75, y = 792 − 1276 ÷ 2 = 154, and so on — gives a crop box of [75, 154, 525, 637]: a 450 × 483-point page, 6.25 × 6.71 inches.
Run the arithmetic on area and the border was 55.2% of the page. More than half the scan was white paper. With a 12-point padding (a sixth of an inch of breathing room), the box relaxes to 474 × 507 points and the cut is 50.4% — the content keeps a slim frame that reads as intentional rather than accidental.
For contrast, a clean office PDF with uniform one-inch margins crops to 468 × 648 points (6.5 × 9 in), removing 37.4% of the page area. Same operation, very different documents, same one-line change per page.
Neither, and it's worth being clear-eyed about that. The content outside the crop is still in the file, so the byte count barely moves — if you need a smaller file, that's compression territory. Text inside the box stays exactly as it was: selectable, searchable, and untouched, because cropping never re-renders content, it only re-frames it. That's also why cropped PDFs don't degrade — there is no re-encoding step to degrade.
Yes — detection is just pixel scanning and writing a box, both of which JavaScript does well. Our Crop PDF tool renders each page with pdf.js, finds the ink bounds, and writes the CropBox with pdf-lib, all in your browser. There's no account, no watermark, and the original file on disk is never modified — the download is a new copy with -cropped appended. Pair it with Organize PDF when the border problem comes with a page-order problem, and with PDF to JPG when the cropped output needs to become images.
Auto-trim white borders or cut fixed margins on every page, preview before you download. Free, private, in-browser.
Crop PDF →Cropping is geometry, not surgery: set the CropBox, keep the MediaBox, and the file both looks right and stays reversible. Auto-trim for raggedy scans, fixed margins for print prep, and never confuse hiding for redaction — sensitive text needs a tool built to destroy it.