Convert PDF to PNG

A page is not an image until something draws it

This is the fact that makes resolution a decision rather than a detail. A PDF page is a set of instructions — draw this line from here to here, set this text in this typeface at this size, place this photograph in this rectangle. Nothing in the file says how many pixels wide the page is, because the page has no pixels. It has a physical size, measured in points, and there are exactly 72 points to an inch.

A US Letter page is 612 by 792 points, which is 8.5 by 11 inches. Ask for it at 72 DPI and you get 612 by 792 pixels — one pixel per point, roughly the size it would appear on an old screen. Ask for 150 and you get 1275 by 1650. Ask for 300, which is print resolution, and the same page comes out 2550 by 3300 pixels. Same document, same page, three images that differ by a factor of twenty in area. That number is shown on screen for every page before anything runs, so the choice is made against the pixel size rather than against a label like "high quality".

Why convert PDF to PNG rather than to JPEG

Most pages are text and lines on a flat background, and that is precisely the content JPEG handles worst. JPEG compresses in blocks and discards the high-frequency detail first — and a hard black edge against white paper is nothing but high-frequency detail. The result is ringing: faint gray halos around every letter, smeared serifs, a soft haze along table rules. It is subtle at a glance and obvious the moment anyone zooms in or prints the thing.

PNG is lossless. What the renderer drew is what the file contains, edge for edge, and its compression is built for exactly this kind of image — large runs of identical pixels, sharp transitions, a limited palette. The file is bigger than the equivalent JPEG would be, sometimes several times bigger, and for a page of text that is the right trade.

The honest exception is a scanned photograph. A page that is one continuous-tone image with no crisp edges is the case JPEG was designed for, and PNG will produce a large file with no visible benefit. If that is what your PDF holds, convert it here and then compress the result to a target size.

Pages in one document need not be the same size

Nothing in the PDF format requires it, and plenty of real documents mix sizes without anyone noticing: a Letter report with an A3 foldout in the middle, a contract with a scanned receipt appended, a deck exported at 16:9 with one portrait appendix. Every page carries its own dimensions.

That matters because "all pages at 300 DPI" means all pages at the same physical density, not all pages at the same pixel size: the A3 sheet comes out more than twice as wide as the Letter ones. Correct, and still surprising, so each page is measured separately and mixed sizes are stated outright before anything runs. There is a ceiling too. Browsers cap how large a canvas can be, and an oversized page at 600 DPI can exceed it, so the limit is checked per page in advance and the highest resolution that page can manage is named — rather than letting the tab die quietly halfway through.

Transparency, and why white is the default

A PDF page is transparent wherever nothing was drawn. There is no white sheet underneath; the white you see in a PDF reader is the reader painting a page behind the content. PNG carries an alpha channel, so that transparency can survive the conversion intact, and sometimes that is exactly what you want — a page composited onto a colored slide, a logo lifted off a letterhead.

It is not the common case. Most people mean a picture of the page, and a transparent PNG dropped into a dark document turns black text on nothing into an unreadable smear. So white is the default, transparent is one click away, and any other color is available for the times you know what sits behind it.

What this tool will not do

It does not do OCR, and the output has no text layer. A PNG is a grid of pixels; there is nowhere in the format to put words, so nothing in the images will be selectable, searchable or copyable. That is not a limitation of this implementation — no conversion to an image can preserve selectable text, because the destination format has no concept of it. If you want the words rather than the picture, you want text extraction, which reads the text objects already inside the PDF. Different job, different output, and not what this page does.

It also does not edit anything, does not reorder or delete pages in the PDF itself, and does not merge documents. And it only runs in one direction. If you arrived here with images that need to become a document, you want PNG to PDF instead, which is the same job in reverse and lets you set the page size and margins before it builds the file.

Locked files, and where the password goes

An encrypted PDF is not treated as a failure. The file stays where it is and a password field appears in place, because being asked for a password is a question, not an error. What you type is handed to the renderer running inside this tab and goes nowhere else — the same reason the file itself goes nowhere. Documents locked only with an owner password, the kind that restricts printing or copying rather than opening, usually need no password at all.

A file that is damaged or not a PDF is named and turned away with the reason. A single page that cannot be drawn — a broken font reference, a malformed image stream — is marked in the grid while the rest are produced. A forty-page document should not be thrown out because page nineteen is bad.

Nothing is uploaded

The rendering is done by pdf.js, running as JavaScript in this tab, on the bytes your browser read from your own disk. There is no upload endpoint on this site for this tool, which is a stronger promise than deleting files after an hour: the file was never anywhere to delete. The renderer itself, along with the font and character-map data some documents need, is served from this domain and fetched only after you hand in a file — never on page load.

You do not have to believe any of that. Open your browser's developer tools, switch to the Network tab, and convert something. You will see requests for the renderer and then nothing leaving with your document in it.