Converting PDF pages to PNG

Where PDF came from, and why it is like this

The choice between PNG and JPG as the target for a rasterised PDF page is not a close call, and it goes the opposite way from the choice for photographs.

A PDF page is typically black text on white, with line art, tables and flat blocks of colour — content with hard edges and large uniform areas. That is precisely what PNG's lossless compression handles well and what JPEG's frequency-based compression handles badly. Rendering a text page to JPEG produces visible haloing around every letter; rendering it to PNG produces exactly the pixels the renderer drew.

The trade is size. For a scanned photographic page PNG will be much the larger file. For a page of text it is often the smaller one, because a page that is 90% white compresses extremely well losslessly.

Is this conversion lossy?

The rasterising is lossy in the sense that matters — the page stops being instructions and becomes pixels — but the encoding step adds nothing on top. What the renderer drew is what the PNG stores, exactly.

So the loss here is entirely about resolution rather than compression. Detail finer than the chosen rendering scale is gone, and enlarging the result afterwards shows pixels rather than revealing detail. Choose the scale for the largest size the image will ever be displayed at, because you cannot go back for more later without re-rendering from the PDF.

What happens to metadata, transparency and colour

The text layer is gone. Same as any rasterising: no selection, no search, no screen-reader access, no copy and paste. The words are a picture of words.

Document metadata, bookmarks, links, annotations and form fields are all dropped. A PNG has no structure to hold them.

Transparency is technically preserved and practically absent. PNG has an alpha channel, so a page rendered with a transparent background could keep it — but PDF pages are rendered onto white, so what you get is an opaque white page.

Colour profiles do not carry. A PDF prepared for print in CMYK with a specific profile becomes an untagged RGB PNG, and colours will shift. This conversion is not a print workflow.

One file per page, delivered as a ZIP for multi-page documents.

Why people actually need this

The distinguishing situation is documentation. A page of a manual, a spec sheet, a form or a diagram that needs to go into a wiki, a bug report, a support ticket or a slide — somewhere that displays images and where the crispness of the text actually matters because someone will read it on screen.

The other case is archival screenshots of a document's appearance: proof of what a page looked like, where lossless matters because the point is fidelity to the original rather than convenience.

And there is a practical one: some optical character recognition works measurably better on PNG input than on JPEG, because JPEG's artefacts around letterforms are exactly the noise that confuses character recognition.

What we measured

This conversion was run end to end before this page was written. Not "should work" — run, with the output checked:

And the ceilings measured for PDF to JPG, the tool that performs it, on the hardware described on the limits page:

Not tested

The page also refuses files above 100 MB before it starts, which is a guard written into the tool rather than something the browser imposes. Why those are different.

When something else is the better answer

If the pages are photographic — scanned photographs, image-heavy brochures — JPG will be dramatically smaller with no visible penalty on that kind of content.

If the text needs to stay text, do not rasterise at all: converting to a word processor format keeps the words as words, at the cost of the layout.

If the goal is reading the text rather than looking at the page, OCR goes straight to text and skips the intermediate image.

If the document is long, check the page-count row for this tool on the measured limits page before starting — rasterising is the most memory-hungry thing on this site, and page count matters more than file size.