Download every PNG on a web page
Paste a page URL and the scanner reads the HTML the server sends, collects every PNG it references, and hands you a grid already filtered to PNG. One thing to know before you look at it: the format label comes from the image's file extension, so a PNG served from an extensionless CDN URL arrives under UNKNOWN rather than PNG — untick the filter and it is there.
What this does
- PNGs are found everywhere the page references them — not just img src attributes, but srcset candidates, picture sources, CSS backgrounds in stylesheets and inline styles, lazy-load attributes, favicons, Open Graph tags and JSON-LD.
- Arriving from this page, the grid is already filtered to PNG — and the filter is a ticked checkbox in the sidebar, not a hidden mode, so one click widens it to everything found.
- What you download is the origin's exact bytes. Nothing is re-encoded, recompressed or flattened, so transparency and interlacing survive the trip.
- Dimensions a page declares show immediately; exact pixel size and file size come from a single 4 KB range request, made only for the images you ask about.
How the PNGs are found
A scan fetches the page once and reads the HTML exactly as the server sent it —
no browser, no JavaScript execution. Every place a page can reference an image
is read on that single pass: the src and srcset on <img>, the <source>
elements inside <picture>, background-image in inline styles and in
<style> blocks, the lazy-loading attributes sites use in place of src,
<link rel="icon"> and Apple touch icons, og:image and twitter:image, and
image fields in JSON-LD. Linked stylesheets are fetched and read too, up to
three of them.
Content inside <noscript> is parsed as well, through the same pipeline. That
is deliberate rather than clever: <noscript> is the site’s own answer to a
non-JavaScript reader, and a static parser is one.
What arrives
Every PNG the page points at, deduplicated by normalised URL, with the format and — where the page declares them — the dimensions already on each tile. Thumbnails load straight from the origin, so previewing costs the site the same requests a visitor’s browser would have made.
Downloading is the only step that goes through our proxy, and it has to: a browser cannot read cross-origin image bytes or honour a download link across origins. The bytes pass through and are forgotten — there is no copy on our side to keep.
What it won't do
- PNG is read from the URL's file extension. A PNG served from a URL with no extension — ordinary on image CDNs — is listed under UNKNOWN rather than PNG. Untick the format filter to see those.
- Only the HTML the server sends is read. A gallery that builds its img tags in JavaScript after load has no PNGs on the page for us to find.
- Three linked stylesheets are fetched per scan. A PNG background declared in a fourth is missed.
- One archive holds up to 500 images. A larger selection has to go in more than one ZIP.
- If the site's robots.txt asks automated tools not to read the page, the scan stops and says so. There is no override.
Questions
Why are some PNGs listed as UNKNOWN?
The format label comes from the image URL's extension. Plenty of CDNs serve paths like /image/abc123 with no extension at all, and those land in UNKNOWN even when the bytes are a PNG. They are still found, still previewable and still downloadable — just not under the PNG filter. Untick PNG to see the whole set.
Will transparency survive the download?
Yes. The proxy streams the origin's response through untouched — no re-encode, no recompression, no background flattening. The file that lands on your disk is byte-for-byte the file the site serves.
The site serves WebP to my browser. Can I still get the PNG?
Often, yes. We read the page's markup rather than what your browser chose to load, so a <picture> element's PNG fallback is found even when your browser took the WebP. One picture is one tile, with its other versions behind the "2 versions" chip in the corner — open it and the PNG is there. Filtering to PNG does the same job in one step: the tile then shows its PNG rather than the WebP.
Does changing the filter re-scan the page?
No. The scan reads everything once and the filtering happens in your browser, so widening from PNG to every format is instant and costs no extra request to the site.
Related
Images belong to their creators. Only download what you have the right to use. See all tools or read what our requests do.