DEV Community: Abby Espinosa

Most "extract images from PDF" tools are just screenshotting the page

Abby Espinosa — Fri, 11 Sep 2026 05:01:31 +0000

If you have ever pulled "images" out of a PDF and they looked soft, that was not extraction. That was a render.

A PDF can hold a real JPEG or PNG as an XObject. The bytes are already in the file. You can lift them out at the original resolution. A lot of online tools do not do that. They draw each page to a canvas and export the canvas. You get a picture of the page, including the text around the photo, at whatever DPI they picked.

Those are different jobs.

How to tell which one you got

Extracted: the file that comes out is a JPEG or PNG. Dimensions match the photo the author placed, not the page. A 4000px camera original stays 4000px.
Rendered: you get one image per page. Text is flattened into pixels. Cropping the photo later is on you.

If the download is page-1.png and it includes the header and the footer, it was a screenshot.

Why the distinction exists

PDF.js (and anything sitting on a PDF parser) can walk the page's XObject map and hand you the raw stream. That is a copy. Rendering the page through a canvas is a photograph of the layout. Both are useful. Only one answers "get the pictures out of this deck."

I built the extract path because that is the query people type. The tool is here:

https://imissfiles.com/extract-pdf-images/

It runs in the tab. No upload. Open Network and watch — same check I wrote about last time.

What still fails

A "photo" that is actually a bunch of vector shapes will not come out as an image. There is no image stream.
Some PDFs wrap images in a form XObject. Those need an extra walk. Easy to miss on the first pass.
A scanned page is one giant image. Extracting it just gives you the scan. That is correct, and also not what people meant.

If you only needed the pictures, use an extractor. If you needed a picture of the page, use a renderer. Do not let a site blur those together and call it a feature.

How to tell if a website uploads your file (Network tab, two minutes)

Abby Espinosa — Wed, 09 Sep 2026 05:26:14 +0000

Most "free PDF" and "free converter" sites take the file. They say they delete it later. You do not have to take that on faith.

This is the check I use. It works on any site, including mine.

Setup

Open the converter in Chrome or Edge.
Press F12 (Mac: Cmd + Option + I).
Click Network.
Check Preserve log.
Click the clear-circle icon so the list is empty.

Leave that panel open.

Run it

Drop a file you do not care about. A 2–3 MB photo is easy to spot. Watch the list while the tool works.

What an upload looks like

Ignore most of the GETs. Those are the page fetching JS, WASM, fonts, ads.

An upload is usually:

Method: POST or PUT
Size: close to your file. A 3 MB photo makes a ~3 MB request.
URL: something like /upload, /api/convert, or an S3 / GCS host.

Click that row → Payload. If you see the file, it left the machine.

What local processing looks like

After you add the file, you should not see a request near the file's size. You may see the page download a WASM engine before you pick a file. That is code coming in, not the document going out.

Size is the tell. Method is the second tell.

Two traps

Ads and analytics. AdSense and GA fire requests. They are small. They are not a 3 MB PDF.

Turning Wi-Fi off. Only useful after the page and its engines have finished loading. If you cut the network too early, a local tool looks broken because the WASM never arrived.

Why this is better than a privacy policy

A policy is a promise. The Network panel is a log. If the site has no upload endpoint, there is nothing to argue about.

I keep the longer writeup, with screenshots, here:

https://imissfiles.com/guides/check-if-a-website-uploads-your-file/

I also run a suite of in-tab PDF and image tools on the same domain. Affiliation: I built it. Same test applies there. Ads are on the page. They do not receive the file.