PDF Image Extractor

Pulls every embedded picture out of a PDF, suggests a filename from the caption or heading text sitting next to it on the page, and lets you review each name before you export. Large, full-bleed images are also flagged as "Hero" shots and tagged Exterior or Interior when the caption says so (or, failing that, a best-effort guess from the image's own colours) — you can always correct the tag yourself before exporting. Each image is also sorted into a folder — Aerials & Views, Location Images, Photography, Renders, Materials, Maps, Lifestyle, Floorplans, or Uncategorised when nothing confidently matches — based on keywords in that same nearby caption text, with the folder shown per image so you can move anything before exporting. Everything runs locally in this browser tab — the PDF never leaves your machine unless you explicitly turn on the Google Drive upload below.

1. Choose a PDF

Click to choose a PDF, or drag one in here.

Extraction quality

Skip anything smaller than 32px

Skip full-page background images

Tag as "Hero" when an image covers at least 30% of the page

Guess exterior/interior from colour when there's no caption

Pages to process (e.g. 1-20,45; blank = all)

Dense line-drawing pages (floor plans, general arrangement CAD exports) can take a lot longer to process than photo pages, and rarely contain images worth extracting anyway — if you know which pages hold the actual renders/photos, listing them above will make a big difference on a large PDF.

2. Review names, folders and export

Images are grouped below by their suggested folder. The folder shown on each card is a best-effort guess from keywords in the nearby caption text — change the dropdown on any card to move it to a different folder before exporting. Anything in "Uncategorised" didn't contain a keyword matching one of your 8 folders (this is common on images captioned only with a project or room code-name).

Optional: apply AI folder suggestions from a chat

This tool never sends your images anywhere on its own — folder sorting above is keyword-only. If the keyword guesses above are missing too much, you can get a sharper pass by attaching this PDF to a chat with Claude (with the companion prompt provided alongside this tool) and asking it to look at the actual images. Paste what it gives you back below and click Apply to sort and rename matching cards automatically — matched by page number and the image's own pixel dimensions, both shown on each card above.

Optional: upload straight to Google Drive instead of downloading a ZIP

Each selected image is uploaded into a subfolder matching its assigned folder (Aerials & Views, Location Images, Photography, Renders, Materials, Maps, Lifestyle, Floorplans, Uncategorised) inside the Drive folder below — those subfolders are created automatically the first time you upload, and reused on later uploads.

This talks directly to Google's own servers from your browser using Google's official sign-in (nothing goes through any third party). It only works if this page is opened over http:// or https:// (for example via python3 -m http.server in the folder you saved this file, then opening http://localhost:8000/<filename>.html) — Google does not allow this sign-in flow for files opened directly by double-clicking (file://).

One-time setup (a few minutes): go to Google Cloud Console › Credentials, create a project if needed, click Create Credentials › OAuth client ID, choose Web application, add the exact address you'll open this page from (e.g. http://localhost:8000) under Authorized JavaScript origins, then paste the Client ID it gives you below. You only need to do this once — reuse the same Client ID every time you use this tool.

Google Client ID

Drive folder ID

Runs entirely client-side. Filenames are a best-effort suggestion from nearby PDF text — always double check before you rely on them.
Works fine opened directly (double-click). Serving it locally (e.g. python3 -m http.server, then open http://localhost:8000/<filename>.html) isn't required for extracting or exporting — it's only needed for the optional Google Drive upload above.