dupfind hashed on your device · photos never leave

The same photo, forty times.
Let's find them all.

Point it at a folder and it finds the duplicate and similar photos that aren't byte-identical — resized, re-saved, WhatsApp'd. All the hashing happens in this tab, and nothing is deleted unless you say so.

Choose a folder of photos. On Chrome and Edge you'll be asked for permission and can delete right here; other browsers scan and hand you a list.

Where the pile came from: my father's laptop had 41,000 photos and about 12,000 actual photographs. Every WhatsApp forward, every "let me back this up" that copied instead of moved, every export from a phone that renamed everything, all piled into one folder called Pictures (new), itself sitting inside Pictures (old). He wouldn't delete any of it because he couldn't tell which was which. The tools I found for him were downloads from sites with three buttons all called Download, and the first one I dared try wanted forty dollars to actually remove anything. His photos never needed to leave the laptop, and neither did his money. A browser can hash forty thousand images in a few minutes; the rest is a decent review screen and the discipline to never delete anything he didn't tick.

— he kept 12,400 photographs and 3 copies of a picture of a cat he doesn't know

How it finds duplicate and similar photos

two fingerprints, both hand-written, both on your machine

  1. Every photo gets shrunk to a thumbnail — because duplicate detection cares about shapes, not detail, and small images hash fast.
  2. Two fingerprints are taken. A gradient hash records where brightness rises and falls across the frame. A structural hash keeps only the coarsest shapes and throws away fine detail and overall brightness.
  3. Photos are compared bit by bit. Two photos are lookalikes only when both fingerprints agree closely — the double check is why different-but-similar photos rarely get grouped.
  4. You review and decide. Groups appear as thumbnails, biggest file pre-selected to keep. Nothing happens until you press the button and confirm.

What it catches, honestly

the strictness dial, explained by what it finds

Rules of the table

what this tool will never do

Asked over the pile

the questions that come with 41,000 photos

Does it handle HEIC from iPhones?

If your browser can decode HEIC it hashes them like anything else; if not, they're skipped rather than silently mis-grouped. Converting them first (our sibling site does this on-device too) makes the whole library easier to live with anyway.

Which copy should I keep?

The biggest file is pre-selected because it's usually the least-compressed original — but check the dimensions on screen: a bigger file at smaller dimensions means someone saved a low-res copy at high quality, and the original is elsewhere in the group.

Can I scan an external drive?

Yes — any folder your computer can see, including drives and network shares, as long as you can pick it in the folder dialog.

Where do deleted photos go?

The browser removes them from the folder. Depending on your operating system they may or may not land in the recycle bin — so treat deletion as permanent, and if these are irreplaceable, have a backup first. This is stated in the confirmation dialog too.

Will it find duplicates between two different folders?

Pick a parent folder that contains both, keep "include subfolders" ticked, and it compares everything as one pile — which is exactly how the "Pictures (new)" inside "Pictures (old)" situation gets solved.

Sheet notes

longer reads from the light table

How a computer sees two photos as one Perceptual hashing explained: why a 4MB original and a 200KB forward fingerprint the same. The duplicate-finder download trap Why searching for this tool leads somewhere unpleasant, and what those installers actually do. Before you delete 40,000 photos A short, unglamorous checklist from people who have regretted a bulk delete.

Sweep log

what changed, newest first

v1.0
table opens: dual perceptual hashing (gradient + DCT structural) in a worker crew, four strictness levels, exact-bytes fingerprinting, thumbnail review wall with keeper selection, confirmed deletion via File System Access, CSV verdict export for every other browser, recursive folders, offline PWA.