Find duplicate photos — right in your browser

What does this tool do?

The tool scans a collection of photos for duplicates and visually similar shots and groups them clearly. For each group it suggests the best photo to keep — by default the highest resolution, with the largest file as a tie-breaker. A summary shows how many groups were found, how many extra copies exist, and how much space you can reclaim.

All processing runs in your browser via WebAssembly. There is no upload, no install, and no account. That is the key difference from most alternatives: desktop programs require installation and lock you to Windows or Mac, while online services upload your private photos to someone else’s server.

How does the two-stage detection work?

The tool works in two stages you can freely combine.

Stage 1 — Instant scan (default, no download). Each photo is turned into a perceptual hash: the image is scaled down to a tiny grayscale preview and reduced to a 64-bit fingerprint. Two shots with almost the same fingerprint — measured by Hamming distance — go into the same group. This reliably recognises the same image even when it has been resized, recompressed, renamed, or converted to another format. The tool also compares the exact file contents to flag truly identical files as such.

Stage 2 — Visual search (optional, one-time model download). With one click the tool loads a small, specialized neural network (about 25 MB) into the browser. It turns each photo into a feature vector; similar subjects produce similar vectors. Using cosine similarity, the tool then finds shots a pure file fingerprint misses: bursts of the same moment, different crops, or slightly changed angles.

Both stages share the same grouping by connected components: if photo A matches B and B matches C, then A, B, and C all land in one group — even if A and C alone would not directly match.

Why doesn’t the tool delete anything itself?

For good reason, a browser is not allowed to delete files on your disk — that would be a major security risk. So the tool is deliberately a preparer, not a deleter: it shows you transparently which photos are duplicated, suggests the best one to keep, and lets you decide per photo.

You can take the result away two ways: as a ZIP of all kept photos, or as an agent-ready removal list in CSV or JSON. That list records, per group, which file to keep and which to remove, along with file size and reclaimable space. A script or AI agent can then handle the cleanup without your photos ever leaving the device.

When does the tool produce good results?

The instant scan is excellent for the most common cases: the same image saved multiple times, an original plus a smaller web version, the same snapshot in JPG and a PNG export of a HEIC original. Here the hash method finds duplicates precisely and without false positives across completely different subjects.

The visual search pays off when you want to thin out burst series — say ten near-identical portraits in a row, of which you keep only the best. It also finds cropped or lightly edited variants that the fingerprint overlooks because of the changed composition.

Set the sensitivity to match: Strict for near-identical photos, Balanced as the default for resized and recompressed versions, Loose when you deliberately want to group looser resemblances too. On “Loose” the tool may group more generously than needed — review the suggestions especially carefully in that case.

Is my photo truly private?

Yes. Neither the instant scan nor the visual search sends your photos to a server. There is no third-party cookie, no signup, and no usage analytics. The only network request is the one-time model download for the optional visual search — and it transfers only the model file, no image data.

For the “visually similar” groups optionally suggested by a neural network, the tool surfaces a fixed note in line with EU AI Act Article 50: the grouping is a suggestion, not a verdict — review it before deleting photos, because a neural network can group images that only resemble each other superficially.

Frequently Asked Questions

The key questions about usage, detection, and privacy:

How do I find duplicate photos without software?

Drop your photos into the tool above — it runs entirely in the browser, with no install and no upload. The instant scan groups exact and near-identical shots immediately.

What’s the difference between duplicate and similar photos?

Duplicate photos are identical or near-identical files. Similar photos show the same subject from a different angle or as a burst — those are found by the optional visual search.

Are my photos uploaded?

No. The entire analysis runs on your device. Your photos never leave the browser tab.

Which image tools pair well with this?

More tools from the kittokit ecosystem that fit cleaning up your photo library:

• Background Remover — cut out subjects after you have sorted out the duplicates.
• Image Format Converter — convert the kept photos into one modern, compact format.
• Image Resizer — scale kept originals to fit the web or your archive.
• Remove Metadata — strip GPS and camera data from kept photos before sharing.
• EXIF Viewer — inspect the capture data of the photos you keep.