dupfind

Before you delete 40,000 photos

sheet 03 · 12 jul 2026

Photographs are the one category of file where the mistake is permanent and the loss is personal. A deleted spreadsheet is an inconvenience. A deleted photo of someone who has since died is a different kind of event, and no amount of reclaimed disk space compensates for it. So before the satisfying part — watching thousands of redundant copies disappear — here is the unglamorous list, assembled from people who have made these mistakes.

1. back up first, even though you won't want to

The whole reason you're here is that photos are taking up space, so "copy them all somewhere else first" feels absurd. Do it anyway: an external drive, a cheap cloud tier, anything. Bulk operations are exactly when a slip becomes catastrophic, and the backup only needs to exist for the hour it takes to be sure. Delete the backup afterwards if space is truly that tight — but have it for the hour.

2. start strict, then loosen

Run the exact and tight settings first and clear those groups: they're the safe wins, and they're usually most of the reclaimed space anyway. Only then move to normal, and treat loose as a review exercise rather than a bulk operation. Deleting 3,000 byte-identical duplicates is arithmetic. Deleting 300 "nearly the same" photos is editing, and editing deserves your eyes.

3. burst shots are not duplicates

The single most common regret. Ten frames from a burst look like ten copies to any hashing algorithm — and one of them is the one where the child is laughing and the other nine are blinks. The looser the setting, the more of these appear. If a group's thumbnails show tiny differences in expression, position, or timing: that's not a duplicate group, that's a moment, and the keeper is a decision only you can make.

4. keep the biggest — but check the dimensions

Biggest file usually means original, and the table pre-selects it for that reason. The exception worth catching by eye: a file that's large in bytes but small in pixels is a low-resolution copy someone saved at maximum quality — the real original is elsewhere in the group, smaller in bytes, larger in dimensions. The review wall shows both numbers precisely so this trap is visible.

5. never trust a tool that won't show you

The final rule, and it applies to every duplicate finder including this one: if a tool proposes to delete files without showing you thumbnails of exactly what it's removing, close it. "Auto-clean" on a photo library is a promise no algorithm can safely make, because the algorithm doesn't know which face in which frame matters to you. The right division of labour is old and boring: the machine finds candidates in seconds; the human makes the judgement in minutes. Any tool that reverses this — asking you to trust it and click one big green button — is optimising for the wrong thing, and it isn't your photographs.