How to Cull Photos Fast: The 4-Pass Method for Big Libraries

August 14, 2026

Culling is where photo organization projects go to die.

You open a folder of 4,000 vacation photos determined to clean it up. Twenty minutes later you've evaluated maybe sixty images, you're exhausted from the decisions, and you've deleted eleven. At that rate the folder takes twenty hours. You close the laptop and it stays closed.

The problem isn't willpower. It's that you're making one hard decision 4,000 times. The fix is to make four easy decisions 1,000 times each — passes that get progressively more selective, where each pass has exactly one criterion and requires almost no thought.

Done this way, 4,000 photos takes about ninety minutes.

Before you start: three rules

Rule 1: Move to an archive folder, don't delete. Create _ARCHIVE-CULLED and send everything there. Wait three months. Then empty it. This single change removes almost all the anxiety, and anxiety is what slows you down. You're not destroying anything — you're moving it out of the way.

Rule 2: Back up first. You're about to move thousands of files quickly. One mis-click on a select-all can do real damage.

Rule 3: Never cull the same photo twice. If you looked at it in Pass 1 and kept it, don't reconsider it until Pass 2. Re-litigating decisions is where the hours go.

Set up your view first

This matters more than people expect. The right view makes culling three times faster.

Sort by date taken, oldest first. Photos from the same moment sit together, which is what makes bursts and near-duplicates visually obvious.

Use a large grid view — big enough to judge, small enough to see 30 to 50 at once. You want to scan a screen, not click through images one at a time.

Learn three keyboard shortcuts: select, shift-select a range, and delete/move. Reaching for the mouse for every action is the single biggest speed killer.

Turn off any preview pane that loads slowly. If you're waiting on thumbnails, you'll lose momentum.

Pass 1: The obvious garbage (fast, ~10 minutes per 1,000)

One criterion: is this a photo of nothing?

Delete without hesitation:

Accidental shots — the inside of a pocket, a blurred floor, a thumb
Completely black or completely white frames
Photos so blurry the subject is unrecognizable
Duplicate frames where you can see they're identical
Screenshots of things you no longer need
Photos of documents, receipts, or screens you've already dealt with

This is not a judgment call. If you have to think about it for more than a second, keep it and move on — it'll come up in Pass 2.

Expect to cut 10 to 20 percent here, and to move very fast. You're scanning a grid and selecting the obvious failures, not evaluating anything.

Pass 2: Burst and near-duplicate reduction (~25 minutes per 1,000)

One criterion: of these near-identical shots, which is best?

Sorted by date, bursts appear as runs of visually similar frames. Five shots of the same sunset. Eight attempts at a group photo. Twelve frames of a kid blowing out candles.

For each run:

Click the first, shift-click the last to select the whole run
Ctrl/Cmd-click to deselect the one or two you're keeping
Move the rest

How many to keep:

Landscapes, objects, food: keep one. There's no version where you want three shots of the same plate.
Groups of people: keep two or three. In any given frame somebody's blinking, and you won't know which until you look closely later.
Kids and animals: keep three or four from a burst. Expressions vary enormously and the "best" one isn't obvious at thumbnail size.
Anything genuinely irreplaceable: be generous. A burst from a wedding ceremony is not the place to be ruthless.

This pass produces the biggest reduction — typically 30 to 40 percent — because near-duplicates are the bulk of most modern libraries. Phones encourage rapid-fire shooting and nobody ever goes back.

Pass 3: The "would I ever look at this again" pass (~30 minutes per 1,000)

One criterion: is there any plausible future in which someone opens this photo?

Now you're making real judgments, but on a much smaller set.

Cut:

Scenery with no people that isn't remarkable
Photos of a place you visited that look like every other photo of that place
Meals, unless the meal was the point
Photos from events you don't remember and that contain no one you know
Poorly framed shots where a better version of the same subject exists
Anything work-related from a job you left

Keep, even if imperfect:

Anything with a person in it. This is the most important rule in the article. Bad lighting, awkward pose, closed eyes — none of it matters in twenty years. What matters is that the person is in the frame.
Anyone who has died, or anyone elderly. No exceptions, no quality standard.
Children at any age. Kids change so fast that even a mediocre photo captures something that no longer exists.
Houses, cars, rooms, streets — ordinary environments that will look extraordinarily dated in thirty years. Nobody photographs their own kitchen on purpose, and everyone wishes they had.
Anything that makes you feel something, even if you can't say why.

The asymmetry is the whole point: deleting a photo you'd have wanted costs you something irreversible. Keeping one you didn't need costs you three megabytes.

Pass 4: The highlight selection (~15 minutes per 1,000)

One criterion: is this one of the best 20 or 30 photos from this event?

This pass doesn't delete anything. It promotes.

Go through what survived and mark the genuine standouts — favorite, star, flag, whatever your software offers. Or copy them into a highlights subfolder.

2024-07 Lake Weekend/
  highlights/
    (20 photos)
  (the other 340)

Why this matters more than the deleting:

The realistic future of your photo library is that nobody scrolls 340 photos from a lake weekend. But 20 photos? Those get looked at. Those go in a photo book, a slideshow, a frame, a text to your sister.

A library with highlights marked is a library that gets used. A library without them is an archive that gets stored and forgotten, no matter how well you culled it.

If you only have energy for one pass, make it this one. Promoting the best is more valuable than removing the worst.

Realistic expectations

Starting with 4,000 photos from a big trip:

After	Remaining	Time
Pass 1	~3,400	40 min
Pass 2	~2,200	100 min
Pass 3	~1,400	120 min
Pass 4	~1,400 + 60 marked	60 min

Roughly a 65 percent reduction, in about five hours across a few sittings.

For a whole library rather than one trip, don't try to do it all. Pick one year, do the four passes, stop. Come back another time. A library culled one year at a time over three years is a finished project; a library you tried to do in one weekend is an abandoned one.

When not to cull

Some situations call for keeping everything:

You're culling within a month of the event. You don't yet know what mattered. Which photo becomes precious often isn't clear until later — sometimes years later, after someone dies or a place changes. Wait at least six months.

You're culling someone else's photos. Inherited collections are not yours to thin out. Someone else may want them, and you can't know what a cousin will find meaningful. Organize them; don't delete them.

You're upset. People delete photos of exes and estranged relatives in a bad week and regret it in a good year. Move them to the archive folder instead and revisit later.

Storage isn't actually a problem. If you have room, the case for aggressive deletion is weaker than it feels. The real benefit of culling is findability, not space — which is exactly why Pass 4 matters more than Passes 1 through 3.

Culling scanned and inherited photos

Different rules apply entirely.

Be far more permissive. These are already survivors — someone chose to print and keep them, often across decades. The selection has been made.

Never delete a photo with an unidentified person in it. Somebody knows who that is. You just haven't found them yet.

Delete only: true duplicates of the same print, badly damaged photos where a better copy exists, and photos of landscapes with no people or context.

That's it. When in doubt with an inherited photo, keep it.