Culling in photography is the process of reviewing everything you shot and deciding which frames to keep, which to reject, and which handful deserve your editing time. A wedding photographer might come home with 3,000 frames and need to reach a final gallery of 400. That first sorting pass, the one that happens before a single slider gets touched, is culling. For most photographers it is also the slowest and least loved part of the whole job.
This guide explains what culling means, why it matters more than it looks, how the manual process works, and where AI culling genuinely helps. It also covers a question the tool makers rarely raise: what happens to your unpublished photos while software is busy grading them.
What culling means, and why the word feels odd
To cull originally meant to pick out or gather from a larger group. The word still carries that sense in farming and conservation, which is why looking it up on its own tends to surface livestock and gardening long before it surfaces cameras. Photographers borrowed the term and narrowed it to one job: separating keepers from rejects inside a batch of very similar shots.
Three activities get muddled together, so it helps to keep them apart:
- Culling answers a yes or no question. Keep this frame, or drop it?
- Editing (or developing) comes next, and only applies to the survivors. Exposure, colour, cropping, retouching.
- Organizing is the long game. Filing, tagging, and being able to find a specific photo two years later.
Get culling wrong and you pay for it twice: once in wasted editing hours spent polishing frames you should have cut, and again in a bloated library that hides your best work behind hundreds of near misses.
Why culling is worth doing well
Speed of delivery is the obvious reason. A tighter first pass means fewer photos entering the edit queue, so clients get their gallery sooner. Decision fatigue is the quieter reason. Judging thousands of almost identical frames drains the same mental energy you need for creative editing, and by frame 1,500 most people start keeping shots they would have rejected at frame 50.
Storage is the third cost. Modern cameras fire long bursts, and a ten frame burst of the same moment eats real disk space while adding nothing to the final set. Culling those bursts down early keeps your archive honest and your backups smaller. Done consistently, a good cull is what separates a portfolio that reads as sharp and intentional from one that buries its highlights.
How photographers cull by hand
A manual cull usually runs in passes rather than one exhausting sweep. Many photographers start with a fast rejection round, flagging anything obviously unusable: missed focus, a blink, a badly clipped exposure, a frame someone walked into. Only after the clear rejects are gone do they go back and rate the survivors with stars or colour labels to mark the genuine standouts.
Bursts are where the real work lives. When you have eight frames of the same smile, judging which one wins comes down to small differences: whose eyes are fully open, which frame is critically sharp on the lash line, where the hands sit. Reviewing that at 100 percent zoom, burst after burst, is precise and slow. Anyone who has culled a full event knows exactly how long an evening it can be.
What AI culling actually evaluates
AI culling promises to take the first pass off your hands. Marketing pages rarely say how, so here is what the underlying signals really are. A culling model does not have taste. What it has is a set of measurable image qualities, and it scores each frame against them:
| Signal | What the software measures | Why it matters |
|---|---|---|
| Sharpness | Local contrast and detail, often via a Laplacian variance calculation | Separates critically focused frames from soft or motion blurred ones |
| Exposure | How many pixels are blown out to pure white or crushed to pure black | Flags frames that clipped highlights or lost shadow detail |
| Closed eyes | A dedicated eye state model checking each detected face | Catches blinks in group shots and portraits before you commit to a frame |
| Burst grouping | Capture time gaps combined with visual similarity (perceptual hashing) | Bundles near identical shots so you compare within a burst, not across the whole shoot |
Notice what is missing from that list. A model can tell you a frame is sharp and the eyes are open, but it cannot tell you the subject's expression is the one the client will love. Good culling software understands this, which is why the honest tools present their picks as suggestions and leave the final call to you.
The question the tool makers skip: where do your photos go?
Most of the well known AI culling products run in the cloud or behind an account. You upload a shoot, their servers grade it, and the results come back. For a wedding or a commercial job, that means an entire set of unpublished, often private client images leaves your computer and sits on someone else's infrastructure while it is processed.
Plenty of photographers are fine with that trade. Others are not, especially those shooting sensitive events, medical or legal work, or anything under a strict client contract. If your photos never being uploaded matters to you, the location of the culling engine is not a footnote. It is the whole decision. That is exactly where a local-first approach changes the maths, because the grading happens on your own machine and the files never travel.
How Memora approaches culling
Memora is a local-first photo manager for Windows, and its AI culling runs entirely on your own computer. No frame, thumbnail, or score is ever sent anywhere. The model scores each photo on the signals above, sharpness through a Laplacian variance measure, exposure through blown and crushed pixel counts, closed eyes through an on-device eye model, and it groups bursts using capture time gaps plus perceptual hashing so you review similar shots side by side.
Four sliders (Subject Focus, Eye Focus, Eyes Open, and Exposure) let you set how aggressive each judgement is, and the thresholds adapt to the folder you are working in rather than to some fixed global standard. One rule sits above all of it: Memora never deletes anything. It proposes rejects and highlights likely keepers, then you decide. If you want to clear out true copies as well, its duplicate finder groups exact and near identical files so you can remove them safely, and that feature is available for free.
AI culling itself is part of Memora Pro, while browsing, RAW preview, and duplicate detection are free to use. Because the whole workflow is on-device, culling folds naturally into everything else you do in the app, from developing the survivors to running AI semantic search months later to pull up a specific keeper by describing it. For the wider picture of how this fits a tidy library, our guide to photo organizing software walks through the full flow.
A repeatable culling routine
Whether you cull by hand or with help, a consistent order beats improvisation. Here is a routine that holds up shoot after shoot:
- Back up the full take first, untouched, so nothing you reject is ever truly at risk.
- Run a fast rejection pass for the obvious failures: missed focus, blinks, clipped frames.
- Work burst by burst, keeping the single strongest frame from each near identical group.
- Rate what survives, marking the genuine standouts you would show a client or print.
- Hand only that shortlist to your editor, and file the rest without deleting the originals.
Follow those steps and the dreaded first pass turns into something closer to muscle memory. If you want a tighter, faster version of this, our walkthrough on how to cull photos fast drills into the shortcuts.
Common questions about culling
Does culling delete my photos?
Not by itself. Culling is a decision about what to keep, and good software treats it that way. Memora, for instance, only flags likely rejects and never removes a file for you. Deleting anything stays a deliberate, separate step that you control.
Is culling the same as editing?
No, and mixing them slows you down. Culling picks the frames worth your attention; editing is the developing work you then invest in those chosen few. Keeping the two stages apart means you never waste retouching time on a photo that should have been cut.
How many photos should I keep?
Enough to tell the story and no more. Ratios vary wildly by genre, a portrait session lands very differently from a wedding, so chase a tight, strong set rather than a target number. If two frames say the same thing, keep the better one and let the other go.
Can software cull automatically for me?
Software can do the heavy first pass, scoring sharpness, exposure, and open eyes far faster than you can. What it cannot do is read the moment or your client's taste. Treat AI culling as a fast assistant that hands you a shortlist, then bring your own judgement to the frames that matter.
Culling will probably never be the fun part of photography. Done with a clear routine and, where it helps, a local tool that keeps your files on your own machine, it stops being the part you dread and quietly becomes the habit that makes everything downstream faster.