Anyone who shoots more than a handful of frames ends up with soft, out-of-focus shots buried in the library. Reviewing every image at 100 percent to catch the fuzzy ones is the classic advice, and it is genuinely miserable across a few thousand photos. A better approach hands the tedious part to software: let a sharpness score flag the weak frames, then you make the final call. Below is how that scoring actually works, how to review the results without losing anything you meant to keep, and how to do the whole thing offline on a Windows machine.
Why blurry shots pile up faster than you think
Burst mode is the usual culprit. Fire off ten frames of a kid running and maybe two are tack sharp; the rest range from slightly soft to clearly blurred. Add low light, a missed autofocus lock, or a bit of camera shake, and the reject pile grows with every shoot. Manually zooming into each frame to judge focus does not scale, and by the hundredth photo your eyes stop being reliable anyway.
Sorting by filename or date will not help here, because blur is a property of the pixels, not the metadata. You need something that looks at the image itself and estimates how sharp it is.
Let a sharpness score do the first pass
Focus can be measured. A common, well understood technique runs a Laplacian filter over the image and looks at how much the result varies: a crisp photo has strong edges and high variance, while a blurred one has soft edges and low variance. Software can compute that number for every frame in seconds and rank your library from softest to sharpest, so the likely rejects float to the top instead of hiding in the middle of a folder.
Sharpness alone is not the whole story, though. A photo can be perfectly in focus and still be a throwaway because someone blinked or the exposure was blown out. That is why the better cleanup tools score several things at once rather than treating blur in isolation.
How Memora surfaces blurry and otherwise weak frames
Memora is a Windows photo organizer with an AI Culling feature built for exactly this problem. Instead of deleting anything, it scores your shots and flags the likely rejects so you can review a short list instead of the whole folder. Four signals feed the scoring:
- Subject focus (sharpness). Laplacian-variance scoring estimates how sharp each frame is, pushing soft and blurred shots to the front of the review queue.
- Eye focus and eyes open. Local ONNX eye models catch closed eyes and unsharp faces, the two failures that quietly ruin otherwise good portraits.
- Exposure. Blown highlights and crushed shadows are detected from the pixels, so badly exposed frames get flagged alongside the blurry ones.
Burst frames are grouped together using capture-time gaps and a perceptual hash, so a run of near-identical shots is treated as one set and the best frame is highlighted. Four sliders (Subject Focus, Eye Focus, Eyes Open, and Exposure) run from 0 to 200, and the thresholds adapt to the folder you are working in, which means a dim indoor set is judged against its own baseline rather than a studio one. Dial the Subject Focus slider up when you want to be strict about blur, or ease it down for a looser first pass.
Review first, delete second: nothing is removed for you
Automatic deletion is where cleanup tools get scary, and it is worth being clear about how Memora handles it. Culling proposes; it never deletes. Flagged frames land in a Rejects view that you scan and confirm, and only then do you decide what actually goes. Treat the score as a fast filter that narrows thousands of images down to a reviewable handful, not as a verdict. A shot the algorithm calls soft might be an intentional motion blur you love, so the final judgment stays with you.
Blurry versus duplicate: two different cleanups
Getting rid of near-identical copies is a separate job from removing blur, and mixing them up leaves clutter behind. For exact and near-duplicate copies, Memora has a free Duplicate Finder that groups matching files with SHA-256 and perceptual hashing. Run culling to cut the soft and mis-focused frames, then run the duplicate pass to collapse the redundant copies. Two focused sweeps beat one vague one.
A candid note on "fixing" blur
Searches for blurry photos split into two very different wishes: some people want to find and remove the duds, and others want to rescue a single precious shot that came out soft. Be realistic about the second one. Memora finds, flags, and helps you cut blurry frames; it does not un-blur them, and honestly, no tool fully reconstructs detail that the sensor never captured. If a frame is unrecoverable, the kindest thing you can do for your library is stop letting it take up space and attention.
Do the whole thing privately, on your own machine
Photo cleanup often means opening your most personal images, so where the analysis happens matters. Every part of Memora's culling runs on your own CPU or GPU, with no network call involved: the sharpness math, the eye models, the exposure checks, and the burst grouping all execute locally. No photo, thumbnail, or score is uploaded anywhere. Compared with a cloud service that scans your library on someone else's servers, keeping the entire pass on your PC means your pictures never leave your computer. Curious how the rest of the app is built around that principle? The photo organizing software overview walks through it, AI Culling is part of the Pro feature set, and once your keepers are sorted you can group them into smart albums.
Frequently asked questions
How does software decide a photo is blurry? It measures sharpness from the image data, typically with a Laplacian-variance score: strong edges read as sharp, soft edges read as blurry. Memora combines that with closed-eye and exposure checks so a frame is judged on more than focus alone.
Will it delete my blurry photos automatically? No. Culling flags likely rejects and puts them in a Rejects view for you to confirm. You choose what to remove, and an intentional motion blur you want to keep stays put.
Can it un-blur or fix a soft photo? Finding and removing blur is the job here, not repairing it. Memora does not reconstruct focus that was never captured, so treat it as a way to clean the library rather than a rescue tool.
Do my photos get uploaded for the analysis? They do not. All scoring runs locally on your Windows PC, and because it all happens on your machine, your photos never leave your computer.
Ready to stop pixel-peeping every frame? Point Memora at a folder, run a cull pass, and let the sharpness score do the hunting so you can spend your time on the shots that deserve it.