Photo selection software: picking the best frames from a shoot of thousands
Good photo selection software does three jobs that take a person too long. It discards the failed shots. It selects one frame from every burst. It spreads the chosen frames across the full day, so the album does not hold two hundred pictures of one group pose. Yet it cannot tell which people matter. That work remains with you.
The group photo nobody blinked in
The reason a shoot runs to thousands of frames is not carelessness. It is arithmetic. In 2006 Nic Svenson and Piers Barnes of CSIRO worked out how many shots it takes to get one group photo with nobody blinking, and won an Ig Nobel Prize for it. Their rule of thumb, as ABC Science reported it: for a group under 20, "divide the number of people by three to get the number of shots you need to take", and by two in poor light. Near 50 people, a blink-free frame becomes essentially impossible.
A photographer shoots a table of fifteen at a dim reception. He takes seven or eight frames of that same second on purpose. Multiply that across a day and the folder is huge for good reason. The job of selection is to find the one frame in each burst where the eyes are open. You must do that for every burst in the shoot.
Sharpest is not best
The simple shortcut for automatic photo selection is to rank by sharpness and keep the top 150. This gives an uneven mix. The sharpest pictures gather where subjects were slow and well lit: the couple motionless on stage, posed family lineups, and rings on a table. The dance, the dim hall, and candid frames lose to those lineups every time.
Frame-level sharpness has a second flaw. The standard measure, the variance of the Laplacian, comes from microscope autofocus research by Pech-Pacheco and colleagues in 2000 and was popularised for ordinary photographs by Adrian Rosebrock's blur detection tutorial. It scores the whole frame, so a crisp background can carry a soft face. Measuring sharpness on the people fixes that.
Three passes over every frame
You run our free Photo Selector in your browser. It works in three passes.
- Faces first: Every face is checked for sharpness at the eyes, open eyes, expression, head pose, and exposure on the skin. The entire frame gets scored as well. But a soft face will always lose to a sharp one, no matter the background.
- One per moment: capture time puts frames into bursts. Bad shots leave first, each with a reason. The best shot in each burst stays.
- Coverage: each part of the day gets its share of the total you ask for, and nearly identical shots never take two spots.
Slide the count to the number you want. The set rebuilds itself. Every pick tells you why it won its moment. Your photos never leave your computer.
The part that stays human
Software is good at the mechanical rejects and poor at meaning. When we measured one real 245-frame wedding session, the defects a machine can clear without argument came to about 7 per cent of the shoot, against the 70 to 75 per cent that culling vendors such as Narrative say a wedding cull removes. The rest of the cut is choosing, which we wrote up in what photo culling actually involves.
Use the automatic selection as a first pass. Run through it once. Bring back the slightly soft picture of the grandmother that the software could not know mattered, then export. If a couple is picking pictures for an album, the story-first method in choosing wedding album photos beats any score.
A short glossary
- Burst: a few shots taken within a second or two of one another, mostly of the same moment.
- Keeper: a shot that makes the cut and moves to editing or delivery.
- Coverage: how well the chosen pictures spread across each part of the day.
- Cull: the pass that drops bad and duplicate frames before you choose anything.
Frequently asked questions
How do you automatically select the best photos from a large shoot?
Choose photo selection software that looks at the people in the frame rather than the frame alone. It checks for sharp faces, open eyes, and good expressions. It groups bursts by capture time. It keeps the best frame from each moment and spreads the count across the day. Check the result yourself once before you export.
Why not just keep the sharpest photos?
The sharpest frames come from posed shots in good light. A set sorted by sharpness is mostly lineups. Whole-frame sharpness can also favour a sharp background over a soft face. Score the faces and spread picks across the day. Then the photos look like the event.
How many photos should you take of a group?
CSIRO researchers offer a rule of thumb for groups under 20. Divide the number of people by three in good light, or by two in poor light. Past about 50 people, a frame with nobody blinking is very unlikely. Selection must then judge eyes frame by frame.
Is the Photo Selector free?
Yes. It runs inside your browser. You upload nothing and you create no account. AccioPix makes its money from photo delivery, not from the free tools.
References
- ABC Science, 6 October 2006. Blink-free photos win Ig Nobel, on the work of Nic Svenson and Piers Barnes, CSIRO Industrial Physics. Read 29 September 2026.
- Pech-Pacheco, J. L., Cristobal, G., Chamorro-Martinez, J. and Fernandez-Valdivia, J., 2000. Diatom autofocusing in brightfield microscopy: a comparative study. 15th International Conference on Pattern Recognition (ICPR), vol. 3, pp. 314-317.
- Rosebrock, A., 2015. Blur detection with OpenCV, PyImageSearch.
- Narrative, Photo culling: the complete FAQ for photographers. Read 20 September 2026.
- The 245-frame measurement is described, with its script, in what photo culling actually involves.
The ratios, the reject categories and the two-pass method behind every cull.
A hands-on framework for cutting thousands of shots down to a tight, varied album.
Don't need an exact count? Cull every soft, out-of-focus shot in one pass with the Blur Culler.
More browser-based tools for culling, converting, and compressing — no uploads, no subscriptions.
AccioPix
← Back to Blog