Closed Eyes Ruining Group Photos

Why blinks are a numbers game

People blink every few seconds without knowing it. One subject, one frame: usually fine. Twelve subjects, one frame: someone's eyes are closed more often than not. Family formals are the worst case — large groups, time pressure, and the highest expectations, since these are the images grandparents frame. And blinks are cruelly easy to miss during culling: a closed pair of eyes on a face that occupies 2 percent of the frame simply does not read at grid size. You find it when the mother of the bride does.

Beat blinks at the shoot

• Shoot at least three frames per grouping. Blinks are random; redundancy is the cure. For big groups, five frames is cheap insurance.
• Use a countdown. "Close your eyes… open on three" synchronizes everyone's blink cycle and buys you two clean seconds.
• Watch for the serial blinkers. Most groups have one. Once you spot them, glance at their face before you move on to the next grouping.
• Confirm before releasing the group. A two-second LCD zoom on faces costs less than a reshoot that can never happen.

Catch blinks in the cull

1. Cull each formal grouping as a set, side by side, exactly as you would a burst.
2. Zoom to each face once, left to right, on your candidate frame — a disciplined sweep, not a glance.
3. If no frame is fully clean, note it: a head-swap composite between two frames of the same grouping is standard practice, but only if you kept both frames.

Consider a 14-person family formal shot four times. Frames one and three each have a blinker — different people. The disciplined face sweep finds frame two clean in thirty seconds; without the sweep, frame one "looked fine" and shipped.

What eye checks cannot do

Manual face sweeps are reliable but slow, and they collapse under fatigue late in a cull. Automated eye detection has limits of its own: heavy shadow, glasses glare, profile angles, and tiny distant faces can defeat it, and squints, laughing eyes, and deliberate closed-eye moments (a first-dance embrace) are judgment calls, not detections. No system should auto-reject a frame purely because eyelids are down.

How Cull AI Studio handles this

Cull AI Studio runs face and open/closed-eye detection locally on your computer, using a bundled OpenCV YuNet face detector with eye cascades — no internet involved. Frames with detected closed eyes are scored down, and because every decision carries a confidence score, ambiguous cases — small faces, glasses, expressive squints — go to REVIEW for your judgment rather than a silent verdict. Rejected frames stay intact in their folder, which matters for formals: the blink frame you need for a head-swap composite is still there. Corrections you make (marking a "closed eyes" reject as a keep, or vice versa) are logged with reasons and feed your personal preference model over time. It will still miss some blinks in tough conditions — the face sweep on your final formal selects remains a good habit.