On culling: narrowing thousands of frames to the keepers

A day on a working set can leave us with four thousand frames and a client who wants forty. Most photographers will tell you the edit is where the work is really made, and they are right, but they rarely tell you how much of the edit is just triage: clearing the floor of everything that was never going to make it, so the frames that matter have room to breathe. That first pass is not taste. It is bookkeeping. And bookkeeping is exactly the part of the craft we have learned to systematise without apology.

We run three passes before a human looks at anything for its merits. The first is focus: a frame that missed critical focus on the eye is gone, no discussion, because no amount of colour work will rescue a soft portrait and pretending otherwise wastes an afternoon. The second is the blink: closed eyes, half-lidded eyes, the micro-expression a subject makes a quarter-second before they settle. The third is the near-duplicate. On a burst you might hold nine frames of the same gesture; eight of them are noise around the one that is alive. We collapse the burst down to a shortlist and let the shortlist compete.

None of this is glamorous and all of it is honest. A frame that is out of focus is out of focus whether a person or a script flags it. What we are buying with the triage passes is attention: the finite, expensive kind. By the time an editor sits down with a coffee, they are not scrolling past four thousand mediocrities to find the good ones. They are choosing between two hundred candidates that all cleared the bar. That is a different job, and it is the job we are paid for.

Here is the line we do not cross, and we want to be plain about it. The triage decides what is technically viable. It never decides what is good. It does not know that the frame where the subject looked slightly away is the one that carries the whole story, or that the "imperfect" gesture is the human one and the clean one is dead. Those calls belong to a person who was in the room, or who at least understands the room, and we keep them there on purpose. A rule that is allowed to pick favourites will pick the safe frame every time, and the safe frame is almost never the memorable one.

We have watched what happens when a studio forgets this. The edit gets faster and quietly worse. Everything is sharp, everyone is smiling, and nothing is felt. The frames that would have made someone stop scrolling get culled in the second pass because they broke a rule the rule could not read the reason for. So we tuned our passes to be generous, keeping too much rather than too little, and to surface the borderline calls to a human instead of resolving them silently. A false keep costs a few seconds of review. A false cut costs the best frame of the day, and you never even know it is gone.

The result is unremarkable to describe and enormous in practice. We deliver faster because the machine does the sorting a machine should do. We deliver better because the person does the choosing a person should do. And the client never sees any of it. They see forty frames that feel chosen, which is the only outcome that was ever the point.