How much time AI culling actually saves: the honest math

You have seen the number: AI culling gives you back ten hours a week. Nobody has measured your week, so that is an advertisement, not a finding. What follows is the honest version. The one sourced figure worth citing, the five costs that get left out of the sums, and the arithmetic to work out your own answer over three shoots.

Where the ten hour number comes from

Search for AI culling and the same promise comes back in different fonts. Ten hours a week. Half your post-production. Your weekends, returned. None of it arrives with a method attached: no sample size, no definition of what was being timed, no note on what the photographer was doing before.

We are not adding another one to the pile. Nobody has measured your week, and that includes us. What this piece does instead is give you the only sourced figure in the whole conversation, list the costs that quietly fall out of the marketing math, and hand you the arithmetic so the number you end up quoting is one you can defend.

That is less satisfying than a headline. It is also the only version you can plan a season around. If what you want is a target to beat rather than a promise to believe, our culling time benchmarks are built for that, and how long culling takes answers the plain question first.

The only survey worth citing

One number in this field comes from asking a lot of working photographers rather than from a vendor's own blog. A UK survey of more than 300 wedding photographers, reported by PetaPixel, asked where the working year actually goes.

A UK survey of 300 plus wedding photographers, published 2020. Respondents averaged 28 weddings a year at about £1,590, roughly $2,050, for a full day.

Two things follow, and neither one is a headline. First, culling is real but it is not the biggest slice. Editing is five times larger. A tool that perfectly handled your first pass still could not save you more than the culling share of your time, so any claim that starts at half your post-production is describing something other than picking photos.

Second, eleven percent of a 40 hour week is a little over four hours. Eleven percent of the 60 hour weeks you work in August is closer to seven. Your answer moves with your season, which is exactly why one universal figure should make you suspicious.

Read the percentages as shape rather than as a timesheet. Run them backwards and they break: 28 weddings at ten hours of coverage is 280 hours behind the camera, and if that were literally four percent of the working year, the year would run to 7,000 hours, or 134 hours a week. That is not a bad survey, it is a survey about how the year feels, and how it feels is that shooting is a rounding error and post-production is the job. Note the date too: 2020, before this generation of tools existed. It describes the problem, not the fix.

Five costs the marketing math leaves out

An honest comparison times the whole route, not the fast part in the middle. Five stretches of that route get skipped almost every time.

1. Getting the photos where the tool can see them

Cards still have to come off the camera, and that is the same wall clock whichever way you work. If the tool is cloud based, add the upload: Imagen runs in a browser and your images go up before anything happens, which on an ordinary domestic connection is a real part of the evening. Desktop tools like Aftershoot and Narrative process locally, so you trade upload time for however quick your own machine is.

2. Waiting for the machine

Previews get built, frames get scored, near-identical shots get grouped. On a 3,000 frame wedding that is not instant on any tool. The honest way to count it: if you have to sit there, it is your time. If you can walk away and come back, it is calendar time. Both matter, they just matter differently, and only one of them belongs in an hourly rate.

3. Reviewing the shortlist

This is the cost that goes missing most often. AI does not end the reviewing, it changes what you review. You go from making a few thousand judgements to checking a proposed selection. That is genuinely faster, but it is not zero, and it still needs your eyes and your full attention.

4. The disagreements

Every tool in this category will set aside a frame you wanted, and hand you two near-identical ones where you only needed the better one. Sorting that out is where the minutes hide, and it is heaviest in your first month while you learn what the model can and cannot see. How the pipeline works is worth twenty minutes if you want to know why the disagreements cluster where they do.

5. Learning it

Budget two or three shoots of being slower than usual. Everyone is slower at first, everyone forgets to count it, and it is the single most common reason a photographer decides in week one that a tool did nothing for them.

The arithmetic, with your numbers in the blanks

Here is the whole calculation. Five lines, no borrowed figures, because the only numbers that mean anything here are the ones you measure on your own shoots.

Hours saved in a year = ((A minus C) multiplied by D) divided by 60. What that is worth = that number multiplied by E. If line E is where you get stuck, work the other side of the ledger first: our cost calculator shows what your current software stack takes out of the year, which is a concrete number rather than a feeling.

Three rules keep this honest. Time the whole route both ways, or you are comparing a stopwatch against a memory, and memory always flatters the old way. Use the same type of shoot for both, because a 900 frame family session and a 3,400 frame wedding are not the same test. And measure after the learning period rather than during it. If you would rather see this walked through on one job than described in the abstract, we did exactly that in one shoot, both ways.

What actually gets faster, and what does not

The saving is not spread evenly across the job. It sits almost entirely in the mechanical checks: is it sharp, are the eyes open, is this the fourth frame of the same burst. A model does that at frame 3,000 with the same attention it gave frame 30, which is more than any of us can say at midnight.

What does not get faster is judgement. Which of the two good frames tells the story. Whether the shot the couple specifically asked for made it in. Whether your second shooter's frames sit next to yours without a visible seam. That work is the same length it always was, and it should be, because it is the part clients are paying for.

So the honest sentence is this: AI culling takes the mechanical half of the first pass off your plate and leaves the judgement where it belongs. That is worth real money if your first pass is long, and worth very little if you already shoot tight and cull in forty minutes. Which of those two you are is a question with a real answer, and it is worth settling before you pay anyone anything.

There is a second effect that is harder to time and probably larger. Research suggests repeated decisions degrade the quality of later ones, and while parts of that literature are debated, no photographer needs convincing that the last 200 frames of a four hour session get judged differently from the first 200. Handing the mechanical checks to something that never gets tired changes the quality of what you deliver, not only the clock, and that is the part photographers tend to notice in the second month rather than the first.

The saving you actually feel

Ask a photographer what hurts about culling and almost nobody says the total minutes. They say Sunday night. The complaint is rarely that the first pass takes four hours, it is that the four hours cannot begin until you are home, at the desk, with the previews built, at an hour when your judgement is at its worst.

That is a scheduling problem wearing a time-saving costume, and it is why two tools with identical processing speeds can feel completely different to live with. Every established option in this category is a computer program: Aftershoot and Narrative Select are desktop apps, Imagen wants your files uploaded first. All of them assume the work starts when you sit down. We compared the two working styles honestly in phone versus computer culling.

This is the part Kepla is built around, and it is worth being precise about where it stands. The picking app for iPhone, iPad and Mac is still being built. The idea is that the first pass happens on the phone while the day is fresh, and the same shoot opens later on the Mac for the frames that genuinely need two windows and a zoom. Nothing is ever deleted, moved or renamed: Kepla marks what it picked, every original stays where you put it, and anything it did not pick is one tap away. The booking page is live today and free while we build the rest.

A three shoot test you can run this month

Every serious tool here gives you enough free access to answer this question for yourself, which is convenient, because your answer is the only one that applies to you.

Prices checked August 2026 from each vendor's own pricing page.

Run the trials against shoots you have already picked through by hand, so you have something to compare against. Pay attention to the disagreements rather than the agreements: agreement tells you the easy frames are easy, disagreement tells you whether you can trust it on a Sunday when you are tired. When the three shoots are done you will have a real hours-per-year figure, and you can put it against what the manual version is costing you, which we lay out line by line in the true cost of manual culling.

FAQ