Skip to content

Culling

AI culling vs a human cull: where each one breaks on a wedding

The AI kept 1,500 frames and you still had to check them all? What AI culling does well, where photographers report problems, and when a human cull is worth it.

Published 22 October 20265 min readFor European studios

Photo: Bride standing with her bridesmaids in sage dresses

You ran the AI cull on a 4,000-frame wedding, it came back in minutes, and you felt good about it for about ten seconds. Then you saw it had kept 1,500 images. Eight versions of the same kiss. A soft ring shot rated five stars. And somehow it had dropped the frame where the groom’s dad finally cried.

So you went through all of them anyway.

If that sounds familiar, this post is for you. It isn’t an argument that AI culling is bad. It’s a look at what it does well, where photographers say it struggles on a wedding, and where a human cull earns its cost.

What AI culling does well

AI culling tools have real strengths, and plenty of photographers use them every weekend.

  • Speed on the obvious. Blinks, missed focus on a static subject, frames where someone walked in front of the lens. A machine finds these fast.
  • Grouping. Similar frames get stacked together, so a burst of twelve looks like one decision instead of twelve.
  • Consistency. It applies the same rules to frame 40 and frame 3,940. You don’t get tired at the reception and start keeping everything.
  • A head start. Even when you review every pick, starting from a sorted set is faster than starting from a raw card.

If your weddings are fairly consistent, your light is good and you’re comfortable reviewing the picks, AI culling may be all you need.

Where photographers report problems

The complaints that come up most often are about judgement, not speed. Reviews and photographers’ reports of AI culling tools describe the same few problems:

  • Over-selection. The tool keeps far more frames than the photographer would deliver, so the “cull” still needs a second pass.
  • Bursts. Eight or more near-identical frames kept from the same sequence.
  • Soft focus rated highly. Frames that look sharp at thumbnail size but are soft on the eyes.
  • Missed expressions. Subtle moments, a glance or a half smile, passed over for technically cleaner frames.

None of these are surprising. A machine can measure whether eyes are open. It’s much harder for it to know that the slightly blurred frame of a grandmother laughing is the one the family will print.

The result, for a lot of photographers, is that the AI cull saves the first hour and then hands back a second cull. You still look at every keeper. The work moves rather than disappears.

The training question

Many AI editing and culling tools learn your personal style from your past work. That usually means feeding them thousands of your previously edited images before the results feel like yours. Check the requirement on the tool you’re considering, because it varies.

That’s fine if you’ve been shooting for years and have a deep catalogue. It’s a real hurdle if:

  • you’re new and don’t have thousands of finished frames yet,
  • you’ve recently changed your style, or
  • your weddings vary a lot, so past edits don’t predict the next one.

A human cull doesn’t need a training set. It needs a conversation: what matters to you, what you always keep, what you never deliver.

Culling criteria that actually matter

Whether you cull yourself, review an AI’s picks or hand the cull to someone else, these are the decisions that shape a gallery.

Keep:

  • The sharpest frame of each moment, judged on the eyes, not the whole image.
  • The best expression in every burst, even if it’s not the most technically perfect.
  • Every person in the family groups, with everyone’s eyes open.
  • One or two frames of each important detail: rings, dress, flowers, stationery.
  • Candid reactions from guests during the ceremony and speeches.

Drop:

  • Near-duplicates once you’ve chosen the best one.
  • Frames where someone important is mid-blink or mid-word in an unflattering way.
  • Test shots, lighting checks and accidental frames.
  • Anything you’d be uncomfortable seeing printed in the album.

Ask yourself every time: would the couple miss this frame if it weren’t there? If not, it probably doesn’t belong.

When a human cull is worth paying for

A person is slower than a machine, but there are weddings where that slower judgement saves you more time overall.

Situation Why it’s hard to automate
Large family groups Every face has to be checked across every frame. One closed eye ruins the shot.
Low light receptions Grain, motion blur and focus misses look similar at a glance, so it takes care to tell a keeper from a near-miss.
Emotional moments The best frame is the one with the feeling, not the one with the cleanest exposure.
Multi-event days Knowing which moments families expect to see from each event, and how many frames each deserves, needs context.
A specific target count Cutting to exactly what you deliver means weighing every moment against every other.

Using both

You don’t have to pick a side. Plenty of photographers use AI as a first pass and a person, themselves or an editor, for the final decisions. A few ways to combine them:

  1. AI first, human final. Let the software remove obvious rejects, then review what’s left with your own criteria.
  2. Human cull for the hard days. Use AI on straightforward weddings and hand off the big family-group days and dark receptions.
  3. Send selects. Cull however you like, then send only the selects to an editor for grading.

How BackstagePost handles the cull

Our Gallery packages include a human cull. Send the full card and we cull to the count in your package: 300 to 500 images from up to 3,000 raws for a Wedding gallery, or 100 to 200 from up to 1,000 for an Engagement gallery. We look for the best expression, sharp focus and no near-duplicates, then grade and retouch the keepers to your look.

If you want a lower count, or you always keep every family combination, say so in the brief. If you’d rather cull yourself, send your selects and we grade only those.

There’s no training set. Three to five finished frames, your preset and a line about what matters is enough.

Your next step

Pick a wedding where the cull is the part you dread, such as one with a big family-formals session or a dark reception, and send it as your free trial. Compare our selection to what you’d have kept.

What we cull and grade is set out on what we edit, and you can see a mixed set of ceremonies and receptions in our wedding day selects gallery. Prices are on the packages page, and the stage-by-stage process is on how it works.

For the bigger picture on where the hours go, read how long it takes to edit wedding photos.

Questions photographers ask

More in the FAQ.

Is AI culling good enough for wedding photos?

For many photographers it is a useful first pass: it removes obvious rejects quickly and groups similar frames. Where photographers report problems is in the judgement calls, such as choosing the best frame in a burst, catching soft focus and keeping subtle expressions. Most still review the AI's picks before grading.

How many photos should a wedding gallery have after culling?

There is no single right number. It depends on the length of the day and your style. Our Wedding gallery culls to 300 to 500 images from up to 3,000 raws, and you can ask for a lower target in the brief.

Do you need my past edits to train on before you cull?

No. A human cull needs no training set. Send three to five finished frames, your preset and a line on what matters to you, such as keeping every family group or putting candid moments first.

Can I use AI culling and then send you the selects?

Yes. Send your selects instead of the full card and we grade only those. Tell us in the brief if you want a second look at the rejects.

The edit team your clients never see.

Your name on the gallery. Your look on every frame. We stay backstage.

Your first project is free. Prefer email? hello@backstagepost.com

Start your free project

Or send a paid project.