Lightroom Assisted Culling vs Kepla: an honest comparison

The awkward part first. Kepla's photo picking app is still being built, so you cannot download it today. Adobe's Assisted Culling is sitting inside Lightroom right now, on software you probably already pay for. That makes a speed test impossible to write honestly, so this is the other thing: what Adobe ships, who it suits, and how to judge us later.

What this comparison can and cannot be

Kepla's photo picking app is still being built. It is not in the App Store, you cannot install it tonight, and we are not going to write around that. Adobe's Assisted Culling, meanwhile, is already in Lightroom Classic and in cloud Lightroom, labeled early access, and it will score a folder for you this afternoon.

So there is no test to report. We have not run our app against Lightroom on a wedding folder, and if we had, the result would be a number produced on a machine we control, on shoots we chose, with a keeper standard we set. That is a press release, not evidence. Any page showing minutes per thousand frames for an unreleased app is describing its own ambitions.

What this page can be instead: an account of what Adobe genuinely ships today, a plain recommendation to use it if it fits your work, a description of the different bet we are making and why we think it is worth making, a list of the claims we are not allowed to make yet, and the test we want you to run on us when we ship. We wrote the same page about the incumbent culling app in Kepla vs Aftershoot, and the shape is deliberate. The one part of Kepla that works today is the free booking page, which holds a date and takes a deposit. Different job, same app, and we will come back to why.

What Assisted Culling actually does in Lightroom

If you last looked at this two years ago, the picture has changed and a lot of comparison pages have not caught up. Lightroom used to be a place where you flagged everything by hand. It is not any more.

Assisted Culling runs in Lightroom Classic and in cloud Lightroom, and Adobe currently labels it early access. In Classic you switch it on at Catalog Settings, then Metadata, then Assisted Culling, or from the Auto Analysis option in the identity plate menu. It scores subject focus, eye focus and whether eyes are open, and it flags the obvious failures: exposure problems, accidental captures, and the documents and receipts that end up on a card. Adobe's own documentation describes it as identifying the best images and rejecting low quality ones, then letting you apply flags, ratings and labels in batches from that analysis. Since version 15.4 in June 2026 it can also run as photos come in at import rather than only afterwards.

Two consequences worth stating plainly. First, the old line that Lightroom cannot see blur or closed eyes is now false, and you should distrust any page that still says it. Second, this is included with the plan you already pay for, which is a strong position that no separate tool can match on price. We go through the settings in how to set up Assisted Culling and give it a full assessment in our Assisted Culling review, so this page will not repeat either.

The philosophy is what matters here. Assisted Culling proposes and you audit, inside the catalog, with the whole sequence still in front of you and the develop controls one key away. When two frames are close you can check crop room and pull the exposure a stop before you decide. That is a real advantage, and it only exists because the analysis happens where the editing happens.

What each one costs right now

One side of this table is easy to fill in and one side is not, which is itself the honest answer to the question most people arrive here with.

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

For most photographers reading this, the Adobe line is already spent. You bought the plan for Photoshop or for the catalog, and the culling feature arrived inside it. That means the interesting cost is not the subscription at all, it is the evening. What the first pass takes out of your week, and whether a second tool could ever earn that back, is arithmetic you have to run on your own jobs, and we set out how in the cost of culling inside Lightroom.

Who should just use Lightroom this month

Plenty of you should, and saying so costs us nothing, because we do not have an app to sell you yet.

• You already pay for it. A feature included in a plan you hold beats a $120 subscription that does a similar job, unless the difference is large enough to notice on every shoot.
• You shoot portraits and headshots. That is the work Adobe has aimed the feature at, and eye focus and eyes open are exactly the checks that matter on a headshot card.
• Your choices depend on edit potential. If picking between two close frames usually means checking crop room, highlight recovery or how a skin tone will land, you want to be inside the catalog when you decide, not in a separate app that cannot show you.
• You distrust a preselection boundary. Some photographers want the full sequence visible, with filters they can switch on and off, and no earlier stage that has already narrowed the field. That is a reasonable preference and Lightroom serves it well.

The honest exceptions are about volume and about where you are when the work starts. A 4,000 frame wedding has to be imported before Lightroom can show you anything, and on a laptop that is sharing itself with a backup, that wait is the job. If your version of slow is the machine rather than the decision, most of it is fixable in settings. If your version of slow is that Sunday night is the first time you sit down, no setting fixes that.

The different bet we are making

If Adobe now ships assisted selection inside the app you already own, why build anything at all?

Because the whole category has agreed on a constraint that is not a law of physics. Lightroom Classic is a desktop application. So is Aftershoot. So is Narrative Select. Imagen wants your files in its cloud before it starts. Every one of them assumes the work begins when you are home, at the desk, with cards ingested and previews built. For a lot of photographers the slow part was never the analysis. The slow part is that the desk is three days away.

Our bet is about where the work happens, not about being faster at the same desk.

• The first pass on the phone. iPhone and iPad, in the car park, on the train, on the couch. Not a companion app to a desktop program, the actual first pass. Whether that idea holds up is a fair argument, and we put both sides of it in phone versus computer culling.
• Then the Mac when the shoot needs it. A phone is not where you compare two near identical frames at full size. The same shoot opens on the Mac and you carry on: side by side, zoomed in, nothing redone.
• Nothing is ever deleted. Kepla marks which photos it picked, and that is the only thing it does to your files. It never deletes, moves or renames anything. Every original stays where you put it, and every frame it did not pick is one tap away.
• It also books the job. The booking page is live and free today. No culling tool sends a client a page, holds a date or takes a deposit, because that is not what any of them are for.

That last point looks strange in a culling comparison, and it is deliberate. Adobe is building a photography suite. We are trying to build the thing that carries a booking through to a delivered gallery, with the picking stage happening in the gaps of your day rather than at the end of it. Whether that is the better idea is not settled by us asserting it here.

What we are not allowed to claim yet

Every unreleased product is tempting to describe in the present tense. Here is the list of things we will not say until you can run them yourself and check.

• No speed claim. No minutes per thousand frames, no chart against Lightroom, no before and after. There is no test, and a timing from our hardware would not be evidence about your wedding on yours.
• No accuracy claim. No percentage, ours or anyone else's. Accuracy in picking is not one number anyway. It depends on your work, your taste, and what you personally count as a miss.
• No maturity claim. Adobe has decades of odd folder structures, unusual RAW formats and enormous catalogs behind it. We have none of that yet and there is no way to fake it.
• No Lightroom integration claim. It is planned. Planned is not shipped, and we will describe it as planned until the day it is not. If you need selections to land in a catalog today, getting selections into Lightroom covers the routes that exist now.
• No price claim. The picking app is not priced. Founding members get it the day it ships, free while we build, and that is the whole of what we can say.

Apply the same filter to everyone, including the pages that outrank this one. A vendor publishing its own accuracy scores against named rivals is running marketing, not a study. FilterPixel does exactly this, so read its figures for other companies' products as its own material. If a page cannot tell you how a number was produced, treat it as a claim.

How to judge us when we ship

Here is the test we would like you to run, written down now so we cannot move the goalposts later. It has a control built in: you already have Lightroom, so you can run both on the same folder.

1. Bring your own shoot. A wedding or a session you already picked by hand months ago, so there is a known answer to compare against. Not a demo folder, not a sample set.
2. Run Assisted Culling on the same folder. Same files, same standard, on the machine you actually own. That is your baseline, and it is a fair one, because it is free to you.
3. Count the two kinds of miss separately. Frames it did not pick that you would have kept, and frames it picked that you would have left out. The first kind costs you a moment from someone's wedding day. The second kind costs you a few seconds. Never let one number hide both. The full method is in how to test culling accuracy.
4. Time the whole job, not the analysis. From card in the reader to a shortlist you actually agree with. A fast analysis that still eats your Sunday evening has changed nothing.
5. Test the handoff. Start on the phone, finish on the Mac, and check the second screen picks up exactly where you stopped. If any work has to be done twice, the whole premise fails.
6. Try to break the promise. Go looking for a frame it did not pick. If it is not still sitting there, untouched and one tap away, we got the most important thing wrong.

If we fail that test, do not use it, and say so in public. Until then the fair recommendation is the one above: if Assisted Culling suits your work, it is included in a plan you already hold, and there is nothing clever about paying twice for a job that is already handled.

FAQ