(Early Access) Assisted Culling (Lightroom Ecosystem - Desktop)

Great concept with real potential, but the current execution has some friction worth addressing.

First, focus detection feels inconsistent. Rather than applying a fixed standard, it'd be more useful to weight sharpness relative to the best image in the set — then calibrate from there.

Second, the bigger issue: the feature re-analyzes images mid-session. If it decides to re-evaluate while you're previewing one-by-one, it bumps you back to the start of the gallery. When that happens every 20–30 seconds, the workflow breaks down entirely.

Hoping to see these ironed out — the thinking behind it is solid.

It seems very broken for me. This image was flagged as a reject for both subject and eye focus both set to 70%.

I tested in the Feb 2026 update and get identical results in Lightroom Desktop and Classic.

 

and these images were rated as follows:

Left - Subject 62, Eye 42
Right - Subject 98, Eye 98

Something very wrong there!

• App Version 9.1  
• Apple MacBook Pro M1 Max (2021) 64GB
• 2,090 images of volcano photography Sony RAW files (ARW)

The issue I had was that, yesterday, when I tried culling for the first time using only the 100% Sharpness parameter box checked, it instantly sorted all files within a few minutes into about ½ and ½ Selects versus Rejects. I had also tried to copy paste a grade from one photo to the other 2,170 photos (yes there were more, but today only 2,090 images would load) but the grade somehow looked messed up by leaving blue-dot artifacts in the darker areas of most of the other 2,169 images, so I had deleted all these from my LR.  

Today however, I tried again by importing all images (like I said, only 2090 of the original 2170 loaded) and tried culling based on 100% sharpness again. This time, it took almost 3 hours and no accumulating results showing in neither the Selects or the Rejects image count, and after three hours, I had a 0 Selects 0 Rejects result. It gave me some sort of message to the effect of (Uh oh, a black hole….” but I can’t see the exact wording anymore, because as I write this, it has re-started the 3-hour process of analyzing all 2,090 image s again.

What is going on, or what am I doing wrong (did I corrupt my files from yesterday?) and why wont assisted culling give me results today?

Worked reasonably well on photos of squirrels. However, it didn’t seem to work nearly as well on macro photos of plants.

 i MADE THE MISTAKE OF ANALYSING MY ENTIRE PORTFOLIO WHICH TOOK 7 HOURS, SILLY ME! THE ONE THING I DID NOTICE IS MANY SHOTS WITH A BLURRED  FOREGROUND OR BACKGROUND EITHER SHOT THAT WAY OR PUT IN VIA AI WERE REJECTED BUT MAYBE I DID NOT PUT IN THE REIGHT SETTINGS? HOWEVER, I THINK IT IS A GREAT TOOL AND I WILL TRY AGAIN. MAYBE A CANCEL BUTTON WOULD BE A GREAT ADDITION ONCE ONE REALISES THEIR ERROR?

Feature has worked well so far, but it is incredibly frustrating that the system doesn’t score culling scores or attributes against XML files for RAW photos. It’s pretty brutal to have the system re-evaluate every single photo if I want to use the Subject Focus scores to filter for my sharpest pictures and apply a 5-star rating to them after an initial cull of slightly lower standards. Additionally, every time I browse through photos after the initial cull is finished, the system starts re-analysing again from that point. This latter issue has caused at least two crashes and many instances of lag/freezing.

From both a usability and general logic standpoint, this doesn’t make sense. If a photo has sharp focus, it wouldn’t really change. Same with photos that have AF eye focus or open/closed eyes. A default behaviour of one single analysis of each photo, with the ability to manually re-evaluate would make much more sense. It would also make the app far more efficient and avoid crashing or lag.

So far so good. Culling several years of family photos (scaned and digital) of people from “art photos”. 

Though it runs very slowly ( 12k out of 30K of pictures after 3 days runing time on an older 2017 i7 laptop) but accurately and way faster then me!  

Can not wait for image analysis AI for other content to be availble. Perhaps even a tunable asthetics/composition function?

Thanks

I have been using the AI Culling Assistance in Lightroom, and while it is a great start, there are a few key features missing that would make it much more powerful for wildlife and bird photographers.

When shooting bursts of high-speed action, I would like to suggest the following improvements:

1. Persistent and Sortable Metadata Scores Currently, the AI scores aren't easily saved or used for organization. It would be a huge improvement if the AI "Quality Score" was saved into the metadata. This would allow us to Sort by Score, making it much faster to identify the best images from a burst and decide which ones to keep or delete.

2. Comparative Scoring for Selected Sequences When reviewing a burst of images, I often need to find the "sharpest of the bunch." I would like to see a feature where I can select a group of photos and have the AI identify which one has the highest sharpness score relative to the others in that specific selection.

3. Balancing Sharpness with Composition (The "Memory" Factor) Sometimes, an entire burst sequence might not reach a "perfect" sharpness score, but a specific frame captures a unique posture or a rare moment. It would be great if the AI could help highlight the "best of a bad batch" so I can ensure I keep at least one or two memory-worthy shots from a sequence, even if they aren't technically perfect.

I hope the Adobe team considers these additions to make the culling workflow more efficient for those of us dealing with thousands of high-speed action shots.

I have been using the AI Culling Assistance in Lightroom, and while it is a great start, there are a few key features missing that would make it much more powerful for wildlife and bird photographers.

When shooting bursts of high-speed action, I would like to suggest the following improvements:

1. Persistent and Sortable Metadata Scores Currently, the AI scores aren't easily saved or used for organization. It would be a huge improvement if the AI "Quality Score" was saved into the metadata. This would allow us to Sort by Score, making it much faster to identify the best images from a burst and decide which ones to keep or delete.

2. Comparative Scoring for Selected Sequences When reviewing a burst of images, I often need to find the "sharpest of the bunch." I would like to see a feature where I can select a group of photos and have the AI identify which one has the highest sharpness score relative to the others in that specific selection.

3. Balancing Sharpness with Composition (The "Memory" Factor) Sometimes, an entire burst sequence might not reach a "perfect" sharpness score, but a specific frame captures a unique posture or a rare moment. It would be great if the AI could help highlight the "best of a bad batch" so I can ensure I keep at least one or two memory-worthy shots from a sequence, even if they aren't technically perfect.

I hope the Adobe team considers these additions to make the culling workflow more efficient for those of us dealing with thousands of high-speed action shots.

Great in theory… it would be great if there was the ability to stop it form analyzing while still seeing the results. It’s difficult to edit because when you start scrolling, it starts to re-analyze everything and then kicks back to the beginning.