Photographers

How to cull photos in Lightroom Classic, fast

By Ayush Patnaik, founder of AccioPix

Published September 4, 2026 · Updated September 29, 2026 · 10 min read

You aren't the reason Lightroom culling is slow. The previews are. Import a card with default settings, and Lightroom Classic builds its own preview of every raw file before you can flip through them at full size. On a 3,000-frame wedding, you just watch a progress bar while your tea goes cold. That's why so many photographers cull in a different programme and only move the keepers into Lightroom.

You do not have to. Lightroom Classic culls fast once you fix three things: the right preview option on import, six keyboard shortcuts, and a two-pass habit that stops you comparing and rejecting at once. This guide covers those. It then shows how AI culling tools bring picks back into your catalog, because that part trips people up.

Step 1: import with Embedded & Sidecar previews

Every raw file already contains a JPEG. The camera made it for the back screen, and on most current bodies it's full size. Lightroom Classic can use that JPEG instead of rendering its own. In the Import dialog, open File Handling and set Build Previews to Embedded & Sidecar, which Adobe's import documentation describes as the option to "Display the largest possible preview available from the camera".

The difference is huge. Photo Taco timed the import of 1,000 raw files: 27 minutes 35 seconds with Standard previews, and 3 minutes 7 seconds with Embedded & Sidecar. They used the same files on the same machine. The embedded route is fast because Lightroom does almost no work. It only reads a JPEG that already exists.

Two settings keep things fast after an import. Under Preferences > General, clear "Replace embedded previews with standard previews during idle time". Without that, Lightroom rebuilds previews in the background and your laptop fans begin to whirr. Do your culling only in the Library module. Opening a picture in Develop forces Lightroom to render the raw, which works for one frame but drags across three hundred.

There is one catch. The embedded JPEG shows your camera's picture style, with its contrast, saturation, and in-camera noise reduction. When Lightroom finally renders the raw file, colours and shadows will look a bit different. Use the embedded preview to cull for focus, blinks, expression, and framing. Do not cull for colour. That is an editing choice anyway.

Step 2: the six keys that do the whole job

You do Lightroom Classic culling on the keyboard. Keep the mouse for the rare frame you need to inspect. Here are the keys, the same on Mac and Windows unless noted:

Adobe's Julieanne Kost keeps a complete list of flag, star and label shortcuts if you want the long version, including the Painter tool for spraying a label across a selection. The seven rows above are the ones a cull actually uses.

Step 3: reject first, in Loupe, with Caps Lock on

Hit E for Loupe, turn on Caps Lock, and begin. There is only one question on this pass: "is this frame obviously unusable?" The floor, a blink you can spot at fit-to-screen size, a frame where the flash failed, the fourth identical picture of the signage. Press X and Lightroom moves to the next shot. If the frame is fine, press the right arrow and leave it alone. Don't rate. Don't compare. Don't press Z unless focus is genuinely in doubt. Zooming is where the minutes go.

Spend roughly a second on a frame, and 3,000 frames takes 50 minutes. In the end, you reject a third of them. Photographers hate this pass. An AI culler does it best, and we'll get to that.

Step 4: pick one winner per burst with Compare and Survey

Go back to Grid (G). Work moment by moment now. Select the surviving frames of a burst, say the six frames of the garland exchange, and press N for Survey. All six sit side by side. Press Z to check the eyes at 1:1 on any of them. When you find the winner, press P. The other five do not need an X; the next step handles them.

Compare (C) is faster than Survey when two frames compete. It puts a Select on the left and a Candidate on the right. Arrow keys step the Candidate through the rest of the burst. When the Candidate wins, make it the Select and move on. Compare fits a bride's portrait series where every shot is close and one eyelid decides it.

Take shortcuts where you can. If a burst shows a clear winner in the Grid thumbnails, you can skip Survey. Group photos are different. Never pick a group portrait from a thumbnail. Someone in the back row always blinks.

Step 5: Refine Photos, then delete the rejects

Your catalogue now has picks for winners, rejects for pass-one failures, and unflagged files for the rest. Library > Refine Photos (Cmd+Opt+R on Mac, Ctrl+Alt+R on Windows) handles this in a single step. It turns every unflagged image into a reject, and every pick into an unflagged frame. The burst losers you never bothered to X become rejects. Your keepers are now clean, unflagged frames, ready to rate.

Then Photo > Delete Rejected Photos (Cmd+Delete on Mac, Ctrl+Backspace on Windows). Lightroom asks whether to remove the photos from the catalog or delete them from disk. Choose Remove. Keep the raws on the drive until the gallery is delivered and the client has stopped asking for "that one photo of the cake". Disk is cheap; the frame you deleted in week one and got asked for in week three is not.

Stars and labels without slowing down

The mistake is using stars during the cull. A cull needs two states, keep and reject, and flags already give you those. Stars are for the pass after: 5 for album candidates, 3 for the standard delivery, nothing for the frames you're keeping in the catalog but not sending. Colour labels are for status rather than quality: red for "needs retouching", green for "sent to the client", whatever your studio agrees on. Because stars and labels are numbers on the keyboard and Auto Advance applies to them too, rating 700 keepers is a ten-minute job once the cull is done.

To make bulk changes, use the Painter tool (Cmd+Opt+K on Mac, Ctrl+Alt+K on Windows). Click any thumbnail to give it a rating or label. The Library Filter bar toggles with the backslash key. It lets you isolate five-star frames or hide rejects while you work.

AI culling and Lightroom: how the picks get back in

Lightroom Classic does not cull with AI on its own. Cloud-based Lightroom has a "Best Photos" tool that selects from an album, but Classic does not. Tools like AfterShoot, Narrative Select, FilterPixel and Imagen run before Lightroom. The thing that matters is how their decisions enter your catalog.

The answer, for almost all of them, is XMP. When the tool rates a frame, it writes an .xmp sidecar file next to the raw file with the star rating and colour label inside. Import the folder afterwards and Lightroom reads the sidecars as part of the import. If the folder was already in your catalog, select everything and choose Metadata > Read Metadata from File, and the ratings appear. Narrative's help pages spell out the same two steps, and their metadata sync article is the clearest description of it.

Flags are the odd one out. For years pick and reject flags lived only in the catalog and never touched XMP, which is why every culling tool speaks in stars and labels instead. Since Lightroom Classic 13.2 in February 2024, flags are read from and written to XMP with the other metadata. Lightroom Queen's release notes note that this exists mainly for sync with the cloud-based Lightroom and that other apps such as Bridge ignore the field. In practice, stars and labels remain the language the tools use.

JPEGs are another trap. Lightroom Classic looks at sidecar files only for proprietary raw formats. With JPEG, TIFF, PSD and DNG, it wants metadata inside the file itself. It ignores any sidecar sitting next to a JPEG, as an Adobe community expert explains here. If you shoot JPEG and cull somewhere else, your tool must write into the file, or you must move your choices over as a list.

A free pre-cull that hands off to Lightroom

Our Photo Culler does the mechanical work of culling in your browser, before Lightroom opens the folder. Drop in the raw files. The tool reads the embedded previews, the same JPEGs Lightroom uses for Embedded & Sidecar. It finds faces and checks their sharpness at full resolution. It tests for closed eyes, expression, and skin exposure. It puts frames into bursts by capture time, rejects the bad shots with a reason, and picks the best of each burst. Nothing uploads.

Its export is built for the workflow above. Alongside a plain keepers list and a CSV with the reason for every frame, it writes an .xmp sidecar for every raw: picks at five stars, other keepers three, stacked duplicates two, rejects one. Put the sidecars in the folder with the raws (in Chrome or Edge it can write them there itself), import with Embedded & Sidecar previews, and filter the Library by stars. Your pass one is done and most of pass two with it; what's left is the human check over the picks, which is the part worth your time. If you'd rather only strip soft frames and decide the rest yourself, the Blur Culler does that one job with a threshold slider. There's more on what the scoring can and can't judge in what photo culling is and how it's done.

A note for event photographers

A cull that takes an hour instead of an evening is only worth having if the gallery leaves your studio quickly too. That's the part AccioPix handles. You upload the finished gallery once, guests send a selfie on WhatsApp, and each of them gets only the photos they're in. No app to install, no password page, no forwarding the family group's requests one by one.

It's credit-based rather than a subscription: a 500-photo event costs about ₹100, and credits don't expire, so a slow month costs nothing (details on the pricing page). The photographers who get the most out of it treat it as the last step of the workflow: cull, edit, upload, and let the guests fetch their own photos while you're already on the next job. How to grow your photography business covers the referral side of that.

Frequently asked questions

Why is culling in Lightroom Classic so slow?

The default import builds Standard previews for each raw file before you can browse them. When you pick Embedded & Sidecar previews on import, it reads the JPEG your camera already put inside each raw. In one published test, that reduced import time for 1,000 raws from about 27 minutes to about 3.

Does Lightroom Classic have AI culling?

No. Lightroom Classic has flags, stars, labels and fast views but no automatic culling. The cloud-based Lightroom has a Best Photos feature for albums. AI culling tools run before Lightroom Classic and hand their picks over as star ratings and colour labels in XMP, or through a plugin.

Are pick and reject flags saved to XMP?

Since Lightroom Classic 13.2 (February 2024) flags are read from and written to XMP along with other metadata, mainly so they sync with the cloud-based Lightroom. Bridge and most third-party tools ignore the flag field, which is why culling tools still exchange decisions as star ratings and colour labels.

Can Lightroom read XMP sidecar files next to JPEGs?

No. Lightroom Classic reads .xmp sidecar files only for proprietary raw formats. For JPEG, TIFF, PSD and DNG it expects the metadata to be embedded inside the file and ignores a sidecar sitting next to it.

What is the difference between rejecting and deleting in Lightroom?

A reject flag is only a mark in the catalog. The file is untouched. Photo > Delete Rejected Photos reads those marks. It asks to drop the files from the catalog alone or delete them from disk. Dropping them from the catalog is the safe choice until the gallery is delivered.

References

• Adobe, How to specify import options in Lightroom Classic. Read 29 September 2026.
• Kost, J., June 2024. Applying flags, stars and color labels in Lightroom Classic.
• Photo Taco podcast, Slash the import time of Lightroom Classic by 90%.
• The Lightroom Queen, What's new in Lightroom, February 2024.
• Narrative help centre, Syncing metadata between Lightroom Classic and Narrative.
• Adobe Community, Importing a JPG with metadata in an XMP sidecar. All links checked 29 September 2026.