Culling product photography: precision over speed
Almost everything written about picking photos assumes you shot a wedding: thousands of frames, a tired photographer, a deadline. Product work is the opposite problem. The folder is small, the deadline is usually fine, and the thing that will hurt you is picking a frame that is very slightly wrong and not noticing until it sits on a catalogue page next to three hundred others.
Product work is checked against a spec, not felt
A wedding day produces somewhere in the range of 2,000 to 4,000 RAW frames and asks you to hand back 400 to 800 you love. The whole craft there is judgment under volume. You are choosing between many good moments, and if you pick the second best frame of the first dance, nobody on earth will know.
A product shoot produces a folder that would fit inside one hour of a wedding. Forty products, six angles each, a few frames per angle. Perhaps five hundred files. And every file is a candidate for a slot that already has rules attached: this size in frame, this shadow, this white, this side of the label facing camera, because the client has three hundred other product photos and yours has to sit among them without anyone noticing the join.
So the question at the desk changes. It stops being which frame do I like and becomes which frame is correct. That is a checking job, not a taste job, and checking rewards care rather than pace. If you have only ever picked from event or portrait folders, the habit to unlearn is the sweep: skimming for the frames that jump out. On a product table nothing jumps out. Two frames of the same bottle look identical at thumbnail size and one has a fingerprint on the shoulder of the glass.
The unit of work is the SKU, not the shoot
Do not work through a product folder front to back. Work product by product, the same way a headshot day is really forty small shoots stapled together rather than one big one. A catalogue shoot is forty small shoots too, and each has its own list of required frames.
Before you open anything, get the list from the client and write it as a column: front, three quarter, back, label detail, scale shot, lifestyle. Then for each product you answer one question per row, and the pass is finished when there are no empty cells.
- Split by product first. If you tethered, your capture folders already did this. If you shot to card, put a separator frame between products, your hand over the lens or a card with the SKU on it. Ten seconds on set saves twenty minutes of squinting later.
- Fill the required angles before you have an opinion. One frame per row, marked. A gap here is the only true emergency in product work, because a missing angle means the sample has to come back.
- Then pick the hero. The one that carries the listing, the one that gets retouched properly. Usually the three quarter. Mark it differently so the retoucher can see at a glance where the hours go.
- Keep an alternate for anything with a reflection. Chrome, glass, glossy black plastic. These are the frames that fall apart when someone looks at them at full size on a Tuesday.
Working this way also tells you when to stop shooting. If the checklist for a product is full and the hero is marked while the sample is still on the sweep, you are done with it. Writing criteria down before you look at files is the habit that makes any genre go faster, and we made the general case in building your own culling criteria.
Focus stacks and brackets are picked as a set, never as a frame
This is where product photography breaks most selection software, and it breaks it quietly.
A stack of twenty slices at f/8, each focused a millimeter further into the watch, is not a burst. There is no best frame in it. Slice one is sharp at the crown and soft everywhere else, and it is supposed to be. The set is the deliverable. The same is true of an exposure bracket for a reflective surface, or a light painting sequence where each frame lights one part of the product.
Two things follow. A tool that scores frames on sharpness will rank the middle of a stack highly and the ends of it poorly, which is a confident answer to a question you did not ask. And near duplicate grouping sees twenty nearly identical frames and offers to reduce them to one, because that is exactly what it was built to do on a burst of a bride laughing. Accept that offer and you lose the stack. Those behaviors are genuinely useful elsewhere, as how duplicate detection actually works explains.
What you check in a stack is different from what you check in a single frame:
- Did every slice arrive? Tethered captures drop frames. Count them against what the rail or the camera says it shot.
- Did anything move between slices? A bumped tripod, a settling product, a shifting reflection. One drifted slice ruins the blend, and you see it only after twenty minutes of processing.
- Does the stack cover the whole product? Front edge to back edge. One that stops short is one you shoot again.
- Is exposure steady? Flicker from continuous lights becomes a brightness ripple through the blend.
Keep stacks together and mark them as sets rather than as candidates. Give them their own color label so no later pass treats them as loose frames.
Matching across the catalogue beats the best single frame
Here is the failure that separates a first catalogue job from a fifth. Every product got its strongest frame. Every frame is sharp, clean and well lit. The set still looks wrong on the shop page, because the kettle is bigger in its frame than the toaster, two products cast shadows to the left and the rest to the right, and one white sweep is a touch cooler than the others.
Individually those are fine photographs. The client is not buying photographs, they are buying a grid. So the check that matters happens with everything on screen at once, small, exactly the way a customer will scroll it.
| What has to match | How it breaks | Where you catch it |
|---|---|---|
| Size in frame | Products shot at their own comfortable distance, so a small item fills the frame and a large one floats | Grid view, comparing thumbnails side by side |
| Shadow direction and length | A light moved between setups, or someone flagged one side on the third product | Grid view, and it is obvious once you see it |
| White point of the sweep | Ambient light changed, or a colored product bounced onto the background | At full size, against a neighboring frame |
| Label and logo orientation | One bottle turned a few degrees more than the rest | Per product, before you leave the SKU |
| Horizon of the surface | Camera height drifted, so some products sit above the line and some below | Grid view, worst on anything with a visible base |
| Cleanliness | Dust, lint, a fingerprint, a scuff that was only on this sample | Full size, one hundred percent, every hero |
Leave that grid open while you work, because every swap changes the set. It is the standard the whole selection is measured against, not a step at the end. The same truth runs through picking for a cohesive body of work: the frame that wins alone often does not belong in the set.
Why precision beats speed here, and only here
Most arguments for fast selection are really arguments about volume. A UK survey of more than 300 wedding photographers, reported by PetaPixel, put culling at 11 percent of working time and editing at 55 percent, against 4 percent actually shooting. When you are moving thousands of frames a week, shaving that 11 percent is real money.
Product work does not have that shape. Five hundred frames is an hour of clicking even at a leisurely pace, so the time you could save by going faster is small. The cost of being wrong, on the other hand, is enormous, because it is not measured in photographs.
A mis-picked wedding frame means a slightly weaker gallery. A mis-picked product frame means the sample has gone back to the client's warehouse, the seasonal line has shipped, and the only fix is a reshoot somebody has to pay for. Add the downstream work: that frame was retouched, clipped, exported at four sizes and uploaded to a store. All of it repeats.
Which is why the useful upgrade to a product workflow is almost never a faster tool. It is a second look at full size on the frames that matter, and a written definition of correct to check against instead of remembering. We collected the numbers for higher volume genres in culling time benchmarks, and the honest summary for product work is that they do not apply to you.
What automatic picking can and cannot see on a still life
Automatic selection tools are trained on people. That is where the demand is, but it sets the limit of what they can do on a sweep table. Adobe says plainly that the Assisted Culling now shipping in Lightroom is optimized for portraits and headshots, and it scores subject focus, eye focus and whether eyes are open. A bottle has none of those.
What still helps. Blur and shake detection is genre neutral, so it will catch the frame where the tripod was knocked or the shutter fired mid vibration. Basic exposure flagging catches the misfire where the strobe did not recycle. Both are worth having, and both are things you would rather not check by eye four hundred times.
What does not transfer. Anything face based is inert. Expression scoring, blink detection and subject grouping have nothing to work with. And the feature that saves the most time on a wedding, near duplicate grouping, is the one to be most careful with, for the reasons in the stacking section above.
What no tool can judge. Whether the label is turned the way the brand guidelines say. Whether the shadow matches the other three hundred photos on the site. Whether that speck is dust or a genuine surface feature you must not retouch away. Those decisions are yours.
For that reason many product photographers get more from a fast browser than from an AI tool. Photo Mechanic is $14.99 a month, $149 a year, or $299 for a perpetual license, and it makes no picks at all: it just makes you much faster at looking. FastRawViewer is $23.99 once for two computers and does a similar job for less. If you do want AI scoring, Aftershoot and Narrative Select both start at $10 a month on annual billing with a free trial, which is the only honest way to find out whether their scoring means anything on your subject matter.
Prices checked August 2026 from each vendor's own pricing page.
Name the picks so the next person does not have to guess
Product selections are rarely delivered to the person who chose them. They go to a retoucher, then to whoever loads the store. Both of those people work from file names, not from your memory of the day, so the naming is part of the pick.
Use the client's SKU, then the angle, then the state. Something like SKU4471-front-hero and SKU4471-detail-01. Two rules make this work: the SKU comes first so a sort puts every frame for a product together, and the word hero appears in exactly one file per product so nobody has to ask which one carries the listing. If your naming is ad hoc, our file naming system guide is twenty minutes well spent.
Send stacks as folders with the slice count in the folder name, so a retoucher can tell at a glance whether anything went missing in transit.
And keep everything you did not pick, in place, for the length of the client relationship. Product clients come back with odd requests months later: a different angle for a marketplace that demands a square crop, a shot of the base because a customer asked about the fitting. If your archive is only the frames you chose, every one of those is a reshoot. Our view on what to do with the photos you do not pick applies double here, where the samples may no longer exist.
The same checklist thinking carries across commercial work. Corporate headshots and events are picked against a brand page in much the same way, and real estate selection is another genre where a missing required frame is worse than a mediocre one.
FAQ
How do you cull product photography?
Work product by product rather than front to back through the folder. For each item, fill in every required angle from the client's list first, then choose the hero frame that will carry the listing, then keep an alternate for anything reflective. Only when every product has a full set do you compare all the heroes together as a grid, checking that size in frame, shadow direction and background white match across the catalogue.
Should AI culling software be used for product photography?
It helps in a narrow way. Blur, shake and exposure flagging work on any subject and will catch the frames where the tripod was knocked or a strobe failed. Everything face based, which is most of what these tools do well, has nothing to work with on a still life. Treat AI as a first sanity check rather than as the thing that chooses, and check any grouping feature carefully before it touches bracketed or stacked sets.
How do you handle focus stacks when culling?
Keep them together and judge them as a set, never as individual frames. Check that every slice arrived, that nothing moved between slices, that the focus range covers the whole product front to back, and that exposure is steady across the sequence. Give stacks their own color label so no later pass treats them as near duplicates, since automatic grouping tools are designed to reduce near identical frames to one and that is the wrong answer here.
How many photos should you deliver per product?
Ask the client, because most e-commerce platforms and marketplaces have a fixed number of slots and a required order. A common shape is one hero, three to five supporting angles, one detail and one scale or lifestyle shot. Agree the list before the shoot and treat it as a checklist during selection, because a missing angle usually means the sample has to come back rather than being something you can fix at the desk.
Why is consistency more important than the best single frame in catalogue work?
Customers see your photos as a grid, not one at a time. A frame that is beautiful on its own but sits at a different scale, with a shadow falling the other way, reads as a mistake on the page even though nothing about it is technically wrong. On catalogue jobs the set is the deliverable, so the correct pick is the one that matches its neighbors rather than the one that would win on its own.
How long should culling a product shoot take?
Less time than a wedding, and you should not try to make it much less. A typical catalogue day produces a few hundred frames, which is an hour or so of careful review. Speed is not the bottleneck in product work: accuracy is, because a wrong pick discovered after the samples have gone back means a reshoot rather than a slightly weaker gallery.
Never sort photos at 1 AM again.
Kepla for Mac clears the obvious misses from a card, names the reason on every frame it sets aside, and leaves the choosing to you. Nothing is ever deleted, moved or renamed. Free through the private preview · the first hundred photographers keep it at $99 a year.