Near identical product frames: how to actually pick one
You have forty frames of the same bottle on a locked tripod. They look the same at thumbnail size, they look the same at full size, and you have been staring at frames 12 and 19 for four minutes. This is not indecision. It is a specific problem with a method, and the method is not looking harder.
Why the folder stops you dead
Almost every article about picking photos assumes variety. A wedding gives you thousands of different moments and asks you to find the good ones. A locked off product setup gives you the opposite: one composition, repeated, where the only differences are the ones you cannot see from a normal viewing distance.
Two things are working against you, and neither is a lack of discipline. Human eyes are far better at comparison than at absolute judgment: you can reliably say which of two frames is brighter, and you cannot reliably say whether one frame on its own is bright enough. And more options make the decision worse rather than better. The classic work here is Iyengar and Lepper's jam study and Barry Schwartz's writing on the paradox of choice, both of which found that adding options raises regret and lowers satisfaction with whatever you end up choosing. Forty near identical frames is a machine for producing that feeling. Decision fatigue then does the rest.
This article is about that moment, the one setup and the forty candidates. The other half of product selection, matching frames across a whole catalogue so the kettle and the toaster look like they belong on the same page, is a different job and it lives in our guide to culling product photography.
If the camera did not move, the frames are identical by design
Start by being honest about how forty frames happened. On a tripod, under strobes, with a locked focus point, frames three through forty are supposed to be the same photograph. You did not shoot forty options. You shot one option, forty times, because firing again is free and reshooting a sample that has gone back to the client's warehouse is not.
That reframes the whole comparison. You are not looking for the best frame among forty. You are looking for the frames that got worse, and then taking any of the ones that did not. Which means the useful first question is not which do I prefer, it is what was actually able to change between these frames.
On a locked setup there are only a handful of live variables: the product settled or was nudged, dust or lint landed, a fingerprint arrived with the assistant's hand, a reflection moved because someone walked past a flag, the strobe failed to recycle, or something in the scene was genuinely in motion, like a pour or a splash. If none of those applies to your forty frames, they really are the same photograph, and picking a favorite is a fiction you are performing for yourself. The habit that produced the forty is worth examining too, and we wrote about that in the difference between a safety shot and overshooting.
Name the one variable before you compare anything
Before you open two frames side by side, write down which variable you are judging. One at a time, in a fixed order, so your eye cannot wander to whichever frame happens to feel nicer. The order below runs from the most expensive problem to the cheapest, so you eliminate on cost first.
| What can change | How to see it | What it costs later |
|---|---|---|
| Product position or lean | Compare edges against the frame border, not against each other | Transform work that shifts shadows and softens type |
| Label or seam rotation | Pick a fixed landmark on the label and watch only that | Often unfixable, and the reason for a reshoot |
| Reflection shape on gloss or chrome | Full size, on the curved surface, looking for room shapes | A rebuilt reflection, the most expensive retouch on the list |
| Focus plane | 100 percent on the text the customer has to read | Sometimes recoverable, usually not on type |
| Dust, lint, fingerprints | Full size sweep, worst on dark and matte surfaces | Minutes of cleanup, priced per speck |
| Exposure or strobe misfire | The histogram, or a flick between frames | Usually just eliminates the frame |
Working down that list rather than around it does two things. It stops you re-litigating the same pair from six different angles, and it means that when you eliminate a frame you can say why in four words. A pick you can explain is a pick that survives a note from art direction three weeks later.
Compare two at a time, and eliminate rather than rank
Nobody can hold forty images in their head. Trying to is what turns a five minute decision into forty minutes of scrolling. Run it as a knockout instead.
- Put frames one and two side by side at full size, on the first variable only.
- Keep the survivor and bring in frame three. Compare, keep the survivor, continue.
- When a frame is beaten, mark it as not picked and never open it again.
- Once you have run the whole set on the first variable, run the survivors on the second.
This feels slower and is dramatically faster, because each comparison is a small question with a real answer, and because your eye is doing the thing it is good at. Survey view in Lightroom Classic works well for three to six survivors. In Capture One, make a variant of a promising frame rather than a virtual copy of every near match, or you will be comparing your own copies within the hour.
Note what "eliminate" means. The frame is unpicked, not gone. Nothing moves, nothing is deleted, and every original stays on the card and in the archive, because product clients come back six months later asking for the base of the jar or a square crop for a marketplace. Committing to a decision without destroying the alternative is most of the skill, and we made the case in rejecting photos without second guessing yourself.
When two frames are equal, the tiebreak is retouch cost
Sooner or later you reach two frames that are genuinely both correct. Do not go back to taste. Ask which one is cheaper to finish, because that answer is real money and your preference between two identical bottles is not.
Count the work attached to each: how many specks of dust, whether the reflection needs rebuilding or just tidying, whether the label needs a nudge in transform, whether the shadow has to be extended for a wider crop. The winner is the frame with the least invisible labor behind it, even when it is fractionally less pretty. That is also the frame your retoucher would choose if you asked, and asking is not a bad idea.
Two cases complicate this. A slightly soft frame is not automatically out, because softness on a background is different from softness on type, and the threshold is worth knowing rather than guessing at: when a soft photo is saveable covers where the line sits. And a single speck of dust on an otherwise perfect frame is usually cheaper to remove than a rebuilt reflection on the cleaner sample, so compare the total, not the count of flaws.
Write the estimate next to the pick while it is fresh: light cleanup, standard cleanup, or complex composite. When the client asks on Thursday whether forty selects can be delivered by Friday, that note is the answer, and a star rating is not.
A stopping rule, so the loop actually ends
The comparison needs an exit, or it will keep expanding to fill the evening. Use a rule you set before you started rather than a feeling you have at the end of it.
The twenty second rule
Put the two finalists side by side at 100 percent and give yourself twenty seconds to name the difference out loud. If you can name it, that is your decision criterion and you are done. If you cannot, the difference is below the threshold that will ever reach a customer scrolling a shop page on a phone. Take the earlier frame, mark it, and move to the next SKU.
Take the earlier one because time on a product table only makes things worse: dust settles, fingerprints accumulate, gels drift warm, and the sample gets handled. When nothing distinguishes two frames, the earlier one is very slightly more likely to be clean, and that is a better tiebreak than a coin toss.
The other half of the rule is a note. Record the reason next to the pick, even if it is four words. It stops you reopening the same forty frames when the retoucher asks a question, and it means the second campaign for the same product starts from your reasoning rather than from scratch. Building that instinct is slow, honest work, which is roughly the argument in training your eye.
What software can and cannot see on a still life
Automatic selection tools are trained where the demand is, which is photographs of people. Adobe says plainly that the Assisted Culling now shipping in Lightroom is optimized for portraits and headshots, and what it scores is subject focus, eye focus and whether eyes are open. A bottle has none of those things.
What still transfers is genuinely useful. Blur and shake detection does not care what the subject is, so it will find the frame where the tripod was knocked. Exposure flagging catches the strobe that did not recycle. Both are checks you would rather not do by eye four hundred times.
The feature to be careful with is near duplicate grouping. On a locked off setup every single frame is a near duplicate, including your bracket frames and your focus stack slices, and a tool built to reduce a burst of a laughing bride to one frame will happily offer to reduce your stack to one slice. That is a correct answer to a question you did not ask. How duplicate detection actually works explains why it behaves that way and how to keep sets intact.
And no tool can judge whether the label is turned the way the brand guidelines require, whether that speck is dust or a genuine surface feature you must not retouch away, or whether the reflection reads as steel or as a softbox. Those are the decisions the whole job is made of.
Where Kepla fits, said honestly
Kepla is built around the shape of problem this article describes: a run of nearly identical frames, one of which you have to choose. It checks each frame against clear criteria, groups the ones that are effectively the same photograph, and hands you a shortlist to approve. It picks, and that is all it does to your files. Nothing is deleted, moved or renamed, ever, so a frame it did not pick is still exactly where you left it, one tap away.
Being straight about the state of it: the picking app for iPhone, iPad and Mac is still being built, and it is not shipping today. On product work the phone would handle the first pass, the obvious misfires and the grouping, and the Mac is where the real comparison happens, because judging a reflection on chrome at 100 percent is not a phone job and we are not going to pretend otherwise. The booking page is live and free right now, which is a different part of the same week.
The nearest relative of this problem in a completely different genre is a corporate headshot day, where six frames of one face have to become one file, 200 times over. The method is the same and the stakes are different, and it is written up in headshot day for 200 employees.
FAQ
How do I choose between product photos that look identical?
Do not look for the best one. Work out which variables could have changed between the frames, usually product position, label rotation, reflection shape, focus, dust or a misfired strobe, then compare two frames at a time on one variable and eliminate the loser. If you cannot name a difference in twenty seconds at full size, there is no difference the customer will see, so take the earlier frame and move on.
How many frames should I shoot of a locked off product setup?
Fewer than most people do. Once the camera, light and product are locked, extra frames are insurance against a nudge or a misfire rather than genuine options, so a small number plus a deliberate check on screen beats firing thirty. The cost of the extra frames is not storage, it is the comparison you have to run later on every one of them.
Should I reject a product shot because of a speck of dust?
Not automatically. Compare the total cost of finishing each candidate rather than counting flaws. A single removable speck is often much cheaper than rebuilding a reflection or correcting a lean on the alternative frame, and it may be far cheaper than getting a discontinued sample back. Mark the cleanup clearly so the retoucher knows it was a decision and not an oversight.
Can AI pick product photos for me?
Partly. Blur, shake and exposure checks are subject neutral and work fine on a still life. Anything face based is inert, since expression and eye detection have nothing to work with. Be careful with near duplicate grouping, because on a locked off setup every frame is a near duplicate, including bracket and focus stack frames that only work as complete sets.
Why is picking product photos so much harder than it looks?
Because you are making an absolute judgment, which human eyes are poor at, on differences that are near the limit of perception. Comparison is much easier than judging a single frame on its own, which is why pairs work and scrolling does not. Research on choice also suggests that more options tend to increase regret rather than confidence in the option you pick.
Forty frames, one decision.
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.