Category guide
Photo culling and image ranking software
Four different kinds of tool get called “culling software”, and they solve different halves of the problem. This is what each one is for — including the cases where the answer is not Prevue.
AI culling in your editor
Software that scans a shoot and rejects the closed eyes, the missed focus, the frames nobody would keep. Built into or bolted onto Lightroom-style catalogues.
- Good for:
- Getting from 2,000 frames to 200 without looking at all of them. The technical rejects are a machine's job and it is faster than you at it.
- Where it stops:
- It cannot tell you which of the 200 are your best, because that is a judgement about the picture, not the exposure.
Head-to-head ranking tools
Show two frames, pick the stronger, repeat. The choices build an order. Prevue, Prioneer and the free browser sorters all work this way.
- Good for:
- The last stretch — the shortlist where every frame is technically fine and the question is which ones actually ship.
- Where it stops:
- It costs you attention: someone has to make the calls. Tools differ mostly in whether they tell you when to stop and who else can help.
Paid panel testing
Post two options to a rented panel of strangers, pay per response, get preferences with written reasoning. PickFu is the category benchmark.
- Good for:
- Thumbnails, listings, covers — anything where a specific market's opinion is the answer and yours is not.
- Where it stops:
- Priced per response, so it suits a handful of finalists rather than a whole shoot.
Browser sorters
Load a folder, click through pairs, get an order. No account, no cost, nothing stored.
- Good for:
- A personal set, right now, with no stakes attached.
- Where it stops:
- No confidence scoring, no second opinions, no protection for images you do not want circulating.
Where Prevue sits
Prevue is the judgement pass, for images specifically. Cull the technical rejects however you already do; bring the shortlist here, compare two frames at a time, and end up with the set in order and a confidence score on every position — so the near-duplicates you keep flip-flopping on get flagged as unsettled instead of arbitrated by whichever one you saw last. Evaluators only ever see watermarked copies on URLs that expire.
It is the wrong tool if you want a machine to reject 2,000 frames unattended, or if you need opinions from a market you do not have access to.
Prevue compared, one at a time
Competitor details on those pages are what each published as of 2026-08-17.
Questions people ask
What is photo culling, exactly?
Cutting a shoot down to the frames worth delivering. It usually happens in two passes: a technical pass that throws out the blinks and the missed focus, and a judgement pass that decides which of the keepers are actually your strongest. Most tools do one or the other, and most of the pain is in the second.
Do I need a culling tool if my editor already has one?
For the technical pass, probably not — your catalogue software is fine at rejecting the obvious. The gap is the judgement pass, where you are staring at eight near-identical frames that are all sharp and all fine, and picking by whichever one you happened to look at last.
Why pairwise instead of rating each image out of ten?
Because your ten drifts. The score you give the first frame and the score you give the hundredth are not on the same scale, and neither are yours and a colleague's. A choice between two frames is a decision you can make in under three seconds and repeat consistently, and enough of those choices produce an order — Prevue uses a Bradley-Terry model, the standard way of turning head-to-head results into a ranking.
How many comparisons does ranking a set take?
For a 20-image set, roughly 60 to 80 head-to-heads — four to eight minutes — before the order stops moving. Not every pair: a good pairing engine spends your attention on the frames that are still contested rather than re-confirming the obvious winner.
Is it safe to upload client work to a ranking tool?
Ask each tool what an evaluator's browser actually receives. In Prevue's case it is a copy watermarked on the server, on a URL that stops working after 15 minutes; the unwatermarked original is stored separately and only you can reach it. If a tool cannot answer that question, that is the answer.