Dupe takes a catalogue and gives back the handful of pairs worth a second look — narrowed in plain arithmetic for free, then adjudicated on our own GPUs into five verdicts, one of which is the only one that means anything should go.
Nothing opens in this browser — this one takes a job and does the work on our own graphics cards in Oregon. The free allowance below is how you try it: 60 pairs, no card and no payment, on the key you are issued.
It deletes nothing and it never looks at your video. The output is a short list with a verdict and a reason on each row.
The problem
The same scene uploaded twice under two titles. A trailer listed as a title. Part two of a shoot sold as a standalone and also inside the bundle. Every one of those splits the search traffic, splits the reviews, and makes the catalogue look padded — and finding them means reading two thousand rows against each other, which is a comparison nobody does by hand.
The narrowing. Arithmetic over titles, durations and dates costs nothing, because it is arithmetic — and it is what takes two thousand rows down to the forty pairs worth a model's attention.
Same scene, re-cut of the same scene, part of a series, trailer of a title, and unrelated. Only the first means something should be removed, and that call is yours to make.
Dupe has no write access to anything. It cannot remove a listing, cannot merge two, and has no button that would.
A re-cut is a different product. Part three of a series is supposed to look like part two. A trailer is supposed to share a title with the thing it advertises. A duplicate finder that calls all four of those duplicates would have you delete half a working catalogue, which is why the verdicts are five and not two.
How it works
Titles, durations, dates, and a description where you have one. No video, no stills — Dupe never asks for the file itself and could not read it if you sent it.
Arithmetic reduces the pairs to the ones that are genuinely near each other. You can stop here and pay nothing; plenty of people do, because the list is often short enough to read.
Each surviving pair goes to the model, which returns one of the five verdicts and the reason for it. Priced per pair adjudicated, never per pair considered.
What it will never do
Every bad outcome in a deduplication tool is the same outcome: something got removed that should not have. The only reliable way to prevent it is to not be able to.
Pricing
Two different free things, and they are worth keeping apart. The narrowing is free for everybody, always, because it is arithmetic. On top of that every key starts with 60 adjudicated pairs that cost nothing — which on a mid-sized catalogue is often the whole job.
A small catalogue's duplicates, usually in full.
A working catalogue gone through properly, once.
A library several labels deep.
The narrowing is free and always will be. You are paying for the model's judgement on the pairs the arithmetic could not settle, and for nothing else. Your licence key is emailed to you within two working days of payment.
Checkout is our own bitcoin server, not a card processor. The units you buy are added to your key: buy from the Buy more box inside the program and they arrive on their own once the payment confirms, or buy here and paste the invoice number from your receipt into that box.
Questions
Because duplicates in a catalogue are almost always visible in the metadata: two rows with the same runtime to the second, uploaded four months apart, with titles that share most of their words.
A video-based matcher would be more thorough and would also require us to hold your work, which we will not do. The free narrowing plus a model reading your own descriptions catches the pairs that are costing you money.
Then nothing happens, which is the design. The verdict is a line on a CSV, not an action.
The reason is printed beside every verdict so you can see what it keyed on — usually a shared runtime, a shared phrase, or a date gap.
Because series are the most common false positive and the most expensive one to get wrong.
Three parts of one shoot look nearly identical in metadata and are three products. A tool that cannot tell those from a double upload is not usable on a real catalogue.
No. Adjudication runs on Tower 1's own graphics cards in Oregon. Your catalogue text is not sent to a hosted model and is not in anybody's training set.
A key. One line of text, issued to you and shown exactly once — only its hash is kept here, so it cannot be looked up later and we could not tell you what yours is. Every key arrives with a free allowance already on it, which is the tier at the top of the pricing above, and the work starts the first time you send a job with that key on it.
Each job is metered in the unit named in the pricing, and a unit that fails — a file that will not read, a row the model held back — is not charged and is named in the job report rather than silently skipped. When an allowance runs out the next job is refused before any work starts, with the number it used printed on the refusal.
More units are bought in bitcoin. The checkout is a bitcoin server we run ourselves. Buy from the Buy more box inside the program and the units are added to your key on their own as soon as the payment confirms; buy from the plan buttons above and paste the invoice number from your receipt into that same box. Each payment adds its units once, to one key.
Send the catalogue and read the free list. If it is short, you are done without paying. If it is long, you have just found out why the search traffic is split.