Every wardrobe app has made the same promise: take a picture, and we'll organize your closet. Almost none of them survive contact with a real closet. The typical flow still asks you to confirm or type in the category, the color, the fabric, the fit — one item at a time — and by item twelve, most people close the tab and never come back. That's not a design flaw you can fix with a nicer UI. It's the wrong architecture.
Clad9 was built around a different bet: the app should do the identifying, not the user. Here's what actually happens between the moment you upload 30 closet photos and the moment they're a searchable, tagged wardrobe.
Step one: group the photos by item
Most closets get photographed in bursts — front, back, a detail shot of a pattern — so the first job is figuring out which photos belong to the same physical item before anything gets tagged. Get this step wrong and everything downstream is wrong too: one garment gets counted twice, or two different garments get merged into one confused entry.
Step two: identify against a real taxonomy, not a guess
Each item is then identified against a curated taxonomy — type, color, pattern, fit, fabric weight, formality, and layering role — and given a one-line style description. This is what turns a photo into something a recommendation engine can actually reason about later: not "a shirt," but a slim-fit navy oxford, business-casual, layerable.
The duplicate problem, with real numbers
The hardest part of this pipeline isn't identification — it's telling apart "this is the same item again" from "this is a different item that happens to look similar." On a real closet, we measured two genuinely different garments coming back as close as 0.199 apart on similarity, while two photos of the exact same watch measured 0.255 apart. Similarity alone cannot reliably tell you it's the same item.
That's why Clad9 requires closeness and matching type and matching color before it ever calls two entries a duplicate. It's a stricter bar than similarity scoring alone, and it's the reason re-photographing your closet doesn't quietly double it.
The free check that runs before any AI call
Before any of that identification work happens, Clad9 fingerprints the image itself. Re-uploaded photos land within a bit or two of their originals; unrelated photos land twenty-plus apart. Exact or near-exact repeats get dropped instantly, before they ever reach the more expensive identification step — which keeps the pipeline both fast and cheap to run at scale.
What all of this is actually for
None of this is organizing for its own sake. Every piece ends up as a searchable item with enough real detail that a request like "something smart-casual in a cool color that works with brown boots" returns actual garments from your actual closet — not a generic suggestion, not something you'd have to buy. You can read more about how the recommendation side of this works on the wardrobe capture page, and see the underlying approach on the methodology page.
If you're deciding whether decluttering or digitizing comes first, we've written about that trade-off directly: how to organize and declutter your wardrobe.
Ten minutes of photos, one time
The entire capture step takes about ten minutes for an average closet, done once. Here's how to shoot the photos well so the pipeline above has the best possible input to work with.
Ten minutes of photos. A closet that catalogs itself. Try Clad9 free →
Clad9 is an AI wardrobe and personal styling app. Photograph your closet once and Clad9 catalogs every piece, then recommends what to wear each day around your body, your colors, the weather and what's actually on your calendar. Start at clad9.com.