Every app that finds clothes from photos says it works. Almost none of them show you a run. So on 4 September 2026 we did the least glamorous thing a product team can do: we picked sixteen photos we did not choose to flatter the product, pushed them through the exact search the Remode app and the browser trial use, and wrote down what came back. All of it. The good results, the honest "we are not sure" results, and the two listings with prices that were plainly wrong.
The full method and the row-by-row table live on our How we test page, together with the raw JSON. This post is the readable version.
What we put in
Sixteen images. Five were our own assets: a retailer-style product photo of a white Nike Air Force 1 (searched twice, once for the United States and once for Switzerland), a beige trench coat worn outdoors (searched for the US and for Germany), and a casual full-outfit photo with several pieces in frame. Eleven were openly licensed photos from Wikimedia Commons, chosen for realism rather than convenience: a pair of Converse Jack Purcells shot by an amateur, blue-and-white sneakers on a pavement, a used Hermès Kelly bag, a 1935 to 1945 Prada handbag photographed in a museum, a streetwear hoodie over a hockey jersey, a plain hoodie portrait, two candid photos of women in floral dresses, and two close-ups of leather work boots.
The photos went in as file uploads, the way a phone sends them. A fresh guest identity was used every three searches so the free-tier limit never became the thing being measured. Regions: fourteen searches for the US, one for Switzerland, one for Germany.
What came back, in numbers
| Measure | Result | What it means |
|---|---|---|
| Searches that returned shoppable offers | 16 of 16 | Every photo produced at least one store result for its first detected piece. |
| Offers and stores | 291 offers, 178 distinct stores | Counted across all sixteen searches for the first piece only. |
| Pieces detected | 31 in 16 photos | Eight photos contained more than one piece; each can be searched separately in the app. |
| Verdict on the top result | 1 exact · 14 best visual match · 1 similar | Remode claimed an exact match once. Everything else it labeled as a strong candidate or a look-alike. |
| Time per search | 14 to 26 seconds, median 17 | Measured end to end from our client; the app shows progress while it works. |
The one exact match was the hardest photo
We expected the product shot of the Air Force 1 to be the easy exact. It was not. Remode called it a best visual match and put the Nike Air Force 1 '07 at the top, at Hibbett for USD 125 in the US and at Decathlon Switzerland for CHF 139 in the Swiss run. Right shoe, right colorway, but the model refused to call it identical, because a product photo does not show the details that separate one Air Force 1 release from another.
The exact came from the photo we thought would fail: a used Hermès Kelly, photographed by its owner. Remode detected an "olive ostrich leather Kelly bag", labeled the top result exact, and returned a vintage Kelly Sellier 32 in Vert Olive ostrich at Madison Avenue Couture for USD 22,500, with eighteen more listings from resale specialists including Christie's. A distinctive object with a distinctive material is easier to be certain about than a mass-produced sneaker. That is uncomfortable and also correct.
Where honesty showed up
The museum Prada bag from the 1930s was the control. Nobody sells it. Remode detected a "brown embossed leather tote bag", labeled the top result best visual match, and returned vintage textured leather handbags from Etsy and The RealReal. It did not say Prada. It did not invent a listing. That is the behavior the whole product is built around: when we're not sure, we don't pretend.
The casual outfit photo produced the widest detection of the day: six pieces, a beige trench, a white ribbed top, light-wash jeans, a navy baseball cap, white sneakers and a beige crossbody bag. The top result for the trench was labeled similar, a Macy's double-breasted belted trench at USD 270, which is the right label for a coat whose brand the photo does not reveal.
The streetwear photo is the best example of what "best visual match" does and does not promise. Remode detected a red Chicago hockey jersey, a white hoodie and a white durag. The top result was a Mitchell & Ness New Jersey Devils 1999 jersey at BSTN for USD 195. Same silhouette, same era of design, different team. The label said best visual match, not exact, and that distinction is exactly why the label exists.
What we found that was wrong
Publishing a test means publishing the flaws. Among the 291 offers, two carried prices that were clearly wrong: one Air Force 1 listing showed USD 70,634 and one Jack Purcell listing USD 3,631, both almost certainly a currency or quantity field mis-read at the source. A few "stores" in the boot searches were not retailers at all but sites that happened to host a picture of a boot, and one US dress search put a listing from Amazon India at the top. None of these break the search, but all of them are ours to fix, and they are now on the engineering list with this page as the reproduction.
What this run does not prove
It is one day and sixteen photos. It does not measure whether a human would agree with every label, it does not compare Remode with any other tool, and openly licensed photos are cleaner than the compressed Instagram screenshots most people actually search with. Prices in the table were correct at the time of the search and will have moved by the time you read this. We will re-run the battery when the search changes materially and publish the new date at the top of the results page.
Run your own
You do not need to take our word for any of this. The browser trial gives you three searches without installing anything. Take three screenshots you actually care about, run them, and compare the labels with what any other tool tells you for the same images. If you want a fair protocol for that comparison, we wrote one: how to test a clothes-finder app yourself in ten minutes.
See what it says about your screenshot.
Three free searches in your browser, or the full app on iPhone. Every result labeled exact, best visual match or similar.
