Can an AI image of a real sneaker get every logo, eyelet and hole right? Yes, but not from a prompt alone. It took us five rounds, a written list of every part and a check by eye on every count.

27 Oct 2026 · by LS · originally published at longlabs.tech

TL;DR

We made one AI image of a real sneaker and checked 17 parts against the real shoe: logos, counts, badge text, the pair.

Round one, from 3 photos and a vague prompt, passed 7 of 13 visible checks. The logo sat on the tongue and the side logo was gone.

Photos from every side and a written fact sheet fixed the placement. Letters, badge and hole rows needed one edit at a time on a crop. Round five passed 15 of 15.

One vision model run on the whole image approved 6 of 13 parts we had broken on purpose (8 test images). Counts go to a pixel counter and a person.

One real sneaker and a fact sheet of 17 points

The shoe is a purple mid-top sneaker from a big sports brand. I own a pair.…