1. It starts with the photographs

Every result is built from the photos you upload. Please follow the photography guidelines we have provided for your garment type. Most fidelity problems we see begin with an input photo rather than with the AI. If you are unsure how to shoot a particular piece, reach out to us before you photograph it and we will tell you exactly how.

2. What your rating is for

Marking an image Garment Same or Garment Changed is a quality evaluation. It is the strongest signal we have for tuning the AI to your specific products, and it is the reason the output keeps improving batch after batch.

It is not a billing action. Images are charged at the point they are generated, so an image you mark Garment Changed is not refunded or re-credited. Rate freely and honestly — an honest rating costs you nothing and makes your next batch better.

We run this evaluation in the open, with you, and we do it deliberately. A figure you can see and check yourself is what makes this service predictable: you know what to expect from a batch before you commit to it. Most of this market will not show you its accuracy at all. We would rather earn your trust with a number than ask you for it.

3. What a good score looks like

AI garment reproduction is not a solved problem anywhere in this industry. Across the field, 40–50% garment fidelity is the accepted benchmark — that is the bar serious tools are measured against today. Your own Fidelity figure unlocks once you have rated 100 images; below that the number is noise rather than a measurement. What we commit to is the direction of travel — that figure should climb over time, and your ratings are what move it.