How AI virtual try-on actually works (and why bags are the hard part)
No, it's not photoshopping a cut-out onto you. Here's the honest version of what's happening — and why the majors skipped accessories.
Nabi · 4 min readIt's placement, not paste
Modern AI try-on doesn't cut out a bag and slap it on your photo. It re-renders the object into your scene: matching the direction and softness of your light, casting a real contact shadow, and keeping your face and pose unchanged. The goal is 'shot on the same day,' not 'stickered on.'
The research lineage here is reference-guided object insertion — preserving a specific product and a specific person in one edit — which only became photoreal in the last generation of image models.
Why bags and jewelry are the hard case
Clothing try-on is a warping problem: drape fabric onto a body. A handbag is a rigid carried object, and jewelry is small, reflective, and unforgiving — thin chains and polished metal are where these systems break. That's a genuinely different and harder task than garment try-on.
It's also why Google's and Amazon's try-on shipped tops, dresses, and shoes and explicitly skipped handbags and accessories — the exact things people most want to picture on themselves. That gap is the whole reason Nabi exists.
Honest about the limits
It's not magic and it's not perfect: reflective hardware and very thin chains still vary shot to shot, which is why Nabi generates four variations and lets you pick. Every result is AI-generated; how you post it is up to you.
The best way to understand it is to feel it break and land on your own photo. Your first look is free, no signup.