Fabric forgets how gravity works
AI fashion shoots have one consistent tell. Fabric forgets how gravity works.
A silk blouse hangs in a way that suggests the room has no gravity. The hem swings to one side like an invisible breeze is blowing only inside the AI's compositional center of attention. A dress drapes against the body in a way that real fabric, under real gravity, in real time, doesn't drape.
This is fabric inertia drift. It's the 2026 equivalent of "AI hands" from 2024. Everyone fixes the obvious problem (hands, eyes) and ignores the subtle one that's now the actual tell.
The model knows what silk looks like. It doesn't know how silk falls. Training data gave it billions of static images of garments — almost no video of garments in motion. So the model invented physics that look plausible in a single frame and impossible the moment you look closely.
The fix in my workflow: I generate the base shot, then run a second pass with a physics-aware in-painting model that specifically corrects fabric drape. Or, more often, I composite real fabric movement from a stock video frame into the AI image. Sounds janky. Works completely. Three minutes per image. Total invisibility.
If you're shooting AI fashion for a Dubai brand and the work feels uncanny but you can't say why, look at the fabric. It's almost always the fabric. Fix that and the image becomes invisible AI-craft instead of obvious AI-output.