Why general-purpose image models can't produce manufacturable jewelry
mulapis Research
A general-purpose image model optimizes for visual plausibility: the scene looks real, the light looks real, the material looks real. Jewelry production instead demands that geometry and craft hold up. The two objectives are different. That gap separates a beautiful image from a piece you can actually build.
An image has no geometry
Image models output pixels rather than solid structure. Renders carry no wall thickness, dimensions, or seat depths. To build from an image, a workshop must recreate the design from scratch.
Visually plausible is not craft-sound
From millions of photos, an image model learns what jewelry looks like — but never the physics of casting. It draws floating settings, prongs thinner than the casting minimum, openwork with no tool access — each plausible on screen, none workable on the bench. These are not random flaws; the model's objective simply does not include manufacturability.
Production is a constraint problem, not a style problem
AI jewelry design is a problem of constrained generation. Metal, stones, settings, and ring sizes are physical requirements that generation must satisfy. mulapis pairs an orchestration engine with specification reviews that regenerate failing designs automatically, embedding manufacturing rules into the workflow.
Revisions must be controllable
Real orders run on local revisions: resize the band, keep the stone; swap the side stones, keep the shank. An image model regenerates the whole picture each time and cannot guarantee that one change leaves the rest untouched. Controllable revision requires a design carried by spec and geometry, not by pixels.