A 3D view on a product page competes with photographs, which load instantly and look real. It earns its place by doing something photographs cannot.
What it does better
- Any angle, including the ones photography skipped.
- Configuration. Colours, materials and options without a photoshoot per combination.
- AR placement. Seeing the object at real size in the room it is destined for.
- Interaction. Opening, rotating, examining a mechanism.
What it does worse
- Realism, usually. A photograph of a real object beats a render of it unless the render is very good.
- Load time. A photograph is tens of kilobytes.
- Production cost per item, for a catalogue of any size.
The requirements
- Accurate colour. A customer receiving something the wrong colour is a return, and this is the fastest way for 3D to lose credibility.
- Accurate scale, particularly for AR.
- Fast loading, or people leave before it appears.
- Mobile first, where most shopping happens.
- A fallback for devices that cannot render it.
Where generated models fit
The honest answer is: better for categories where approximate is acceptable, and worse where exactness is the point.
Generation is a poor fit for products where the customer is buying the specific details — jewellery, electronics, anything where a wrong proportion misrepresents the item. It is a much better fit for large soft goods, furniture in context, and for placeholder or category-level visualisation while a photographic pipeline catches up.
Whatever the source, a product model that misrepresents the product is worse than no model at all.
Catalogue scale
The real question is per-item cost across thousands of SKUs. That is where automated pipelines matter, and where the trade between fidelity and throughput has to be made deliberately rather than per item.