Do we really need a $5,000 reshoot?

> WHAT WE WALKED INTO

A baby-category partner needed its product shown on a range of models. The usual answer is a reshoot, and the going benchmark for one is about $5,000 before the first image exists. We kept the real product photograph and had AI build the model and the scene around it: six renders, five new models from one hero shot, $0.12 each, $0.72 in generation cost. The partner saw them as what they were, AI renders, and approved them on fidelity as a proof of concept. The 72 cents is real. It is also the generation line only, and the $5,000 it sits against includes everyone’s time. We print both because the comparison is still lopsided after you add ours.

> THE MOMENT · THE SET, AS BILLED

> STEP 01

The product stays the photograph

The scene and the model are generated around the real product image; the product pixels are the source photo in every frame. This is the rule that makes AI imagery usable on a listing at all, because a generated product is a listing-accuracy problem waiting for a complaint.

> STEP 02

Five models from one hero shot

Six renders in total, five new models from a single existing hero image, at $0.12 each. The range is the point: a reshoot gives you the models you booked, a render gives you the ones you did not think of.

> STEP 03

Reviewed every frame

Every image checked against the real product: straps, seams, colors, proportions. Frames that drift get remade, and remakes are time, not the $0.72. Review is the real cost of this method, and it is still a fraction of a shoot.

> STEP 04

Approved on fidelity, as a proof of concept

No listing performance is claimed for these images, because none was measured, and a render that looks right is not the same as a render that sells. The test that settles it is a click-share read or a shopper poll before the images go into the stack.

> THE NUMBERS, DRAWN FROM THIS FILE ONLY

> WHAT CHANGED · FLIP IT YOURSELF