The reshoot is on its way out. H&M created AI twins of 30 real models in 2025, built with consent and compensation, with the models keeping rights to their own likeness. Zalando has reportedly cut content production costs by around 90% and pulled turnaround down to days, though those figures come through secondary reporting and are worth verifying before you quote them anywhere.
What this buys you isn't cheaper photos. It's a different unit of production. Instead of shooting a look once on one body and hoping it converts across your whole customer base, you generate every SKU on a range of body types, then regenerate the set when the campaign changes. Seasonal refresh stops meaning a studio booking and starts meaning a render queue. A customer on a product page sees the garment on someone built like them, which is the job size charts have been failing at for 20 years.
The build is an imagery pipeline feeding your product pages, plus the thing most teams skip: a public disclosure policy. Label the images. Say what's generated and what isn't, in plain language, on the page and not in a footer.
That disclosure carries the whole risk. If a customer discovers your models are synthetic and you never said so, you lose more trust than the photography budget was ever worth, and transparency rules for synthetic media are tightening rather than loosening. There's a returns problem underneath it too: a generated image that flatters a fit the garment doesn't deliver drives returns, and returns eat the savings faster than a studio ever did. Consent also has to stay live. A one-time buyout buried in a release is how this becomes a story about you rather than a line in your P&L.
Photography was never the point. Showing someone what the thing looks like on a body like theirs was the point, and studios were just the only way to do it.
