Every label in Surat has now seen the demo. Someone uploads a photograph of a lehenga, types a sentence, and a model appears in a palace courtyard wearing it. It looks like a ₹3 lakh shoot and it cost nothing. So the question I get asked is: can we stop doing photoshoots?
The honest answer is partly, and only if you understand exactly where it breaks. We produce the campaign imagery for JMS Studio, a bridal house whose embroidery is the whole point, and the rule we learned there is the one that matters.
The rule
The further the generated pose moves from your source photograph, the more the model redesigns your product.
Give it a straight-on mannequin shot and ask for a model standing in a courtyard, and the garment survives. Ask for her twirling, seated, or turned three-quarters away, and the model has to invent the parts of the lehenga it cannot see. It invents them beautifully. They are not your lehenga. The kali count changes, the border shifts, the cutdana becomes sequins.
For a T-shirt brand that might not matter. For a bridal house where a customer pays for the embroidery, it is the difference between a campaign and a fraud.
What we do for JMS
Two beats per design, from real shots. A mannequin photograph and a close-up go in. Two campaign images come out: the garment on a model in a scene, and a second beat, a different mood or crop, for the grid. Both stay close to the source silhouette. Both are checked against the close-up before they go anywhere.
No generated detail shots. Ever. The zari, the threadwork, the fall of the dupatta are photographs. They sit next to the campaign images on the product page, and they are what the bride zooms into.
Consistency across a collection. Same light, same world, twenty designs. That is where AI genuinely beats a shoot: a real studio day produces variation whether you want it or not.
Square, for the grid. The campaign beats are made at the size and crop Instagram actually shows, not cropped down afterwards.
What it costs, honestly
Third-party figures, not our rates. Product photography in India runs about ₹150 to ₹2,000 per image, and ₹80 to ₹200 in tier-two cities for basic catalogue shots. AI generation runs ₹10 to ₹150 per image (Aivastra, CatalogX, 2026).
But the per-image number is not the saving. The saving is the studio day, the model, the location, the make-up, the retoucher, and the three weeks. Against that, the cost of a good AI campaign is mostly the time of someone who knows the rule above and checks every image against the close-up.
Where it goes wrong, so you can watch for it
- The garment changed. Compare every campaign image to the close-up. Count the kalis. Look at the border. If anything differs, the image does not ship.
- The model has the same face in every image. Buyers notice. Vary her, or use her once per collection.
- The setting is a palace nobody in Surat has. A courtyard the bride recognises beats a fantasy she does not.
- Detail shots were generated to save a day. This is the one that costs a customer. Do not.
If you are doing it yourself
- Shoot the source properly. One mannequin shot, one close-up, plain background, good light. This is the day you cannot skip.
- Keep the pose close to the source. Standing, facing, arms where they were. Resist the twirl.
- Check every image against the close-up before it goes on the page.
- Never generate detail. Photograph it.
If you want to see what this looks like on a real collection, the JMS campaigns are in the case study. If you want it done for yours, message me with your Instagram handle and I will tell you whether your existing shots are good enough to start from.