GenAI Image Transformation

Product photography is slow and expensive. Getting physical samples, booking a studio, arranging models and retouching can take weeks for one collection. Image generation cuts that to hours.

Golden Sea used CycleGAN, a type of model that learns to turn one kind of image into another, for two jobs: putting clothes on models and furnishing empty rooms.

For fashion clients, the system takes a flat photo of a garment and a photo of a model and produces a realistic picture of the model wearing it, with no sample or studio needed. It was trained on thousands of garment and model images to get fabric, drape and lighting right.

For interior clients, it takes a photo of an empty room and a furniture product and returns the room furnished. Clients use these images for online shops, marketing and proposals, instead of paying for 3D renders.

In both cases, going from brief to finished image is 70% faster than a photo shoot, and clients run whole catalogues of hundreds of products overnight.