GPT Image 2 for E-commerce Product Photography

If your catalog team still books a studio day for every new SKU, you are leaving a 90 percent cost reduction on the table. GPT Image 2 on fal.ai handles three of the four shots a product listing typically needs: the white background hero, the in-context lifestyle, and the close-up detail. The only shot worth a human photographer is the one that sets the brand reference for the season, and even that can be relit and restaged via the edit endpoint.

The four-shot catalog

For a typical e-commerce product you need a white background hero at 1:1, a lifestyle shot that places the product in use, two or three detail shots of material and finish, and one group shot if the SKU comes in color variants. GPT Image 2 on fal-ai/gpt-image-2 plus fal-ai/gpt-image-2/edit covers all four without a booking.

The reference pattern

Take one real photograph of the product. This is the SKU reference. Every subsequent shot goes through the edit endpoint with this reference as image_urls. The edit prompt describes only the new framing, not the product. This keeps the product identity stable across the entire catalog.

1const hero = await fal.subscribe("fal-ai/gpt-image-2/edit", {
2  input: {
3    prompt: "Place this product on a seamless cream backdrop, soft studio light, 1:1 framing, subtle contact shadow. Preserve every detail of the product exactly.",
4    image_urls: [SKU_REFERENCE_URL],
5    image_size: "1024x1024",
6    quality: "high",
7    input_fidelity: "high",
8    num_images: 1,
9    output_format: "png",
10  },
11});

Cost per SKU

At quality=high, four shots per SKU costs about $0.50. A thousand SKU catalog is $500 versus $40,000 for a studio shoot. Keep the studio for your top 20 SKUs. Run the other 980 through fal.ai.