"Keep the product identical": the Sume docs background-swap prompt

Sume documents a plain product-photo edit: send the product image as a reference and prompt "Keep the product identical; swap the background to a soft daylight studio". It is one request with one reference photo, and it works as a starting point for any background swap. The result still needs a visual check, because the model can alter labels, edges and shadows.

The request

The Image 1.0 page in the Sume docs uses that prompt on the legacy route. Image 1.0 is retiring, so here is the same idea on POST /v1/images. References go in input_references as image_url objects, and aspect_ratio "auto" matches the reference shape.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "Keep the product identical; swap the background to a soft daylight studio",
    "input_references": [
      {"type": "image_url", "image_url": {"url": "https://example.com/product.png"}}
    ],
    "aspect_ratio": "auto",
    "quality": "medium"
  }'

Why those fields

The reference URL must be public HTTPS; Sume rejects localhost, private-network and non-HTTPS URLs before submission. On edit calls, aspect_ratio "auto" matches the reference, and omitting the field is not the same. Quality medium is a middle setting; use high for finals and dense text. The docs example for Image 1.0 uses medium with a 4:3 ratio.

What to check in the result

A background swap changes pixels you did not ask about more often than people expect.

• Logo, label and barcode text: compare with the original.
• Edge halos around the product after the swap.
• Shadow direction against the new background light.
• Colors: compare a swatch of the product with the source photo.