Compose image

How 19 models handled “Compose image” from our eval suite, FLUX.2 Klein 9B Image Generator first, then newest model first.

What we look for

Combines the referenced woman, white coat and handbag into one coherent photorealistic fashion shot. Identity and products match the references.

Prompt

“full body fashion photo of a woman wearing a white coat and holding a hand bag”

Inputs

Tests for

Reference following, consistency