GPT Image 2 for Product Photography: Matched Reference Test
GPT Image 2 produced one of the strongest reference-led campaign frames in the matched Sloane test. It followed the diagonal flash-lit direction, kept the identity recognizable, preserved the dress's pale-yellow sleeveless construction, and held the circular-detail bag as a distinct product.
This review now uses the same visible inputs and one-output rule as the rest of the article family. Older weak-source examples are no longer part of its public evidence.
The complete input reference set
The identity, garment, and accessory are the approved references from Maison Urban prompt 46557199, copied into workspace 896 (Articles).
The product sources exceed 2K on the long edge. Sloane's 679×722 portrait is the disclosed exception to the 1600px source-quality gate, so the identity result is directional.
Prompt
Dress hero shot. An off-kilter, flash-lit photograph of the referenced woman in a dynamic, almost off-balance leaning pose against a dark wooden bar counter or pillar at night. Her body creates a diagonal line. She wears the reference dress and holds the reference bag. The strong flash casts a sharp shadow. Blurred shelves with bottles and bar lights define the out-of-focus background. Cool, edgy vibe.
The matched GPT Image 2 result
The downloaded original is 3792×2160. The article serves an optimized 1920px-wide WebP derivative and does not describe that derivative as the master.
Fidelity review
The output passes both publication gates. It looks clean at article size and keeps the central source relationships: Sloane's face remains recognizable, the dress stays pale yellow with a sleeveless V-neck, two front buttons, belt, and long skirt, and the gold bag is not merged into the garment.
The result still interprets rather than reproduces. Pleats and belt geometry shift; some bag discs and chain details change; and the accessory occupies less of the frame than in the strongest accessory-led result. That makes this a strong campaign image, not automatic catalog evidence.
GPT Image 2 or GPT Image 2.5?
In this shared brief, GPT Image 2 remains competitive. GPT Image 2.5 Sunburst pushes the diagonal composition harder, while Flare gives the accessory more room. The useful workflow is to keep the same source set and run the compatible variants side by side rather than infer an upgrade from the name alone.
Bottom line
GPT Image 2 passes this reference-led fashion test at high output resolution. Its best qualities here are art direction, identity handling, and a coherent garment-led composition; product construction still needs inspection. Compare it with both GPT Image 2.5 variants in the full matched benchmark.
For current Astria rates and billing terms, see Astria pricing.