A polished copper pan is the single hardest reflection in product photography. It is a large, curved mirror, so it shows the entire room, the window, the counter, the photographer, and the curve bends all of that into a warped reflection that has to stay physically plausible. Real cookware shoots use light tents and hours of retouching just to control what the metal reflects. It is exactly the kind of subject you would expect AI to turn into garbled chaos.
It did not. I ran one brief, a polished copper saucepan with a riveted steel handle on a marble counter by a window, the copper mirroring the kitchen, through four of the strongest image models with the same prompt: Nano Banana 2, GPT Image 2, Seedream 4.5, and FLUX.2 Pro. All four rendered a coherent, believable mirror reflection, with one beautiful catch worth understanding. This is the cookware entry in our product-photography series, alongside the skincare, jewelry, supplements, makeup, food and beverage, footwear, candles, clothing, furniture, electronics, handbags, sunglasses, glassware, flowers, watches, perfume, packaging, pet products, toys, textiles, stationery, drinkware, soap, ceramics, art prints, earbuds, houseplants, knives, and automotive wheels tests and the broader best AI image model for product photography roundup.
Quick answer
- Best overall, and cheapest photoreal: Seedream 4.5. The richest, most coherent reflection and the most convincing copper.
- The hardest reflection is solved: all four rendered a believable curve-warped mirror reflection of the room, no garbled metal.
- The catch: the reflected kitchen is invented. It looks real, but it is a scene the model generated, not your actual environment.
If you only remember one thing: reflective cookware comes out beautifully now, so choose on the copper and the reflection quality, but treat the reflection as decorative, it is a hallucinated room, not a real one.
The test, model by model
One brief, four models, same prompt. I judged the mirror reflection first, then the copper, then the handle and rivets.
Seedream 4.5 rendered the hardest part the best. The copper mirrors a detailed, believable kitchen, a window with light, a cooktop with knobs, jars on the counter, and crucially the reflection is warped correctly by the curve of the pan, the way real metal bends a scene. The copper itself is the warmest and most convincing, and the riveted steel handle is detailed. This is the same material and optics strength it showed on jewelry and glassware, on the largest mirror surface yet. Best result, lowest cost of the photoreal options.
Nano Banana 2 produced a warm, clean copper pan with a coherent reflection: the window and counter mirror believably across the curve, the stainless lining reads at the rim, and the riveted handle is convincing. It is slightly less detailed in the reflection than Seedream's macro, but as a finished product shot of a notoriously hard subject, it is excellent and a third of GPT's cost.
GPT Image 2 gave a bright, tidy result with a coherent reflection of a sunlit kitchen, and it added a believable reflection of the pan on the polished marble counter below, a nice touch of physical correctness. The copper is a touch brighter and cooler than Seedream's, and the reflection is a little less detailed, but it gets the hard thing right at the highest price of the four.
FLUX.2 Pro produced a convincing copper pan at the lowest cost, with a coherent reflection of a bright kitchen and a clean mirror on the marble. The reflection is less detailed than the macros, the usual FLUX softening, but it is plausible and warp-correct. For a quick catalog shot of reflective cookware it is solid and cheap.
The comparison
| Model | Mirror reflection | Copper finish | Handle + rivets | Rough cost/image |
|---|---|---|---|---|
| Seedream 4.5 | Best, richly detailed | Best, warm | Detailed | ~4.8 credits |
| Nano Banana 2 | Coherent, clean | Warm | Believable | ~9.3 credits |
| GPT Image 2 | Coherent, plus marble | Bright, cooler | Clean | ~26.4 credits |
| FLUX.2 Pro | Coherent, softer | Warm | Clean | ~3.6 credits |
Credit costs are first-hand from this test on Masonry; per-image rates move, so check current pricing.
Why reflective metal is no longer the nightmare
Cookware has a fearsome reputation in real studios, and the useful finding is that the thing that makes it hard for a camera is not what makes it hard for AI.
Mirror metal is solved, even at scale. A camera struggles with reflective cookware because the metal honestly reflects the studio, the lights, and the crew, and you have to physically hide all of it. A model has no studio to hide, it invents a clean reflection from scratch, and all four did so coherently, with the curve-warping that sells it as real metal. This is jewelry's reflective-metal result on a far larger surface, and it held.
But the reflection is invented, and that is the real catch. The kitchen mirrored in the copper is a plausible scene the model generated, not your actual environment. It looks completely real, which is exactly why it is worth knowing: you cannot use the reflection to show a specific setting, and you cannot assume it is hiding nothing. For a hero shot it is a feature, a clean reflection for free. If the reflected setting has to be a real place, that is the one part to composite.
The pan is still generic. Each model produced its own proportions and handle style. For your exact SKU, the shape and rivet pattern are the design, so generate from a reference of the real pan.
How to shoot your cookware line without a light tent
The workflow is the roundup approach, applied to the subject real studios dread. Trust the reflection quality and choose on the copper warmth and reflection richness, Seedream for the most detailed. Judge the reflection for plausibility rather than accuracy, since it is invented, and if the setting matters, composite a real reflection. And for your exact pan, feed a reference photo so the shape and construction are yours rather than a generic invention.
With the Masonry CLI you can compare the reflections across models, or pass your real pan as a reference to keep the exact design:
masonry image "polished copper saucepan on a marble counter by a window, mirror reflection, soft light, photoreal" --model seedream-4-5 masonry image "place this exact pan on a sunlit kitchen counter, keep the shape and finish" --ref ./real-pan.png --model gpt-image-2
The bottom line
Cookware is a subject real studios dread and AI handles with ease. The hardest reflection in product photography, a large curved mirror showing the whole room, came out coherent and believable on all four models, with Seedream 4.5 the richest at the lowest cost. The one thing to remember is that the reflection is a hallucinated room, not your studio, beautiful and free, but invented. Choose on copper and reflection quality, use a reference for your exact pan, and composite a real reflection only if the setting must be specific. See how the same fidelity-first logic plays out across every product type in our best AI image model for product photography roundup, or run your own cookware from one place with the Masonry CLI.


