产品视觉 GPT Image 2 提示词：Stacked Corridor Model Comparison

A side-by-side comparison image presented as two stacked documentary-style photographs of the same cramped old Chinese apartment corridor, framed by a black filmstrip border with white sprocket holes on both left and right edges. The top half is labeled {argument name="top label" default="GPT-image-2"} in large bold white sans-serif text centered near the top, and the bottom half is labeled {argument name="bottom label" default="Nano Banana 2"} in the same style. Create a gritty, nostalgic, cinematic scene in a long, narrow shared hallway of a worn residential building from the late 20th century, with peeling paint, stained walls, faded propaganda posters and old newspapers on the walls, dark wooden doorways, hanging towels and laundry, a damp concrete floor, and daylight coming from a window at the far end. In the upper scene, show exactly 4 people: 1 adult standing at the left cooking over a small stove on a wooden table with visible steam, 1 adult man sitting deeper in the corridor beside an electric fan, 1 child crouching in the center playing near a basin on the wet floor, and 1 adult woman seated at a small table on the right writing or reading; include enamel mugs, metal pots, stools, buckets, thermos containers, and cluttered household items. In the lower scene, show exactly 4 people again: 1 adult woman at the left cooking at a stove with smoke rising, 1 adult man seated on the right in a bamboo chair leaning over a tabletop fan, 1 child crouching in front writing on paper on a low stool, and 1 distant adult figure near the back window; add more visible corridor clutter including a bicycle front wheel at lower left, wash basins, baskets, thermoses, shelves, shoes, and hanging clothes. Keep the camera straight-on and centered to emphasize the corridor depth, with realistic human proportions, candid slice-of-life behavior, muted aged colors, soft natural light, high detail, photojournalistic realism, and a subtle difference in arrangement between the top and bottom scenes while preserving the same overall environment and mood.