A comparison of image generation models (GPT Image 2, Recraft, Seedream, Nano Banana Pro) using a detailed photorealistic prompt with a reference face.
提示词
Photorealistic photograph of an unmistakably adult woman. Use the supplied live-action reference photo as the absolute authority for her identity: exact facial structure, short silver-grey bob with straight bangs, hair color and texture, pale skin, real body proportions, silhouette, and overall presence. Preserve clear recognizability. do not replace her face, do not copy the reference photo’s original outfit. full-body standing portrait, smiling at the camera, hands on her skirt as it moves. Vertical crop from grass to above the head. WARDROBE: white off-shoulder lace maxi dress with ruffles and a flowing skirt, aviator sunglasses, simple jewelry and a watch. Do not copy clothing from the reference photo. SETTING: bright valley — fruit trees, wildflowers, a grazing deer, high waterfall, green mountains, birds in a clear sky. LIGHTING AND CAMERA: photoreal midday sun, natural pores, her real face readable under or around the sunglasses. Live-action photograph, 35mm, subject sharp, landscape clear. 2:3 or 3:4 portrait. One adult character only. No extra people, no text, no logos.