原始文档
GPT Image 2 · en
Goal: Create a vertical three-panel comparison collage showing the same Chinese traditional aesthetic portrait generated in three different styles.
Canvas: Portrait format, approximately 4:5 aspect ratio, 960×1200 feel. Divide the image into exactly 3 equal horizontal panels separated by thin white horizontal lines.
Overall subject: In all three panels, show the same young East Asian woman in a warm, cinematic ancient Chinese interior, wearing a soft ivory hanfu robe. She is adjusting a peach-pink flower hair ornament near her ear with both hands. Her black hair is loosely pinned up with wispy strands around her face. The atmosphere is golden-hour, backlit, romantic, and traditional, with wooden lattice windows behind her, soft haze, shallow depth of field, warm sunlight, and floating flower petals or dust motes.
Panel count and labels: Use exactly 3 panels with large white sans-serif text labels in the lower-left area of each panel: 1) “Midjourney V8.2” on the top panel, 2) “GPT-Image 2” on the middle panel, 3) “Seedream 5.0 Pro” on the bottom panel.
Top panel: Label “Midjourney V8.2”. Make this the most dramatic and cinematic version: stronger contrast, moody shadows, glowing rim light, more dynamic pose, windblown hair, petals crossing the foreground, refined facial features, realistic skin texture, premium film still quality. The woman faces slightly toward camera with a serious, elegant expression while placing the flower by her ear.
Middle panel: Label “GPT-Image 2”. Make this cleaner and softer, with a slightly polished beauty-retouch look. The woman is turned three-quarters to the right, looking sideways while carefully adjusting the flower. Keep warm window light, cream hanfu fabric with subtle embroidered floral details, and a more smooth, influencer-like face.
Bottom panel: Label “Seedream 5.0 Pro”. Make this brighter and slightly more plastic/smooth than the top, with a gentle smile and softer realism. The woman holds the peach flower close to her ear, looking down-left or sideways, with intense backlight from the lattice window and a glowing edge around her hair and hands.
Text styling: Large white text, simple sans-serif, no outline, subtle shadow if needed for readability. Place each label consistently near the bottom-left of its panel. Do not add any other text, logos, watermarks, or UI elements.
Customizable details: The subject is {argument name="subject description" default="a young East Asian woman in ivory hanfu"}; the setting is {argument name="setting" default="a warm ancient Chinese room with wooden lattice windows"}; the hair accessory is {argument name="hair flower" default="a peach-pink flower"}; the top label is {argument name="top model label" default="Midjourney V8.2"}; the middle label is {argument name="middle model label" default="GPT-Image 2"}.使用方法
- 复制完整原始提示词，并保留其中的参数。
- 按需求替换主体、文字和构图要求。
- 在兼容的外部工具中使用，核对当前使用条件并检查结果。
开始前需了解
- 先确认是否需要参考图。标注兼容模型，不代表该示例由这个模型生成。
- 核对名称、数字与画面文字；不同模型版本的结果可能不同。
