漂亮可爱的人物墙壁涂鸦，效果如图，用的是 GPT image2，不多废话，提示词如下：
Use the uploaded reference image as the strict identity and outfit reference. Preserve the face, facial features, skin tone, hairstyle, hair colour, accessories, outfit, and overall vibe exactly as shown.
Create a high-quality vertical (4:5 or 9:16) mixed-media portrait featuring:
A realistic full-body version of the person. A black hand-drawn doodle-shadow of the same person on the wall beside them.The real person should look cute, slightly embarrassed, playful, and naturally copying a new random mischievous pose each generation. The doodle-shadow should perform the same pose idea in a much more exaggerated, chaotic, cartoonish way, with manga motion lines, stars, hearts, and sparkles.
Use a clean white/cream studio wall, minimal background, soft natural lighting, and keep both figures fully visible. The doodle should clearly resemble the real person through hairstyle, accessories, outfit silhouette, and pose.
Negative Prompt: outfit changes, identity changes, realistic second person, normal shadow, horror, anime human, cluttered background, repeated pointing or finger-gun poses, stiff pose, extra limbs/fingers, distorted body, text, watermark, logo, AI artifacts.
翻译成中文大概就是：
使用上传的参考图像作为严格的身份和服装参考。精确保留面部、五官、肤色、发型、发色、配饰、服装以及整体氛围，如图所示。
创建一幅高质量的竖构图（4:5 或 9:16）混合媒介肖像，包含：
现实主义的全身人物版本。墙上旁边同一人物的黑色手绘涂鸦阴影。真实人物应看起来可爱、略带尴尬、俏皮，并在每次生成时自然模仿一个新的随机调皮姿势。涂鸦阴影应以更夸张、混乱、卡通化的方式执行相同的姿势概念，带有漫画运动线条、星星、心形和闪光效果。
使用干净的白色/奶油色工作室墙壁、最小化背景、柔和自然光线，并确保两个形象完全可见。涂鸦应通过发型、配饰、服装轮廓和姿势清楚地酷似真实人物。
负面提示：服装变化、身份变化、现实主义第二人称、普通阴影、恐怖、动漫人类、杂乱背景、重复指向或指枪姿势、生硬姿势、多余肢体/手指、扭曲身体、文字、水印、标志、AI 瑕疵。
