Which AI model works best for this prompt?

Models with reference image features, such as Midjourney (using --cref) or Stable Diffusion (with IP-Adapter/ControlNet), deliver the most consistent character replication.

Can I change the location in the prompt?

Yes, simply replace the `location` argument; it defaults to Taiwan MRT, but you can set it to any environment you need.

Will the generated image look exactly like my reference photo?

The AI strives for high similarity, but minor variations may occur. Increasing reference weight or using face restoration tools can improve accuracy.

GPT Image 2AI 提示词

Reference-Based Character Portrait: Taiwan MRT Scene Prompt

Easily place your character into a Taiwan MRT scene while keeping their look consistent with this AI prompt.

提示词 · EN

在 arglocation: 台湾捷运车厢 内，一位年轻女性坐在靠近车门的座位上。她的外貌和服饰必须参考上传照片中的女性（包括面部特征、气质、发型和穿搭风格；服装必须与照片完全一致）。她正低头专注于手机，神情自然，没有看向镜头。角色位于画面的右侧三分之一处，左侧有模糊的前景遮挡。

关于这个提示词

Easily place your character into a Taiwan MRT scene while keeping their look consistent with this AI prompt. 它更适合作为 GPT Image 2 的概念艺术起点：先保留画面结构、主体关系和风格约束，再替换成你的品牌、人物或场景。

使用时优先替换 location 等占位符，并保持镜头、构图和材质描述的顺序不变。这样更容易获得稳定的画面结果，也方便继续做多版本对比。

如何使用这个提示词

Copy the prompt text from this page.
Upload your reference image to a compatible AI model (e.g., Midjourney with --cref, Stable Diffusion with IP-Adapter).
Optionally, modify the location argument to change the scene (default: Taiwan MRT carriage).
Paste the prompt and generate your image.
Adjust reference strength (e.g., --cw in Midjourney) to fine-tune identity consistency.

Reference-Based Character Portrait: Taiwan MRT Scene Prompt

关于这个提示词

如何使用这个提示词

常见问题

更多来自 GPT Image 2

Minimalist Japanese Electric Nail Clipper Product Landing Page — Sora 2 Prompt

Indonesian Bird Contest Meme — Low-Quality Phone Photo

Futuristic Google Search Ad — AI Prompt

用 Sora2Hub 提示词更快创作。