海报排版 / PROMPT
水墨双重曝光人物海报
从公开案例收录的海报排版提示词,适合参考其画面结构、风格控制和细节描述。
适用模型GPT Image 2
已实测灵感来源于 awesome-gpt-image-2 案例 359;当前提示词与效果图由 Fastgoing 使用 GPT Image 2 实测重新制作。
完整提示词COPY / USE
A cinematic character promotional poster of [SUBJECT], vertical composition (9:16), designed with a refined East-Asian ink aesthetic and high-end visual storytelling.
STRUCTURE:
Top-heavy hierarchical layout. The upper half features a large, highly recognizable silhouette of [SUBJECT]'s head / face / mask / upper body, forming a bold, iconic primary shape. The silhouette should be instantly identifiable.
The middle-lower section contains the full-body version of [SUBJECT] as a secondary subject, standing in a stable pose or subtle action stance, forming the visual core.
COMPOSITION STYLE:
Inside the large silhouette and around the character, use double exposure and collage storytelling. Integrate multiple elements:
- key scenes related to [SUBJECT]
- symbolic imagery and environment
- small narrative figures and interactions
- supporting visual motifs
Blend everything seamlessly using clouds, mist, ink diffusion, and negative space.
VISUAL FLOW:
Create a continuous flowing visual path from top to bottom, connecting:
- upper silhouette
- inner collage elements
- full-body subject
Ensure smooth eye guidance and compositional cohesion.
SIDE ELEMENTS:
Add balanced supporting elements on left and right sides to create tension, depth, and spatial variation.
STYLE & ATMOSPHERE:
- Large areas of negative space
- Ink-wash diffusion edges, soft fading, subtle fragmentation
- Eastern aesthetic: balance of emptiness and detail
- Calm, premium, restrained, cinematic tone
QUALITY:
Ultra-detailed, high resolution, layered depth, soft lighting, atmospheric perspective, cohesive series-style design.
OUTPUT:
9:16 aspect ratio, poster-ready composition.复制后,可参考下方“建议替换”项目改成你的主题;带【方括号】的内容请优先替换。
PROVENANCE / 来源与验证
它从哪里来,是否完成实测?
灵感来源于 awesome-gpt-image-2 案例 359;当前提示词与效果图由 Fastgoing 使用 GPT Image 2 实测重新制作。
GPT Image 2 · 2026-08-18 实测HOW TO USE
不用学提示词工程,也能改成自己的。
建议替换
01【SUBJECT】使用小贴士
先只替换变量跑一次,观察最不符合预期的部分;第二轮只补充那一项约束,通常比一次塞进很多要求更稳定。
