中文版提示词
电影级历史现实主义实景摄影，晚清至民初时期（约19世纪末至20世纪初），白天室外街巷集市。主体为密集市集人群，包括传统中式商贩、行人、儿童，以及手推木车、食物摊位、红色商业旗幡。环境为灰砖四合院建筑群，带有牌楼拱门、木雕门窗、灰瓦屋顶、中文招牌与灯笼，构成层叠民居与街巷结构。
摄影构图采用约2:1宽幅比例，全景建立镜头，机位近似平视但略高，利用前景人物与摊位、中景街市活动、后景建筑拱门形成多层纵深，街道透视线与屋檐线引导视线，中央拱门作为视觉焦点。色彩以低饱和棕褐、灰绿、土黄与暗红为主，营造复古历史调性。光线为阴天漫射光，柔和低反差，厚云层过滤的自然日光配合环境反射光，辅以店铺内部微弱暖色灯光。无明显数字特效，强调实景置景、群众调度与烟尘环境质感，后期为电影级写实渲染与数字调色，保留细腻胶片颗粒、自然材质纹理及轻微历史影像褪色质感。
English Prompt
Cinematic historical realism live-action photography, set in the late Qing to early Republican period (circa late 19th to early 20th century), daytime exterior street market. The scene features a dense marketplace crowd including traditional Chinese vendors, pedestrians, children, wooden handcarts, food stalls, and red commercial banners. The environment consists of grey-brick courtyard houses with archway gates, carved wooden doors and windows, grey-tile roofs, Chinese signage, and lanterns, forming layered residential blocks and alleyways.
Cinematographic composition adopts an approximately 2:1 widescreen ratio, with an establishing wide shot, a slightly elevated eye-level camera angle, utilizing foreground figures and stalls, midground street activities, and background architectural arches to create multi-layered depth. Street perspective lines and rooflines guide the viewer's eye, with the central archway serving as the visual anchor. Color palette is restrained to low-saturation brown, grey-green, earthy yellow, and dark red for a vintage historical tone. Lighting is overcast diffuse light—soft, low-contrast, filtered daylight through thick clouds combined with ambient fill and faint warm light from shop interiors. No obvious digital effects; emphasizes practical set design, crowd choreography, and atmospheric dust. Post-production applies cinematic-grade rendering and digital color grading, with fine film grain, natural material textures, and a subtle historical faded-image quality.
该提示词已在以下工具中验证有效：
- Midjourney – 建议使用
--ar 2:1 --style raw --s 200，搭配--v 6.0或--v 6.1，可添加--stylize 250增强环境氛围。 - Stable Diffusion – 推荐写实场景模型（如 Juggernaut XL、Realistic Vision），配合 ControlNet（Depth 和 Lineart）锁定透视与层次，开启高分辨率修复（放大1.5~2倍）。
- DALL-E 3 – 直接输入全文，生成后可裁剪为2:1比例，注意强调“多层纵深”和“中央拱门”以保证构图。
- 文心一格 – 对中式古建与市井场景理解良好，建议选择“写实摄影”风格并提高画质细节。
- 通义万相 – 中文语义响应准确，可完整保留人群与建筑元素，推荐输出宽幅尺寸（如 2048×1024）。
- Adobe Firefly – 适合后期调整光影与褪色质感，初稿生成后可进一步增强胶片模拟。
- NovelAI – 若需强化人物面部与服饰细节，可借助该工具进行局部细化。
使用建议：针对不同工具添加负面提示词（如“过度饱和、清晰边缘、对称呆板、现代元素”），采样步数推荐50~70步以获得最佳景深与颗粒平衡。
◯ 评论 0