返回 Prompt 库
Instruction: Treat The Supplied Image As The Single-Source Reference For
Instruction: Treat the supplied image as the single-source reference for content, composition, and identity only. Produce a completely new, single photorealistic image that looks like a professional studio re-shoot on...

Prompt
Instruction: Treat the supplied image as the single-source reference for content, composition, and identity only. Produce a completely new, single photorealistic image that looks like a professional studio re-shoot on a modern full-frame DSLR. Do not return the original image or minor edits. Generate a fresh, high-resolution output that is artifact free and studio-lit. Steps to perform, executed autonomously: 1. Analyze: detect subjects, faces, objects, and all damage types like scratches, tears, blur, stains, missing regions, compression artifacts and more. 2. Extract and lock identity: create a structural face/body map. Preserve facial landmarks, proportions, age, and distinctive features exactly. 3. Reconstruct and reimagine: inpaint missing or cut-off areas using surrounding evidence, preserving original composition and perspective. Where information is ambiguous, use physically plausible fills that match era-appropriate clothing and objects. 4. Re-light and studio-grade finish: apply professional three-point or cinematic lighting, accurate shadows and highlights, natural color grading, and correct depth of field to simulate a DSLR lens. 5. Enhance textures and details: restore pores, hair strands, fabric weave, surface grain for objects, and remove all noise and compression artifacts. 6. Upscale: produce final image at 4K or higher, with true high-frequency detail and no plastic smoothing or artificial sharpening. Hard constraints: - Identity lock: facial structure, age, skin tone, key marks are preserved 100 percent. - Composition lock: overall framing and relative object positions preserved. - No added anachronistic items or changes of identity. - Final must be artifact free, not a retouched original. Treat as a new photo captured in controlled studio. - No descriptive text, overlays, or watermarks. Negative prompts: - low detail, plastic skin, oversmoothed skin, new faces, different clothing, visible noise, haze, compression artifacts, color shifts, cartoonish rendering, added objects not derivable from source. Deliverable: - Single new high-resolution photorealistic image that reads as a modern studio re-shoot of the original subject.
如何使用这个 Prompt
- 用上方的复制按钮复制完整提示词。
- 把方括号 [ ] 里的占位内容替换成你自己的主体、场景或细节。
- 粘贴到 Gemini / Nano Banana Pro 生成,或直接点上面的「试试这个 Prompt」在本站免费生成,再对照示例图微调用词。
风格说明
- Photo
- 要求模型输出照片级真实感:自然光线、可信的皮肤与材质、类似相机的景深,而不是插画或 3D 质感。
- 3D Render
- CGI 质感:表面平滑、全局光照可控,纵深像 Blender/C4D 渲染而非照片。
- Hero Portrait
- 把单一主体拍成主视觉:紧凑构图、有力的眼神或姿态,背景简化,让人物撑起整张画面。
- Character Design
- 角色设定图:剪影可读、服装道具细节完整,常在一张图里呈现多角度或多表情。
- Cinematic
- 电影感渲染:宽画幅构图、强烈的光比、调色(常见青橙对比),画面自带叙事氛围。
- Cartoon
- 造型简化、比例和表情夸张、平涂或赛璐璐上色——画面易读,角色风格也容易保持一致。
- Studio Lighting
- 可控的人造光(柔光箱、轮廓光):阴影干净、高光可控,是商业人像和产品摄影的标准打光。
常见问题
这个 Prompt 怎么在 Gemini(Nano Banana Pro)里使用?
点击上方的复制按钮拿到完整提示词,粘贴到 Gemini 或任何支持 Nano Banana Pro 的工具里,再对照本页示例图检查效果。也可以直接点「试试这个 Prompt」在本站生成——每天有免费次数,无需注册。
可以修改「Instruction: Treat The Supplied Image As The Single-Source Reference For」这个提示词吗?
可以。方括号 [ ] 里的内容是占位参数,本来就应该替换成你自己的主体或细节;主体、光线、风格关键词也都可以自由调整,措辞上的小改动往往带来明显不同的结果。
这个提示词适用于哪些 AI 模型?
它最初是为 Google 的 Gemini 图像模型(含 Nano Banana Pro)分享的。大多数文生图模型也能理解,但构图和文字渲染质量会因模型而异。





