图像生成
专业照片修复专家
Professional photo restoration expert
去除老旧照片中的划痕、灰尘、指纹和水渍,修复边缘缺损与面部细节,并校正几何变形和对比度。需要进行保守、高精度修复,并以受1940年代 Kodachrome 启发的柔和暖色和自然颗粒感进行历史着色时使用。
- 测试模型:
- GPT-4o
- 有效性评分:
- 90/100
- 提示词归档
中文版本
Act as a professional photo restoration expert. You are tasked with performing a high-precision conservative restoration and historical colorization of a degraded vintage photograph. The final image should resemble a perfectly preserved original print. **IMAGE ANALYSIS & RESTORATION:** 1. **Surface Repair:** - Digitally remove deep scratches, dust, fingerprints, and moisture stains. - Reconstruct missing areas or tears at the edges while preserving the texture of the photographic paper. 2. **Structural Fidelity:** - Correct geometric distortion. - Restore the original contrast without overexposing highlights or excessively darkening shadows. 3. **Facial Clarity:** - Recover facial features with extreme precision. - Avoid the "wax skin" effect; maintain the natural grain and original micro-expressions. **CHROMATIC & AESTHETIC STYLE:** 1. **Historical Color Palette:** - Apply a realistic colorization inspired by the Kodachrome process of the 1940s. - Use soft, warm, and desaturated tones. 2. **Skin Tones:** - Render skin tones naturally, considering the period's ambient lighting. - Avoid uniform digital saturation. 3. **Authentic Grain:** - Preserve a fine, organic photographic grain typical of 35mm analog film. **NEGATIVE PROMPT / WHAT TO AVOID:** - Do not apply modern filters such as Instagram. - Avoid "smooth" or "plastic skin" effects. - Refrain from using neon colors, excessive saturation, or sharpening artifacts (e.g., white halos). - Prevent the appearance of a digital painting or 3D illustration. **FINAL OUTPUT QUALITY:** - Achieve a photorealistic, museum-quality finish with ultra-defined detail (8k resolution style) and absolute historical fidelity.
填入内容并运行
- 在 ChatGPT 中打开 (在新标签页中打开)
- 在 Claude 中打开 (在新标签页中打开)
- 在 Gemini 中打开 (在新标签页中打开)
Gemini 无法通过链接接收提示词,请先复制。
- 在 Perplexity 中打开 (在新标签页中打开)
将在新标签页打开并已填入提示词。可能需要先登录该服务。
如何使用此提示词
- 发送前,请将方括号中的内容替换为你自己的信息。
- 不同模型的结果会有差异。若输出不理想,可换一个模型或补充一个具体示例。
- 请将提示词放在一条消息中发送。分多次发送会削弱指令效果。