UI Mockup Difficulty 5/5 ↻ Reusable
Time-Traveling Filter — Old Photo Restoration
穿越时空的滤镜——老照片修复
Our take
Reusable template with placeholders. Structured JSON prompt with controllable parameters. Strong visual evidence.
The prompt
{
"type": "augmented_reality_filter_synthesis",
"aspect_ratio": "16:9",
"instruction": "CRITICAL: This is an IMAGE-TO-IMAGE task, not a text-to-image generation. You must strictly preserve the pixels of the {{UPLOADED_IMAGE_FILE}} as the background. Do not generate a new scene based on description. Only the area INSIDE the glass filter is allowed to change.",
"global_constraints": {
"source_fidelity": "The background outside the filter must be an EXACT pixel-match to {{UPLOADED_IMAGE_FILE}}. Do not alter, crop, expand, or 'reimagine' the background context.",
"context_isolation": "Ignore all previous images or conversation topics. Focus EXCLUSIVELY on the current uploaded file.",
"modification_scope": "modifications are STRICTLY limited to the area inside the glass filter lens."
},
"dynamic_analysis_module": {
"step_1_focal_point_extraction": {
"task": "Identify the coordinates of the main subject (Face/Object) in the current file to position the filter.",
"output_variable": "[TARGET_COORDINATES]",
"logic": "Find the center of interest solely within the provided image."
},
"step_2_filter_specs": {
"task": "Determine filter parameters based on the subject.",
"output_variable_shape": "[FILTER_SHAPE]",
"output_variable_style": "[HD_RESTORATION_STYLE]",
"logic": "Face -> Circular/Oval + Skin Restoration. Building -> Rectangular + Structure Sharpening."
}
},
"system_prompt_logic": {
"main_premise": "A first-person view of a hand holding a [FILTER_SHAPE] glass lens over the actual physical print of {{UPLOADED_IMAGE_FILE}}.",
"layer_definition_logic": {
"layer_1_background_immutable": {
"z_index": "Background (The Original Image)",
"content": "The original {{UPLOADED_IMAGE_FILE}} displayed as a large physical print or wall projection.",
"strict_rules": {
"NO_GENERATION": "Do not generate new scenery. Use the uploaded image as a fixed texture map.",
"NO_EXPANSION": "Do not attempt to extend the image borders. Show the image exactly as provided, framing it to fill the background.",
"visual_state": "Must retain all original artifacts: noise, blur, scratches, B&W tone, and low resolution. This is the 'Before' state."
}
},
"layer_2_foreground_hand": {
"z_index": "Foreground (Overlay)",
"content": "A realistic human hand holding a [FILTER_SHAPE] optical glass lens.",
"positioning": "The hand enters from the edge (unobtrusive). The glass lens is positioned directly over [TARGET_COORDINATES].",
"lighting": "The hand casts a subtle shadow onto the background print to create depth."
},
"layer_3_filter_throughput": {
"z_index": "Lens Interior (The Transformation)",
"content": "The specific pixels visible *through* the glass lens.",
"visual_transformation": {
"action": "Upscale and colorize ONLY the pixels inside the lens boundary.",
"style": "[HD_RESTORATION_STYLE] (4K resolution, realistic textures, vibrant lighting).",
"contrast_lock": "The separation is absolute. Outside the line = Old/Blurry. Inside the line = New/Sharp. No fading gradients."
}
}
}
}
} Aspect ratio
16:9
How to use
This prompt is a reusable template. Replace the
placeholders (e.g. [INSERT CITY NAME]) with your own
values before sending to your image-generation model.
Source
Similar prompts
See all →
AI Expert Imagery
1. Open ChatGPT / Gemini AI / Grok AI
Transparent Type Garment Tutorial — NanoBanana Ver1.2 Guide
服を文字に変えて透過させる件、Hailuo以外のnanobananaでも精度高くする方法を探ってました。どうも服の色を抽出して文字色に置き換えるところがなかなか難しいらしく、手間ですが、服の色をカラー
Hairstyle Design Prompt Generator
Generate an extremely detailed [hairstyle name, such as: American front thorn] hairstyle design guid


