Old photos hold our most precious memories. Family portraits from decades past, wedding pictures from our grandparents' era, and childhood snapshots that capture moments we never want to forget. But time takes its toll on physical photographs. They fade, tear, blur, and sometimes break into pieces.
Modern AI technology offers a solution. Tools like GPT-Image and Nano Banana can restore damaged photos with stunning accuracy. These AI models analyze damaged areas and reconstruct missing details while preserving the original character of the image.
This guide shows you five proven prompts for restoring different types of photo damage. Each prompt targets specific restoration challenges, from simple color correction to complex reconstruction work.
Key Takeaways
AI photo restoration delivers several key benefits for preserving your memories:
Speed and Efficiency: Traditional photo restoration requires hours of manual work. AI tools complete the same process in minutes.
Cost-Effective Solution: Professional restoration services cost hundreds of dollars per photo. AI tools provide similar results at a fraction of the price.
Consistent Quality: Human restorers vary in skill and technique. AI models apply consistent standards across all images.
Accessibility: You can restore photos from home without shipping valuable originals to restoration services.
Preservation of Original Character: Advanced AI models maintain the authentic look and feel of historical photographs while removing damage.
Tools We Will Use
This guide focuses on two powerful AI restoration tools:
GPT-Image (Inside ChatGPT)
GPT-Image excels at understanding complex restoration instructions. It analyzes photos holistically, considering composition, lighting, and historical context. This tool works best for:
- Black and white photo colorization
- Facial feature restoration
- Complex damage repair requiring creative reconstruction
Nano Banana (Inside Google Gemini)
Nano Banana specializes in technical image processing tasks. It delivers precise results for structural repairs and cleanup work. This model performs best for:
- Distortion correction
- Broken piece reconstruction
- Detail sharpening and blur removal
You can also use our AI Image Editor here which is also free.
List of Photo Restoration Prompts
1. Restore Black and White Image
Recommended Model: GPT-Image
"Restore this attached image, just the person in the portrait not the entire frame. Analyze the person in portrait, the expressions, facial features like eyes, nose, eyebrows, skin, hair etc.
Option #1 - Restore with the following colors [colors]
Option #2 - Restore the image added colors as per your choice keeping in mind that this image is taken [50 years] ago, so here are some instructions to follow:
Keep the expression as it is
Keep every facial structure like in the attached image.
Keep all the elements you see in the image
Don't change anything about person in the portrait.
Don't change the style, aesthetics of the person in the portrait.
Don't add superficial expressions.
Don't change or alter any part of face like eyes, nose, hair etc.
Don't change the clothing style, keep exactly like in the picture."
Winner: GPT-Image - Superior understanding of historical color palettes and facial analysis
2. Restore Distorted Image
Recommended Model: Nano Banana
"Restore this attached image, removing distortion, cleaning up the mess. Analyze the person in portrait, the expressions, facial features like eyes, nose, eyebrows, skin, hair etc.
Option #1 - Restore with the following colors [colors]
Option #2 - Restore the image added colors as per your choice keeping in mind that this image is taken [50 years] ago, so here are some instructions to follow:
Keep the expression as it is
Keep every facial structure like in the attached image.
Keep all the elements you see in the image
Don't change the style, aesthetics of the person in the portrait.
Don't add superficial expressions.
Don't change or alter any part of face like eyes, nose, hair etc.
Don't change the clothing style, keep exactly like in the picture."
Winner: Nano Banana - Excellent geometric correction and distortion removal capabilities
3. Restore Blurred Image
Recommended Model: GPT-Image
"Restore this attached image, removing the blur and restoring the original image. Analyze the picture in deep, for example, aesthetics, expressions, elements, shadows, background etc.
Here are some additional info: [There's white rose on left pocket, it's in the marriage hall, camera flash on face].
Don't add colors, just restore the image in its original form keeping each and every pixel as in original."
Winner: GPT-Image - Better contextual understanding and detail reconstruction
4. Restore Broken Pieces Image
Recommended Model: Nano Banana
"Restore this attached image, it's in broken pieces, keeping in its original form. Analyze each broken pieces, for example, aesthetics, expressions, elements, shadows, background etc.
Instructions: Color correction to its original form, I don't want to add colors, just put the pieces together keeping the original colors, vibe to it. Don't change anything about the image. Don't add any elements by yourself. Don't add any scratches, frames etc just restore original in portrait
Don't add scratches in the output image, just restore the attached image as portrait in its true form"
Winner: Nano Banana - Superior pattern matching for piece reconstruction
5. Restore Damaged Image
Recommended Model: GPT-Image
"Restore this attached image, it's distorted and scratched, I want you to analyze the image and restore to its original form. Analyze the image for its aesthetics, expressions, elements, shadows and background.
Some part of the image cannot be restored, therefore add subtle imagination to it matching the whole pattern of the image.
Don't overdo the imagination, just keep it subtle.
Don't add colors, just restore this image to its true form and color."
Winner: GPT-Image - Better creative reconstruction while maintaining authenticity
Best Practices for AI Photo Restoration
Choose the Right Model
Select GPT-Image for complex restoration tasks requiring creative interpretation. Use Nano Banana for technical corrections and precise reconstruction work.
Provide Clear Context
Include relevant details about the photo's age, setting, and visible elements. This information helps the AI make accurate restoration decisions.
Specify What to Preserve
Clearly state which elements must remain unchanged. This prevents the AI from making unwanted modifications to important features.
Start with Less Damaged Photos
Practice with photos that have minor damage before attempting complex restoration projects. This helps you understand each model's capabilities.
Review and Iterate
Check the restored image carefully. If needed, provide additional instructions to refine specific areas.
Save Your Originals
Always work with copies of your original photos. Keep the damaged originals safe as backup references.
Use High-Quality Scans
Scan damaged photos at high resolution (at least 600 DPI) for best restoration results. Higher quality inputs produce better outputs.
Frequently Asked Questions
Can AI restore severely damaged photos?
Yes, but results vary based on damage type. AI excels at removing scratches, correcting colors, and filling small gaps. Extensive damage requiring significant reconstruction may produce less accurate results.
How do I choose between GPT-Image and Nano Banana?
Use GPT-Image for creative restoration tasks like colorization and facial reconstruction. Choose Nano Banana for technical fixes like distortion correction and piece assembly.
Will AI change the person's appearance?
Both models are designed to preserve original features when given proper instructions. Always specify that facial features, expressions, and clothing should remain unchanged.
How long does AI photo restoration take?
Most AI restorations complete within 2-5 minutes. Complex reconstructions may take slightly longer, but results are still much faster than manual restoration.
Can I restore multiple photos at once?
Both tools work with individual images. For batch processing, you'll need to restore photos one at a time using the appropriate prompts.
What file formats work best?
JPG and PNG formats work well with both AI models. Higher resolution images generally produce better restoration results.
How accurate is AI color restoration?
AI color restoration analyzes historical context and makes educated guesses about original colors. Results are generally accurate for common items like skin tone and clothing, but may require refinement for specific color preferences.
Can AI fix photos with missing pieces?
Yes, both models can reconstruct missing areas. Nano Banana excels at putting broken pieces back together, while GPT-Image is better at filling gaps with contextually appropriate content.
AI photo restoration makes preserving precious memories accessible to everyone. These five prompts give you the tools to restore different types of photo damage while maintaining the authentic character of your historical images.
