AI Photo Restoration Guide

Old Photo Restoration AI Prompt: Restore Photos Without Changing Faces

Old photographs can contain details that no modern camera can recreate. AI can help repair scratches, fading, stains and damaged areas — but the real challenge is restoring the photograph without accidentally changing the people inside it.

Quick Answer The best old photo restoration AI prompt tells the image model exactly what to repair while clearly defining what must remain untouched: facial identity, expressions, apparent age, clothing, composition, lighting and the original character of the photograph.

A simple instruction such as “restore this old photo” sounds straightforward, but it gives an AI image model a surprising amount of freedom.

It may sharpen a face that was originally soft, recreate damaged facial features, change clothing, alter expressions or make an old family photograph look like a modern studio portrait.

That is why a good old photo restoration AI prompt should prioritize preservation before enhancement.

What Can AI Restore in an Old Photograph?

Depending on the original image and the AI model being used, modern image editing tools can help with several common forms of photographic damage.

01 Surface Damage Repair dust, scratches, stains, small marks and minor creases.
02 Image Quality Improve faded contrast, softness and some lost photographic detail.
03 Color Correct faded color photographs or optionally colorize monochrome images.

Why Face Preservation Matters More Than Sharpness

When restoring a family photograph, the face is usually the most important part of the image. Increasing sharpness is useful only if the recovered detail still belongs to the original person.

Generative AI can sometimes fill missing facial information with something that looks realistic but was never actually present in the photograph. That creates an important distinction between restoration and AI reconstruction.

The restoration rule:

If the AI is uncertain about a damaged detail, preserving the original ambiguity is often better than inventing a visually impressive replacement.

Copy-Paste Old Photo Restoration AI Prompt

Upload the original photograph to your preferred image-editing model and use the prompt below. It is intentionally conservative: the goal is to repair the photograph rather than redesign it.

Ready-to-Copy Prompt Old Photo Restoration
Restore the uploaded old photograph with a conservative, realistic photo-restoration approach. Treat the uploaded photograph as the definitive visual reference. Preserve the original composition, people, facial identities, expressions, apparent ages, hairstyles, clothing, poses, objects, architecture, background and overall photographic character. Repair only visible damage and degradation. Remove dust, scratches, stains, minor creases, faded areas, discoloration and scanning artifacts while reconstructing missing detail only when it can be reasonably inferred from surrounding information. IDENTITY PRESERVATION: Every person must remain recognizable as the same person shown in the original photograph. Preserve facial geometry, facial proportions, eye shape, eyebrows, nose, lips, jawline, cheek structure, hairline, apparent age, expression and natural asymmetry. Do not redesign, beautify, rejuvenate or replace anyone's face. If a face is partially damaged, reconstruct only the minimum missing information necessary to connect the damaged area naturally with the surrounding facial structure. Do not invent a new identity. SKIN: Preserve realistic human skin texture appropriate to the original photograph. Do not create plastic-looking skin, excessive smoothing, artificial pores or modern beauty-retouching. DAMAGE REPAIR: Remove visible scratches, dust, stains, paper marks, minor tears, fading and uneven exposure. Reconstruct damaged areas using nearby visual information while avoiding unnecessary changes. DETAIL RECOVERY: Gently improve clarity in faces, hair, clothing and important objects. Do not oversharpen. Do not create artificial facial details that were not supported by the original image. CLOTHING: Preserve the original clothing exactly in design, structure and historical character. Do not replace garments with modern clothing. BACKGROUND: Preserve the original environment, architecture, furniture and objects. Repair damaged areas without adding new people or objects. LIGHTING: Preserve the original lighting direction, shadows, highlights and exposure character. Do not introduce modern cinematic lighting. COLOR: If the source is a color photograph, restore faded colors naturally with restrained saturation. If the source is black and white, preserve the monochrome appearance unless colorization is specifically requested. FILM CHARACTER: Retain an appropriate amount of authentic photographic grain and texture. Remove destructive damage without removing the historical character of the photograph. REALISM: The final result should look like the same physical photograph after careful professional restoration and scanning, not like a newly generated modern portrait. NEGATIVE CONSTRAINTS: No face replacement, no identity change, no age change, no beautification, no rejuvenation, no modern clothing, no changed expressions, no changed poses, no new people, no removed people, no background redesign, no cinematic color grading, no HDR, no plastic skin, no excessive sharpening, no artificial eyes, no CGI appearance, no illustration, no cartoon style, no text, no watermark. FINAL PRIORITY: Preservation is more important than perfection. When information is uncertain or missing, make the smallest plausible repair and preserve the original ambiguity rather than inventing new visual information.

How to Use an AI Photo Restoration Prompt

The workflow is simple, but the order matters.

  1. Start with the highest-quality scan or photograph of the original image.
  2. Upload the image to your chosen AI image editor.
  3. Paste the restoration prompt.
  4. Review the face and important objects against the original.
  5. If something changed incorrectly, make a targeted follow-up edit instead of regenerating the entire photograph.

How to Get Better Results From Old Photo Restoration

Use the Original Photograph Whenever Possible

If you have the physical print, scan it carefully at a useful resolution. If scanning isn't available, photograph the print flat with even lighting and minimize reflections.

Describe the Actual Damage

Instead of relying only on the phrase “restore this photo,” identify what needs fixing.

Remove the long scratch crossing the upper-left background while preserving the person's hair and face exactly. Reduce the yellow color cast while keeping natural skin tones. Repair the crease across the lower-right corner without changing the clothing or background.

Targeted instructions give the AI a clearer editing boundary.

Don't Ask AI to Make the Person Look Younger

Restoration and beautification are different goals. If the original photograph shows an older version of someone, preserve that appearance unless you are intentionally creating a separate creative edit.

Don't Over-Sharpen Faces

A blurry old photograph does not contain unlimited hidden detail. Asking for extreme sharpness can encourage the model to invent facial features. Gentle detail recovery usually produces a more believable restoration.

Restoration vs Colorization

These two operations are often confused, but they solve different problems.

Restoration

Attempts to repair existing photographic information such as scratches, fading, stains, dust and minor damage.

Colorization

Adds estimated color information to an originally monochrome photograph. Unless the original colors are known, colorization is an interpretation.

Black-and-White Photo Colorization Prompt

If you specifically want to add realistic color to an old black-and-white photograph, use a separate instruction rather than combining every possible edit into one request.

Copy-Paste Prompt Natural Colorization
Naturally colorize the uploaded black-and-white photograph while preserving the original photograph as accurately as possible. Preserve every person's identity, facial structure, expression, apparent age, hairstyle, clothing design, pose and original composition. Repair minor scratches, dust and visible damage without redesigning the image. Use realistic, historically appropriate colors with restrained saturation and natural skin tones. Preserve the original lighting direction, shadows, highlights and photographic atmosphere. Do not modernize clothing, architecture, hairstyles or objects. If the original color of an object is unknown, choose a restrained plausible color rather than an unusually vivid color. Do not alter faces, expressions or identities. Do not beautify, rejuvenate, smooth skin, change body proportions or create a modern DSLR appearance. The finished image should look like a naturally colorized version of the original photograph, not a newly generated photograph.

How to Restore a Severely Damaged Family Photograph

Severe damage requires more caution because some information may simply no longer exist in the source image.

If a scratch crosses an entire face or part of the photograph is physically missing, AI may have to reconstruct information. In that situation, the result should be treated as an AI-assisted reconstruction rather than a guaranteed recovery of the original visual information.

For irreplaceable family photographs, keep the original scan untouched and create a separate restored version.

Common AI Photo Restoration Mistakes

1. Using Only “Make It HD”

Resolution language doesn't explain what should actually be repaired. Describe the damage, identity requirements and desired photographic character instead.

2. Removing Every Trace of Film Texture

Old photographs naturally have texture. Removing everything can make the result look like a modern AI render.

3. Using Extreme Saturation

A faded photograph doesn't necessarily need bright colors. Keep color restoration restrained unless you have reliable information about the original colors.

4. Allowing AI to Redesign the Background

If the photograph is historically important, the background can be just as meaningful as the people. Explicitly tell the model to preserve architecture, furniture and objects.

ChatGPT, Gemini and Nano Banana for Photo Restoration

Modern image-generation systems can interpret uploaded photographs and natural-language editing instructions, but their exact behavior varies by model and image.

For restoration work, the most useful capability is not simply image generation. You want an editor that can understand the uploaded reference, follow precise instructions and make controlled changes.

This prompt can therefore be adapted for ChatGPT image editing, Gemini, Nano Banana and other capable AI image editors.

Using a Ready-to-Copy Prompt From Nano AI Prompts

Nano AI Prompts is a free, independent library of practical AI image prompts and photo-editing prompts for tools such as ChatGPT, Gemini and Nano Banana.

When you open a relevant prompt page, you'll find the complete prompt inside the dedicated prompt box. If the prompt is long, simply scroll inside the prompt box and use the visible Copy Prompt button to copy the complete prompt instantly.

You can then paste it into the relevant AI image generator or photo-editing tool and customize the instructions for your photograph.

Nano AI Prompts

Find Your Next Copy-Paste AI Prompt

Explore free AI image generation and photo-editing prompts designed for practical results with ChatGPT, Gemini, Nano Banana and other image tools.

Explore Free AI Prompts →

Frequently Asked Questions

Can AI restore old photos?

Yes. AI image-editing tools can help repair common problems such as scratches, dust, fading, stains and some soft or damaged details. Results depend on the quality and condition of the original photograph.

Can AI restore an old photo without changing the face?

You can strongly instruct the model to preserve identity, facial structure, expression and apparent age. However, severely damaged or missing facial information may require AI reconstruction, so important photographs should always be checked against the original.

Can Gemini restore old photographs?

Gemini's image capabilities can be used for image editing through natural-language instructions. Detailed restoration instructions can help define what should change and what should remain untouched.

Can Nano Banana colorize black-and-white photos?

AI image models can be instructed to colorize black-and-white photographs. Because the original colors are usually unknown, the resulting colors should be treated as an AI-assisted interpretation unless you provide reliable color information.

Should I keep the original photograph?

Absolutely. Keep the original scan or photograph separately from the AI-restored version, especially when the image has family, historical or sentimental value.

Important: AI restoration can reconstruct damaged or missing visual information. For historically significant or irreplaceable photographs, keep the original and treat heavily reconstructed areas as AI-assisted restoration rather than guaranteed historical recovery.

Final Thoughts

The best old photo restoration AI prompt isn't the one that makes an old photograph look perfectly new. It's the one that repairs damage while respecting the photograph's original identity, composition and character.

Start with the best source image available, describe the actual damage, protect every person's identity and use conservative detail recovery. When the original information is uncertain, preservation should win over artificial perfection.