I put ChatGPT and Google Gemini to the test for restoring old photos

A new viral trend is sweeping across social media, as users use advanced AI tools such as Google Gemini and ChatGPT to restore their old, grainy photographs with remarkable precision. The phenomenon has taken over platforms like X, where users are sharing impressive before-and-after images that look almost professionally retouched.

The growing popularity of this trend lies in how easily it allows people to breathe new life into ageing photos. Pictures once faded, pixelated, or blurred can now be transformed into crisp, high-quality portraits, while still retaining their original look and emotion.

Using artificial intelligence, these tools enhance colour, texture, lighting, and detail, turning old digital or scanned prints into versions that resemble modern-day DSLR captures. Many users say the results appear “as if the photo was taken yesterday.”

How to restore your old images using Gemini or ChatGPT

Users can simply upload their old images to ChatGPT and Google Gemini, and afterwards, the prompt will restore their images.

An AI expert and X user @ShreyaYadav_ shared a post on Saturday explaining how to recreate professional-quality results.

“Restore Your Old Photo with Gemini or ChatGPT

Old Photo
AI Gemini
AI ChatGPT

Prompt:
Edit this photo into a professional portrait of very high quality and colour, comparable to the results of Canon EOS R5. Create like a recent photo, with great clarity and no noise. Produce razor-sharp photos. Full HD 32k.”

This detailed prompt is now being widely circulated, helping users achieve studio-grade image restoration through AI models.

Also Read | I tried viral AI car photo hack — ChatGPT x Gemini Nano Banana nailed DSLR looks

Bonus prompt

I created my own prompt using ChatGPT, and the results with this one were clearly much better. The final output was impressive and worked very well in the Gemini.

An photo restored using the Gemini AI.

Here is the prompt that I used:

“Restore this old photo of my dad with advanced AI upscaling and enhancement. Please ensure the restored version looks natural, realistic, and recent — as if it was taken today with a professional camera like the Canon EOS R5.

Preserve Identity and Originality: Keep my dad’s exact facial structure, expressions, skin tone, and hairline intact — do not alter his face or make him unrecognizable.

Clarity & Resolution: Sharpen the image to ultra-high quality (Full HD or higher), remove any blur, noise, or grain, and make the photo crisp and detailed.

Color Correction: Restore accurate, balanced colors — natural skin tone, realistic lighting, and lifelike shadows. Avoid over-saturation or cartoonish tones.

Remove Damage: Eliminate all scratches, stains, cracks, fading, or folds. Clean up the background gently while keeping the original scene authentic.

Texture & Detail: Enhance fabric textures (shirt, jacket, etc.) and facial details (eyes, hair, wrinkles) subtly to make it look like a recent portrait.

Also Read | Kim Kardashian calls ChatGPT her frenemy after AI makes her fail tests

Lighting Enhancement: Add soft, natural lighting — like daylight or studio portrait lighting — that flatters the face while maintaining realism.

Background Refinement: Smooth and clarify the background without changing its type (e.g., keep a vintage studio or outdoor background if present).

Output Quality: Generate a final image that appears professionally shot — high-definition, noise-free, balanced contrast, realistic depth of field, and fine dynamic range.

Deliver a restored, high-resolution portrait of my dad that feels modern, clear, and authentic — preserving the emotional essence and originality of the old photograph.”

ChatGPT vs Gemini: Which performs better?

While both ChatGPT and Gemini deliver impressive outcomes, early comparisons suggest that Gemini Nano Banana offers more natural colour grading and texture retention. ChatGPT, on the other hand, excels in maintaining fine facial details and sharpness, giving images a refined, high-resolution finish.

It is noteworthy that both platforms rely on deep learning models trained on vast image datasets, allowing them to interpret and enhance photos intelligently. The choice between them often comes down to personal preference and the kind of aesthetic one hopes to achieve.