Just days after Google launched Nano Banana Pro, its most advanced AI model for image creation and editing, social media is seeing a wave of AI-enhanced, photorealistic images that are increasingly blurring the distinction between real and synthetic media.
Nano Banana Pro runs on Google’s recently released Gemini 3 Pro, part of the company’s latest suite of foundational AI models with multimodal reasoning abilities. Users have quickly noticed improvements over the previous Nano Banana model, which used Gemini 2.5, particularly its ability to produce text in images more clearly and in multiple languages.
Popular Use Cases
While various creative applications of Nano Banana Pro have gone viral recently, one of the most common uses is restoring old photographs. To get optimal results, prompts should follow a structured approach:
[Type of original photo], [Action required], [Specific defect removal], [Desired final quality/style]
All-in-One General Restoration
This approach is best for standard vintage photos that are faded, blurry, or slightly scratched:
Direct Method: “Restore this vintage photograph. Remove scratches, dust, and creases. Correct fading and yellowing. Sharpen details, reduce noise, enhance texture. High resolution, clean image.”
Descriptive Method: “Complete restoration of damaged photo. Remove paper texture artifacts, blur, and dirt. Enhance details in clothing and background. Balance exposure and contrast. Pristine, sharp, 4k-quality image.”
Repairing Heavily Damaged Photos
For photos with tears, water damage, or deep creases, a stronger repair prompt works best:
Heavy Repair Prompt: “Intensively repair severely damaged photograph. Reconstruct torn edges seamlessly. Remove deep fold lines. Eliminate water stains and discoloration. Restore damaged areas while preserving the original subject. Clean, reconstructed image.”
Scratch and Mold Fix: “Remove heavy scratches, mold spots, and emulsion damage. Smooth paper texture without losing details. Denoise extensively. Clear, restored image.”
Facial Enhancement for Portraits
Vintage portraits can be tricky as AI sometimes over-smooths faces. Use a prompt like:
Realistic Face Enhancement: “Restore old family portrait. Focus on facial details, eyes, and skin texture naturally. Remove motion blur. Avoid over-smoothing skin. Keep the authentic vintage look of the background while making subjects sharp and highly detailed.”
Colorisation: Black & White to Color
For converting black-and-white photos, avoid sepia-only washes. Effective prompts include:
Historical Accuracy: “Colorize this black-and-white photo. Apply historically accurate colors to clothing, foliage, and architecture. Maintain natural, warm skin tones. Vibrant yet realistic color palette. High-definition image.”
Fixing Faded Colors: “Correct color fading and shifts. Neutralize red/magenta tints. Restore natural white balance and vibrant, accurate colors. Remove grain. Sharpened, modern color grading.”
Using Negative Prompts
Another useful technique is applying negative prompts, specifying what should not appear in the final image.
Bonus Tips for Nano Banana Pro
Input Quality Matters: Ensure your scanned or photographed image is high-resolution, flat, and glare-free before uploading.
Iterate: Start with a general restoration prompt, save the result, and then run it through a follow-up prompt focusing on faces or color details.
Avoid Over-Prompting: Too many contradictory instructions—like “make it 100 years old but also look brand new”—may confuse the AI.
Add Negative Prompts: Specify unwanted elements such as scratches, dust, tears, creases, folds, water damage, mold, sepia, yellowing, blur, noise, grain, low resolution, distorted faces, oversaturation, cartoonish effects, artifacts, bad hands, or missing fingers to guide the AI more precisely.
Also Read:
Gemini 3 vs Gemini 3 Pro vs Gemini 3 DeepThink: A quick overview of Google’s newest AI models
Google aims for 1000x compute growth, plans to double capacity every six months: Report
