by Lark M. Robart
Google Query: What is ‘Nano Banana’?
Google Response: Nano Banana is Google’s internal codename for its advanced AI image generation models, primarily referring to two versions: the fast Nano Banana (Gemini 2.5 Flash Image) for quick tasks and the powerful Nano Banana Pro (Gemini 3 Pro Image) for high-fidelity creation, both known for accurately placing text in images, consistent character generation, and complex visual reasoning, allowing users to create everything from product mockups to infographics within the Gemini ecosystem and other Google products.
——————————————————–
I’m currently in the process of writing a book about my mother’s paternal ancestors and I have lots of wonderful photos that I’d like to include in that book. As is the case with many of us, most of those photos are faded, torn or damaged in some way. Or they are snapshots of original photos, likely taken with a camera or copied on a flat bed scanner at Kinko’s in the 1980s and 90s. Definitely not ready for prime time.
I’ve played around a bit with AI photo restoration tools in the last few months, with mixed success. A few weeks ago, however, I heard that Google’s tool – known as Nano Banana – had made a big leap forward in terms of photo restoration. So I decided to give it a try.
I knew the key to success would come from writing a good prompt, so I decided to ask Nano Banana for guidance. Here’s what that looked like:
Me: I’m working on a book that will include many images of my ancestors. Some photos have been damaged and torn, others are a bit blurry, and nearly all are faded and could use a bit of help with restoring sharpness. However, I do not want any of the photos colorized. My goal is to have an image that will look good in a book format. Can you provide me with a set of prompts to accomplish this goal with each picture?
Nano Banana: Restoring old family photos for a book is a wonderful project. To get high-quality results without adding color, we need to be very specific about restoration versus reimagination. 🎨
I will help you build a library of prompts tailored to different types of damage. Since you want to maintain the original look, we will focus on keywords that emphasize texture and tonal range.
🛠️ The “No Color” Foundation
To ensure your photos stay in their original palette, every prompt should start with or include these tonal anchors:
“Professional grayscale restoration”
“High-contrast black and white photography”
“Original sepia tones preserved”
“Monochrome fine art print”
📋 Prompts for Specific Issues
| Issue | Target Prompt Strategy |
| Tears & Scratches 🩹 | “Repair physical damage, mend paper tears and scratches, seamless texture restoration, preserve original facial features.” |
| Blurry/Out of Focus 🔍 | “High-definition sharpening, enhance facial details, refine iris and hair texture, 8k resolution, crisp edges.” |
| Faded/Low Contrast 🌓 | “Restore deep blacks and bright whites, balance histogram, reclaim lost shadow detail, improve tonal range.” |
| Dust & Grain 🧹 | “Clean surface dust, reduce digital noise while preserving film grain, smooth paper texture, clear photographic plate.” |
💡 Tips for Book Formatting
When preparing images for print, “DPI” (dots per inch) and “clarity” are more important than they are for a screen. To help the AI understand the book format requirement, we can add phrases like:
“High-resolution scan quality”
“Publishing-ready detail”
“Clean background isolation”
To give you the most effective prompts, it helps to know which path you are taking. Are you using a specific AI tool like Photoshop’s Generative Fill, Midjourney, or a specialized photo restorer like Remini or MyHeritage?
Would you like to start by building a “Master Prompt” for a specific photo that has multiple issues, or would you prefer to dive deeper into how to use “Negative Prompts” to strictly forbid color?
Me: Attached is a photo that has multiple issues. Please provide a “Master Prompt”.
Nano Banana: Here is a “Master Prompt” designed to restore the image while strictly maintaining its original vintage character and sepia tone.
Master Prompt:
“Professional restoration of the vintage sepia photograph of the young boy, image_0.png. Preserve the original sepia tone and monochrome palette; do not colorize. Repair the large vertical tear running down the center of the photo, seamlessly reconstructing the boy’s dark jacket, white ruffled shirt, plaid bow tie, lace collar, and the painted landscape background. Remove all other scratches, creases, dust, and discoloration spots from the photograph’s surface and the cardboard mount. Enhance the overall clarity, sharpness, and contrast, bringing out details in the boy’s face, hair, and clothing texture, while retaining the original photographic grain. Ensure the text ‘MAUD DAVIS BAKER, HELENA, MONT.’ at the bottom right is crisp and clear. The final image should be clean, sharp, and ready for high-resolution book printing.”
Using this prompt, I added the BEFORE picture you see below. Less than one minute later, I received the AFTER version. (Click on the images to see larger versions.)
NOTE: The cream-colored 4-pointed star icon you see in the lower right corner is AI’s way of letting the viewer know that this is a Gemini-Generated image. The fact that the icon can be easily removed is a topic of conversation in the AI/genealogy world right now. No one seems quite sure how to deal with that fact but the concern is certainly there.
For the most part, I was very pleased with the restoration, except for one thing: the face in the BEFORE image seems a bit brighter, especially the eyes. I asked Nano Banana to try and rectify that issue but the result made the eyes look a bit googly.
(Google, Googly – get it? Hah!)
Moving on.
I spent some time yesterday running the “Master Prompt” through its paces, modifying it to fit each image. Once I got a system going, it all went pretty fast. Below are a few more examples.
Later in the afternoon as I moved into early to mid-20th-century black and white photos, I was getting mixed results. So this morning I had another chat with Nano Banana in an effort to better understand how to restore photos from that time period.
A few tips:
- If at first you don’t succeed, try, try again. At one point I got an AFTER photo that was nothing like the BEFORE photo. After one more failed attempt within the same CHAT, I deleted the entire CHAT and started over.
- I asked for tweeks on a few occasions. For instance, some of the images looked too dark so I asked Nano Banana to “lighten the image just a bit”. Sometimes it worked, sometimes it didn’t. That’s something I can easily manage later using Canva or Google Photos.
- Be sure to keep the AI naming convention for the “restored” files you download to your computer. And I added additional information:
- the date the “restored” file was created; and
- the name of the BEFORE photo so that I can always track back to it if I need to.
- Nano Banana is currently referred to as “Image Creator” in the platform itself; and I used the THINKING mode, which may or may not be included in the FREE version.
All in all, this photo restoration tool has been a success, delivering print-ready images for my book that I’m very pleased with. And I anticipate this tool will continue to improve and become even more user-friendly over time. Although it may not be a universal fix for all damaged and faded ancestral photos, it certainly offers the solution I need.
If you’d like to give Nano Banana a try, click here. I believe they are still offering basic photo restoration for free, with daily limits. I switched over to the PRO version, for which I pay $20/month.
