I believe it was a year ago that I found image restoration models became so easily available. As a kid, I used to watch National Geographic and the Smithsonian Channel where they aired colorised historical footages, which required countless hours of work by teams of colorists, archivers and restoration units to produce. Today, it is just a JPG drop and prompt away.
Considering what these tools are capable of nowadays, I know it was a small experiment, yet I found some interesting insights to share. I wanted to see if I could bring colour and clarity back to old photographs of my grandfather, Catello Spagnuolo Sr., and his brothers. They were all young, smiling, and full of life, but in a greyed out world. I uploaded one of their group photos into a visual restoration tool and wrote a simple prompt: Restore this image, fix tears & dust, enhance contrast, colorize gently.
I utilised two different AI tools: Gemini 1.5 Pro and DALL·E 3. Gemini 1.5 Pro impressed me with how naturally it handled portraits. In close-ups, it maintained the context of light and natural skin tones, revealing subtle details like the warmth in my grandfather’s face, making his eyes shine. But when it came to group photos, it struggled to identify the characters, resulting in a bust with a dash of blue.
DALL·E 3, on the other hand, left me speechless. It lit a thousand colours in the picture and made it feel alive. However, these colours for their suits, shirts, and ties were entirely imagined. Yet the model was able to maintain minute details like the striped fantasy of my grandfather tie, again impressive!
I felt captivated by the hallucinated colors AI decided to use to dress these men.
Seeing my grandfather’s in its youth, its expression in color made me feel closer to that time and the man. It remains a frozen moment, but a bit warmer.
AI didn’t just restore old photographs; it created a connection for me. If you find any pictures around in black and white, I urge you all to try it out.
