A little photo restoration project.

It was actually a year-long or more process. I started working on it before AI came along and ended up with something adequate enough for the landlords to give it to our local library for their archives. When AI was first introduced, I was able to make some additional improvements while leaving the faces along. Then AI improved again and I went in for the finishing touches. I did take some liberties. I wasn’t sure where the cloth with the floral pattern was coming from, whether it was part of the woman’s dress or a blanket, so I made it her entire dress. And I did overlook part of the scarf around her neck. Contrast, vignetting, the sepia tones and some final touches were old-fashion manual fixes in Photoshop.

Long story. I actually started it a few years ago using Photoshop and whatever rudimentary photo restoration apps were available at the time. The results were adequate enough that the landlords gave a copy to the local library for their archives. Then I kept improving and AI kept improving and I’d continue working on it every year or so.

Then Gemini 3 happened. When I work with Gemini 3, I first upscale the image dimensions and ppi to 300 to give the AI enough pixels to work with. Then I take individual square sections of areas that need work (faces, hands, feet, articles of clothing, various objects), and work on each section separately after pasting and resizing the result back onto the original. (By the way, I learned one of the most important prompts one can use when editing with Gemini, for example, “Improve the quality and anatomy of the hands, with particular attention to the fingernails. Leave everything else as is.”) Gemini likes to take liberties on occasion, so that last sentence is important. I even developed a keystroke shortcut for it.

Anyway, to answer your question, years. But all in all, maybe 8 hours total.