I guess this would ammount to local AI training, but if I have a serries of photos of a person, or group of people, thing or things, it would be great if I could feed a “good” photo of a subject into the AI and have the AI use that information to fix photos of the same subject or subjects, rather than rely on generic training models.
So if I have a bunch of photos from a wedding, and get a photo of the bride and groom that would have been great but just turned out blurry or got accidentally photo-bombed, I could drag and drop a photo or photos of the bride and groom (that turned out well) into one panel as “example” or something like that, and tell the program to fix the bad photo.
Then, instead of trying to fix the photo generically, the AI has an idea of what the end result should look like with these particular people.
Now that I think of it, this could work for filling in details of subjects when upsizing by using higher resolution photos or simply photos where the subject is a lot closer, to feed the AI information about the same subject in a photo where they are much farther away. Then when you upscale, the AI isn’t trying to fill in missing pixels generically.
