• 0
  • 0
Andrew Williams's picture

Flag #459

I suspect you have seen mention of the new generative fill capability within the Beta version of Photoshop that came out a week or so ago. I installed it over the weekend and have been experimenting with it to see if and how it might fit into my workflow.

Memorial Day motivated me to document more flags in the neighborhoods as part of my ongoing project, "Wrapped in the Flag." This particular photograph presented a few processing challenges due to my limited mobility and Harry trying to pull me into the street to sniff his friend, Ruby, when I was shooting it.

The original camera image was tilted and had several obstructions. When I straightened the image, I could either severely crop it or turn the rectangle into an octagon. I chose the latter. Previously, I would have used contextual fill to clean up all four corners and restore the image's rectangle-ness. The generative fill was slower (it has to send the file to Adobe's servers to do the work) but resulted in a much cleaner fill. There was a branch intruding into the upper left corner and a mailbox in the lower left. Previously, I would have cloned the adjoining area and hoped I could get everything in alignment. Once again, generative fill to the rescue.

One of my guiding principles is that any manipulation should be either so obvious that viewers know you did it on purpose or invisible so they don't know you did it at all.

Even looking at this under very high magnification, I cannot see any evidence of manipulation, even though I know I did it. Ethical questions aside (delayed, not ignored) this all worked far better than I expected.

FYI, this was shot with an 80-400mm Nikon lens @400mm on my D850.

Log in or register to post comments
5 Comments

Thanks for sharing Andrew. I’ve been meaning to try the new feature and appreciate your take on it.

I think we could go overboard with AI but like the idea of using to clean up problematic areas that might ruin an image otherwise.

It would be nice to see the original so we can appreciate the impact.

This is the original NEF file, vastly reduced in size and converted to a JPEG for posting. Otherwise, unedited.

And you should pay attention to this for the best results with large (almost all) files.

https://www.youtube.com/watch?v=pGJ0KjT4Xds&t=4s

Thanks for including this Andrew. I have been testing GF and was not aware of this issue.

Very nice cleaning job!