Is AI Bad For Landscape Photography?
One of the most polarizing topics in landscape photography these days is the use of artificial intelligence (A.I.) in post-processing software. On one side of this technology, Adobe Photoshop and especially Skylum's Luminar AI have been heavily promoting the ability to do automated sky replacements, adding in fog effects and sky effects such as god beams. While these techniques have been employed by real estate photographers for years, landscape and nature photography has historically been rooted more in capturing natural history than manufacturing scenes that didn't actually exist.
I personally have no interest at all in creating nor marketing work such as this. While I did consider purchasing Luminar AI to illustrate this blog post (maybe I'll update this eventually), the whole reason why I chose to be a photographer is to show people the interesting places and things that I saw with my own eyes. If I start to sprinkle in automated composites then it makes all of my work become suspect if every third image features elements that weren't actually there in front of my camera. The common argument is that if you disclose then you can do whatever you want. Sure, that's technically true but as someone that also enjoys viewing photography, I feel that once you "break the fourth wall" then the magic is lost. The emotional connection I might have had from seeing something that represents a real scene goes out the window once I realize that half of it was never actually there.
Less cheesy and a glimpse into where A.I. can be really beneficial for my landscape photography are the products created by Topaz Labs. Their tools are meant to help correct technical issues and optimize images for print. In the Gigapixel AI example here, I have an 8 megapixel file (approx. 12 inches at 300 DPI) that I've upsized to 60 inches at 300 DPI. While I can make good prints at that size with many images through the use of other upsizing algorithms and sharpening techniques, Gigapixel AI does an incredible job here at creating entirely new pixels based on my original file. I would still comb through this file with a fine-toothed comb if I were to make a print however. More likely, I would use the original file as my base file and go through my usual workflow then blend in parts of the Gigapixel file where I need the extra detail. AI is not 100% perfect at this point but it can do some really amazing things already.
Just about all images benefit from sharpening techniques before publishing. Traditional Photoshop tools include unsharp mask, smart sharpen and high-pass filter. Topaz Labs has their own tool called Sharpen AI. Based on my limited testing of this tool, it tends to be on the aggressive side of sharpening so it's most beneficial on images with minor focusing issues, lower resolution images and to reverse the blurring effect of anti-aliasing filters which many camera sensors employ. This artificial intelligence-based sharpening algorithm is definitely more sophisticated than unsharp mask which I used for edge sharpening during the first decade of my photography career. Again, I don't know if I'm ready to trust this technology 100% at this point due to occasional artifacts that get magnified through enlargement but it clearly has a place in the toolbox.
What really excites me is being able to mitigate the effects of lens corner sharpness. The center of the lens is the sharpest part of most lenses while the corners tend to be less so depending on the aperture. In this example, I have a pretty sharp and detailed image overall that has some corner sharpness issues. I ran this file through Topaz Labs DeNoise AI and it did a near-perfect rendition of my photo while revealing the corner detail that was lost in my original photo. I didn't have a need for this on most of the frame so I would selectively blend in as needed.
With judicious use, I feel that the use of AI in digital photography is here to stay. This technology can help us photographers improve our offering from a technical standpoint. What were considered acceptable flaws in the past might not even exist anymore in the near future due to advancements in technology. Certain tools can also help with efficiency by eliminating redundant or manual tasks. I heard from one photographer on Twitter felt that we're not too far away from the day when AI can learn from our own personal post-processing style and replicate it automatically. Cutting out post-processing time would certainly save me a lot of time though from a creative standpoint I'm not sure if being 100% hands-off would be as satisfying for me as an artist. I predict that I'll continue to use AI-based tools in the future but also still remain in full control of my own photography.