To AI or not to AI?
Vishnu-Somnath temple on a morning pregnant with thick monsoon air in Khandepar village, Goa.
Both the images below derive from the same exposure. The second image shows what my eyes and the lens saw. As a result of ongoing construction a temporary fence had been placed around the temple perimeter. Using AI Generative fill, I cleaned it up and the result of that operation is seen in the first image.
Photographers routinely clone out distractions in images such as power lines, a strewn soda can, an intruding twig, and so on. So long as one is not fundamentally misrepresenting the extant scene, not disclosing these minor touch-ups is a long standing practice.
What about the current scene? Where do the limits lie? Nothing fundamental has been altered but the AI makes guesses in its fill. Should the photographer come clean in this instance and be upfront with the viewer?
Vishnu-Somnath Temple, AI modified
5DS, 24-70L f/2.8 II
Vishnu-Somnath Temple, as seen
5DS, 24-70L f/2.8 II
beautiful photo of a charming little temple !
You have eliminated the scars and done the temple a big favour.
Thank you, Sanjeev-bab.
Are the steps leading up into the temple in the first picture what they actually are? If so, were they manually etched or did AI guess them AI correctly? The fence is obscuring the view, hence wondering. To answer your wuestion, I think the photographer must disclose the use of AI if the primary object of interest (in this case the temple) is not a faithful reproduction of the original.
Dear Rajan, I love the first picture. the temple so beautifully extends into the woods behind and invites any person walking aloing that road. I can imagine entering the temple in the first picture
Dear Prashanth, Thank you for your thoughts.
I like the first picture as it removes the distraction of the fence (temporary any way).And I love the minimalist structure, in the arms of the luxuriant woodland, an irresistible invitation to the passerby for rest, reflection, prayer.
Thanks, Antonio-bab. Yes, I like the modified version, too. The question is – should this change be disclosed or it is sufficiently close to what was there that disclosure is not required.
In my view, the answer, Rajan-bab, is – the latter.
Very effective use of AI to remove the fencing. Combined with that soft light, the end result is print-worthy.
Bob – thank you.
Both are stellar amigo but #1 is my favorite! This is a BEAUTIFUL landmark against the lush, vibrant dark green of the monsoon season!
Amigo Dan – Yes, striking conditions, and a case of being at the right place at the right time.
The first picture is very striking. The second picture imho is nice too, but the first one brings the temple into such great focus.
Manish – thank you.
With the previous version of CAF, the effort to cleanup the distractions sometimes gets so time consuming and tedious and even the results aren’t good, that unless it is quick and easy, I just leave them, sadly, as part of the scene and carry on with the other edits.
Just so we are on the same page: this is not CAF, but Generative Fill AI. The two are entirely different tools.
Nice image! Even with fence, which I think , honestly is not a distraction
Thanks, Arun-bab.
The white, red and gold of the temple set in the background of lush green is lovely. Thank you, Rajan.
Oh to the question to “AI or not AI”, I say AI.
Thank you, Atanu. I use AI judiciously for touch-ups and not to conjure scenes that weren’t there during the exposure. But that’s because my aim is to produce what are conventionally thought of as photographs. Those who want to do digital art are free to use the tool in any they they wish.