Should AI edited photos be labelled?

[pop]

Advertisement (gone after registration)

3 hours ago, Coppereye said:

Now this use of AI should be labelled

Why should it? You shouldn't call it a photograph, though.

[Coppereye]

6 hours ago, jaapv said:

But then , this is not a photograph but a technograph and no Leica  gear was used…

You are right no Leica gear was used - I was making an extreme point - creating an image out of text.

My friend calls his work as art and has been open about how the image was created

BTW Technograph is a good name.  

[jaapv]

Made it up just now. 😉

[Stuart Richardson]

I think often these discussions can drift into binaries, in a ever deepening spiral looking for exact criteria to delineate what is acceptable and what is not. This is not aided by the extremely broad and poorly understood technology behind the word AI. For example, are we talking about large language models and/or image content generating AI, or are we talking about algorithmic tools that have been trained using large amounts of data? The ethics are quite complicated, as some AI was created ethically, while others were created largely off of stolen data. And of course the use is important as well. I am sure no one here would accept the unauthorized use of AI to generate compromising images of real individuals. There is also the question of authorship. I don't think most would consider an AI the creator in an image where it was only used for masking or NR, but when parts of the image are generated authorship is called into question, prompt or not.

I think it gets to Justice Potter Stewart's definition of obscenity as "I know it when I see it". The line is going to be a bit fluid and is going to have to be on a case by case basis.  Personally, I am not particularly comfortable with the use of AI in my editing, but I think Andreas's criteria makes sense to me. I think there is a difference in things like AI noise reduction and masking than there is in generation and removal. I think it is partially because our culture and our brains have not caught up yet to the new reality. The notion of objectivity and "reality" in photographs has been dispensed with by academics for decades, but the general public still has an inherent credulity and trust in images. To pretend they don't is to ignore the evidence.

As a small example, I had a large collage work in a museum show, and the fact that there were manipulated images was fully disclosed in the museum text and artist statement. I thought I had made the collage extremely obvious. I did this collage manually in photoshop based on three different film negatives. It was clean, but not seamless. That was not the intent. Still, many people would come up to me and be like, "wow, where was that taken?". Once you point out that it was collage, people see it immediately. But the eye is trusting, especially when it is not primed for images to be deceptive. So I think in this context it is important to disclose.

As for the difference between traditional removal or generation, I think there is also a component in which skill and effort are important. For example, the people removing purged individuals from pictures with Stalin had to work pretty damn hard and carefully to achieve their result. That work was their own. The difference with AI is that, even when used as a "tool", it does not really obey the original author. It makes its own decisions based on trying to predict what you want, rather than to obey instructions. It uses its huge libraries of data to see what everyone else did...even if the result is the same (the object removal), there is a fundamental difference. I think even in the context of removing a lady from the background, it is important to note. It may not be significant to you in the moment, but in a way it is also just a kindness to the viewer to show your hand... These days I find myself seeing lots of amazing images on instagram reels in particular, and I look further and they are fake. It is honestly just tiring and depressing. As photography does become more and more influenced by AI, I think it is important to plant a firm stake in the ground to show what isn't. 

Edited September 23 by Stuart Richardson

[pgh]

I like the term "technograph" on first reading. 

@jaapv - In your image of the wine tasting, I would call the second image a technograph, and one where you only share authorship. It is not a photograph, and the issue isn't the image itself, but it is an issue if it is presented in a way where it traffics as a photograph and a viewer can reasonably be expected to read it as one. But it's not, so it could be deceptive. This is a case of mild and pointless deception, really. It's not even an image where I personally understand taking the time to manipulate it as such, but that's beside the point.

Paintings are not a good comparison because they look like paintings. The nature of them automatically provides meaningful information to a viewer. The deception with ... I'll say CGI or any sort of pixel imagery that looks like and mimics a photograph in all of the obvious visual ways but isn't actually one presents an inherent level of deception (at least for now). @Stuart Richardson narrative above is plain evidence of this.

This is all a matter of parsing words, really.

What is a photograph? This should mean something - and should not be used to describe how something merely looks, but how something was made. Painting is made with paint. Photograph is made with light. It's called a work of mixed media if the painting incorporates paper, printing etc to make the image. Why should a photograph that has sections (even insignificant) of totally made up computer generated pixels get a free pass and still get to use such a term?

That's why I like "technograph" or an attempt at some other word to be used for imagery that looks like a photograph but isn't. No value judgment on whether it was okay or not to make, it just signals that it is something else. A hybrid of notable manipulation or a computer based fabrication.

[LocalHero1953]

Aside from forum rules, or the rules of competitions and professional/academic qualifications, we are left with the personal standards that we choose to adhere to.
At risk of repetition, I summarise below my own approach, as written in a couple of previous posts, but only evolved over the last year or so as the power and ubiquity of AI has become clear.

1. I will label any image I post here where I have added or removed elements (by AI or traditional means). This means Adobe object removal, generative fill etc. I will not record the use of AI masks and AI denoise (etc).
2. If I see an image made by someone who I know does such addition or removal without stating it, then I will mistrust it as evidence of a person's photographic ability (to see an image, and to capture it with a camera) - though it may well demonstrate their skill with Photoshop (etc) and their eye as an artist.

I emphasise, this is my own approach - anyone else can do whatever they choose.
(And I don't rule out further evolution as others make persuasive arguments!)

[jaapv]

Advertisement (gone after registration)

A sympathetic and laudable point of view; but I think that we are well beyond the point of no return, and if not presently, it will not be long before it will be impossible to adhere to, simply because we will not be aware of what happened in the camera, nor in the processing, because it will not be labelled as such. Not even the photographer will  know. An out-of-camera JPG will be cleaned up before it is exported  and who knows what happens before the raw is written.
Then it only remains to judge whether the result matches the vision of the photographer.

Things like authentication and raw comparison will remain in the domain of competition and PJ work, etc. and will not even reflect in-camera manipulation. I do agree with addition and removal techniques to be declared in select cases, but in my book only if relevant. We don't declare cropping out, changing point of view or blurring the background either. 
Anyway, boring the viewer with technical details is not very productive in general in my opinion, perhaps it is best to state: "My photographs are not a copy of reality, they represent my vision on reality." 

[pgh]

41 minutes ago, jaapv said:

in my book only if relevant

Any time an image is described as a “photograph” or posted in a context where people reasonably expect to see photographs I would say it is relevant. Cropping, pov, dof are all different in kind and inherent subjective devices used at the moment of capture. They are all embedded in the photograph itself. That is quite different and in a way again, that I think is widely understood by the mainstream public. 

If cameras all start making decisions for me like eliminating objects, removing skin blemishes, smoothing out rough edges etc I will…I guess keep my m10s and my m6. If they all go beyond that it’s doing something I have no interest in - manipulating a world based on some designers arbitrary and likely very boring and commercialized sense of what constitutes a pretty picture (pretty ≠ good).

I also think there will be a market for cameras that don’t do this. Film has had a resurgence for a reason.
 

“Because everyone else is fine with it” or “the momentum is going this way anyways” is not a compelling reason to do much of anything in my estimation. I’ve said it before, it is exceedingly cynical. To a person it gives up agency and responsibility. 
 

No one I see is claiming that a photograph is a copy of reality - but a photograph, or photography is something unique as a method of producing a picture and it is rooted in a recording of light that was real, and a digital or analogue image that is not that and doesn’t maintain a pretty thorough fidelity to that isn’t a photograph. You can make a lot of funny looking photographs - a lot of things that don’t appear as a copy of reality - with the way you record light.

[LocalHero1953]

6 hours ago, jaapv said:

it will not be long before it will be impossible to adhere to, simply because we will not be aware of what happened in the camera, nor in the processing, because it will not be labelled as such. Not even the photographer will  know. An out-of-camera JPG will be cleaned up before it is exported  and who knows what happens before the raw is written.

Cameras removing elements and replacing them without the person with the camera knowing it? It might be possible, but, seriously, would you choose to use a camera that way? 

Not me, and I suspect there will be plenty of others that do not so choose. If I want pretty pictures, I'd just buy them, or ask an AI source to make one. And I wouldn't post it here as an example of my photographic ability.

Edited September 24 by LocalHero1953

[jaapv]

If you don’t claim that a photograph is a copy of reality, how can you argue that alterations to that photograph somehow destroys its validity? Coming back to the original question, despite going around in philosophical circles, I think that Andreas gave us practical guidelines and I for one completely agree. As for my personal photography, rest assured that, despite the one example here that I made for sake of the argument, my images will be written with light and refined with honest tools that don’t add outside elements, like they always have been. But that does not mean that my eyes are closed to the march of time. Once, in the nineteenth century, it was unthinkable that a gentleman could step outside the accepted code of honour , now we have a different code of behaviour. Similar shifts are in all walks of life, our concept of photography is equally fluid and shaped by the march of technology and general perception over time. The only thing we can do is to set our own personal boundaries hoping that they resonate. 
Such  is the privilege of the photographer. 

[LocalHero1953]

8 minutes ago, jaapv said:

If you don’t claim that a photograph is a copy of reality, how can you argue that alterations to that photograph somehow destroys its validity? 

If you're responding to me, then you missed that I was setting out what I choose to do and how I choose to view the images of others. What others do is up to them, and depends on their intention when using a camera and their claims (usually implied) when posting here. If I know their photos are likely to have significant undeclared adding and removal, then I will mistrust them as photographs. They might as well be posted here.

Among amateurs, there is already a distinction (though on a spectrum) between those who use cameras because they are interested in capturing a view of the real world as it is (like many/most people in this forum) and those who are constructing an image of the world as they'd like it to be. They tend to use smartphones, take selfies, use creative filters, and post on Tiktok. There's plenty of overlap, and, in the professional world, plenty of other areas (advertising photography, weddings etc) where reality+unreality has a place.

 

[jaapv]

48 minutes ago, LocalHero1953 said:

Cameras removing elements and replacing them without the person with the camera knowing it? It might be possible, but, seriously, would you choose to use a camera that way? 

You forget that we live in an era of self-driving cars that make life and death decisions for us, relinquish control to presets, algorithms and such in our photography making it only possible to approve or disapprove of the result.
We cannot dig in and simply reject the way the world develops, only accepting the way things were. We have to live with it and try and understand and by that understanding attempt to shape and guide. Ignoring never works. 
I do not have to like it, but I cannot close my eyes. 
 

[LocalHero1953]

Just now, jaapv said:

You forget that we live in an era of self-driving cars that make life and death decisions for us, relinquish control to presets, algorithms and such in our photography making it only possible to approve or disapprove of the result.
We cannot dig in and simply reject the way the world develops, only accepting the way things were. We have to live with it and try and understand and by that understanding attempt to shape and guide. Ignoring never works. 
I do not have to like it, but I cannot close my eyes. 

Then keep them open. And if you get pleasure from such tools, that's your choice.

As @pgh wrote, we will have a choice in the tools we use. The law may say your car must make decisions for you; in a free market, I doubt camera makers (and not Leica) will take such choices away from us.

We will also keep a choice over what images we like. I like paintings as much as I like photographs. But with a painting I am only tangentially interested in how it relates to what the painter saw. With a photograph, I am vitally interested in how it relates to the scene before the photographer. I want to know if a photograph shows things that were not there, or doesn't show things that were there. If I suspect the photographer is not being candid, then I will lose interest.

[jaapv]

I was not responding to you specifically but searching for the deviding line. I do not believe in statements like “ that is not photography “ unless we are looking at something that has absolutely no relationship like completely AI generated images. I try to dissociate the input of reality from the intention of the result. The result, to me, is not necessarily linked to the input of reality, but an expression of the vision of the photographer. If that vision is to reproduce reality as faithfully as possible that is perfectly fine and it will obviously limit the method used and necessitate the explanation of areas where the reality was modified. If, however, the intention was to create something new from the original input, how can we object to the methods used? 

[jaapv]

Controversy has the effect of making me research deeper. I still resist elaborate declarations - which depend on honesty anyway- I have found the simple solution:

I'll change my signature: 

"Any image that I publish -if edited- will have Adobe Content Credentials attached, from October 1st 2025."

This will automate the process, plus attach immutable copyright. 

[jaapv]

And just to add - Adobe generative AI only uses photographically (light generated) components, keeping it within the realm of photography. 

[pgh]

23 minutes ago, jaapv said:

And just to add - Adobe generative AI only uses photographically (light generated) components, keeping it within the realm of photography. 

Can you be more specific here? If you’re referring to the fact that generative AI uses previous photographs and/or the picture being edited to predict what pixels would replace it, I understand. It’s in the realm of photography only tangentially thought - it’s just using what it knows from previous photographs. It isn’t actually recording light to do this. It doesn’t maintain the fidelity of the original recording of light.

As for all the rest - the talk of progress, get on board etc etc. - This is actually irrelevant.

This isn’t actually about whether or not we will use AI/predictive algorithms in life etc, it’s about what we call it for what we’re doing here. And it’s about how we read pictures. It’s not a value judgment about whether or not something is in the broader scope acceptable. While not a perfect delineation maybe, we have expectations as readers when we pick up a work of fiction versus a work of non-fiction. This doesn’t mean that the non-fiction is objective, or even accurate to reality. But it is about the author’s intent and about the subject matter. So this is not about denying “progress” or whatever. It’s about what a photograph is, and what it isn’t - for me - at least. Progress can march on, photography can be come a dying medium, even. But that won’t change what photography is. 

On 9/24/2025 at 4:09 AM, jaapv said:

If, however, the intention was to create something new from the original input, how can we object to the methods used? 

I wouldn’t object to the image. I would only object if it were presented in a way that implicitly or explicitly positions it as a photograph, when it isn’t. Actually, if you look at my website you will find a good bit of work that I think shows that I have engaged in methods to these ends quite a bit. I was even, I would say, pretty aggressive in using Adobe’s early versions of some of the more complex algorithmic tools. I just hope the work is contextualized in a way where this is quite clear. 

And you’re right in that it’s not really a novel discussion, it goes back a long way - though it is more pertinent now that corrupting the photograph is so easy and seamless. Anyways, for me, while imperfect, the best way forward for now is to look at a photograph the way a press photographer’s association would. They use the term photo illustration for things that have been collaged or edited beyond global adjustments of contrast or exposure, etc. Technograph, photo illustration, whatever you call it. I’ve often used the term “Digital Collage” - but I’ve never really loved it. Just haven’t been able to think of something better for when I’ve done this sort of thing. So yea, these are fine visuals to produce. But not photographs. Not for me. You can debate where to draw the line and I’ll always remain open to reconsidering it based on photographer’s intent and presentation. But it’s where I start, and pretty much always have, except for my explicitly collage based work.

And at this juncture I am only interested in photographs when it comes to pictures that look like photographs. (I can get into some digital collage if it’s not trying to look like something else). So if the board is full of images that might have elements cloned out for the sake of aesthetic cleansing with no note there I won’t be interested. Because again most of the time it doesn’t even improve the picture - it quite often reinforces some arbitrary version of a prettier world that people have learned but possibly haven’t questioned too deeply. Too far down the slope of just rendering whatever you want to see as made by pixels. Which is just not interesting anymore (it did feel more novel a few years ago). This is the crux of the matter here for me as a viewer. 

[jaapv]

2 hours ago, pgh said:

Can you be more specific here? If you’re referring to the fact that generative AI uses previous photographs and/or the picture being edited to predict what pixels would replace it, I understand. It’s in the realm of photography only tangentially thought - it’s just using what it knows from previous photographs. It isn’t actually recording light to do this. It doesn’t maintain the fidelity of the original recording of light.

Which generative AI are you referring to? Adobe uses real  photographs that they obtained from their Stock. (with permission of the photographers who provided them.) No, it is indeed not recording light, it is selecting from light-generated photographs. So do you in any catalog. You are entering the realm of splitting hairs here. 

Plus, as mentioned, provides authentication. 

[pgh]

7 minutes ago, jaapv said:

So do you in any catalog. You are entering the realm of splitting hairs here. 

 

Splitting hairs for some is how you find out where you draw lines for others.

Whether Adobe is using other photographs to generate predicted pixels as a replacement in what you're editing, or they are just cutting and pasting from another image doesn't really matter - they're both a meaningful corruption of the photograph in question and therefore make it into something else. 

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[jaapv]

But, as I argued before, it is something else than reality anyway, so it is just a matter of degree.