ChatGPT + Photoshop

[Manfred M]

I thought I would test out some of ChatGPT's image functionality, so I took an old photocomposite from a few years ago and had ChatGPT colourize it and turn the image into a coverpage of GQ magazine. I didn't 100% like what came out, so I spent a bit of time polishing the image in Photoshop.

[escape]

Very nice, Manfred. I like it.

Recently, I was editing the photo of the guy on the bike in Lightroom. I then put the image into AI and asked what changes I could make to improve it in Lightroom. It provided me with detailed steps to take, and I asked it to apply those steps to the image. We went back and forth a few times until the image was where I wanted it, which I posted in that thread. It did change the color of the subject’s T-shirt, but that was okay with me.

What stood out to me is how beneficial this kind of workflow can be for everyday photographers—especially those of us who don’t have deep post-processing skills. As AI tools improve, they seem to lower the barrier between having a good image and being able to bring it closer to what you envisioned. That may broaden participation in photography overall, shifting more attention toward subject, timing, and storytelling, while making the technical side less intimidating.

[DanK]

I think this reveals a very big difference among photographers. For me, having control over the output is a big part of the enjoyment. It's a craft, like playing a musical instrument. I like using specific AI tools to accomplish things I specifically want, like removing a power line or selecting a sky. However, I have no interest in having AI create images for me.

My brother, who is also a photographer, said something similar when I was learning how to print. Once one knows how to do it, printing is usually not difficult, but it's a lot of work to learn to do it well. He said most of the photographers he knows fall into one of two groups: they enjoy printing as part of their craft, or they find it a frustrating nuisance. I wasn't sure where I would end up, but I ended up in the first group (except when my printer malfunctions).

I'm not saying one is better than the other. IMHO, in their photography, amateurs like me should do whatever gives them enjoyment. All I'm saying is that this split has always been around. It shows up in whether people use automatic (program) mode, whether they shoot JPEG or raw, whether they devote time to learning how to process, etc. However, AI has taken it to a whole new level.

[AntonioCorreia]

I find this quest for what Daniel is seeking quite fascinating, especially when combined with the use of Artificial Intelligence to achieve results that are, at the very least, surprisingly pleasant.
Although I haven't yet had the opportunity to exhaustively explore the full potential of these tools, that hasn't stopped me from making some successful attempts, as shown in the example I’m sharing here.

The image in question resulted from a curious process: I used an AI application based on a candid photo taken with the smartphone, one of those spontaneous moments where no one was aware the shot was being captured.
I believe these tools represent a valuable resource, not only for the everyday photographer but especially for those wishing to embark on new paths of exploration and creativity.

In fact, I’ve previously had the chance to show you works systematically developed through AI that possess such intensity and realism they seem to have emerged, quite genuinely, from the camera digital or analog.

[DanK]

I don't want to pick a fight, but I'll ask anyway because this question has been nagging at me for a quite a while when I read about people using AI for creative purposes.

If you asked someone how to accomplish something in postprocessing software, and they gave you instructions that you followed exactly, would you consider that your work? If not, how is that different from asking an AI program, which is basically giving you recommendations based on scraping thousands of other people's work?

In the case of actually creating an image, exactly the same question arises. If another person created it for you, would you consider it yours? If Nano Banana creates it based on many other images it has been trained on, how is that different?

I once posted about something a friend of mine suggested before AI image generation. He said that if all a traveler wants is a good picture of the place they are standing, there is no reason for a camera. All you need is a device that records where you are and accesses a library of other people's photos, letting you pick the one you want. He was envisioning a simplistic analog of AI image generation. I reminded him of this recently, and his answer was "my wife actually did that, but in a much lower-tech way. She bought postcards."

the main difference, it seems to me, is that my friend's imaginary device only let you select. AI lets you tell it what you want, and it uses what it has learned from the many thousands of images it has been trained on to try to do what you are asking. Seems similar to asking an artist, "paint a person at a counter leaning forward, with one hand holding a cup of coffee." AI tools are fun because they let you do that almost instantly, as many times as you want, at zero cost until you like what it has created for you. the user is not creating the image; they are just finding ways to give the artist or the software good enough instructions to create what they want.

This analogy is reflected in the rules of large camera club I recently joined, when I moved to a city too far from my old one. Here is their rule about AI, which they put next to using other people's images:

In all sections of competition, images must originate as photographs made by the entrant. They may not incorporate identifiable images produced by anyone else (for example: clip art, replacement skies, or stock images). Images created in whole or in part by image creation software (frequently called 'AI' or 'generative' images) are not allowed.

This is all very different from using AI as a teaching tool to learn a technique that you can then apply yourself. Increasingly often, I use Gemini to give me custom lessons about all sorts of things: setting Photoshop settings, configuring a new computer, and on and on.

[LenR]

I am inclined to agree with Dan on this score.......

[AntonioCorreia]

For me, creating merged/fused and manipulated image, before and/or after AI is, at heart, the same creative joy: looking attentively, choosing with care, and transforming with intention.
Photography has always held a kind of magic: in layering, cutting, darkroom work, and post-production. AI doesn’t erase that; it simply brings new possibilities. That’s why I truly believe the image itself has value with AI or without it, because what supports it is the human eye, sensitivity, and the decisions we make along the way.

I like to think of this as the blending of cultures and the phenomenon of acculturation: when different worlds meet, they don’t cancel each other out, they enrich each other. Brazilian music is a beautiful example of this living mixture, where African, European, and Indigenous roots embrace and create something utterly unique.

And when I look at Frank Gehry’s architecture, I see that same affectionate courage to combine forms and materials, with a modernity full of soul. One may like it or not.
The same goes for writing prompts from pre-AI images: it’s a continuation of photography, a way of building bridges, not replacing origins.
And even within groups of people with shared interests, it’s natural for the group and for each individual to choose different paths towards their own creative place.

[DanK]

Antonio,

I think you are mixing different things together.

All cultures are syncretic, combining things from other earlier cultures, and that's usually a good thing. On a personal note, I think that one of the best aspects of the US, despite the current revolting wave of xenophobia, is the ability to share cultures. My own family is a hodge-podge of cultures. However, I must have been unclear, because that has nothing to do with my point.

My point is about "agency", to use the currently popular American term. That is, who actually creates? My view is that if you ask another agent to create something, you aren't entirely the agent anymore. It seems to me that it makes no difference whether the agent is a real human or an AI program emulating humans. It is still someone or something else.

The principle in my mind is that if you ask the agent to do something specific that you want as part of your vision--removing a power line, for example--then you are still the agent. AI in that case is simply a more powerful tool for doing what you decide to do. But when you ask the agent to create something, you aren't the agent anymore, or at least not fully the agent. The dividing line is sometimes unclear, but that's the general principle.

When you ask an AI agent to create something, what it is in fact doing is using a great many of other people's images to do so. You probably know that it is the nature of machine learning that even the creators of the model can't tell you how it is combining bits and pieces of other people's images, but that's what it's doing nonetheless.

Frank Gehry did not ask another agent to design his buildings. His genius was that HE came up with creative ideas that no one else had. If someone had asked an AI agent to design the Walt Disney concert hall before he lived, it would not have come up with what he did because that wasn't in any of the data it scraped.

Off topic: Some people would use the verb "stole" rather than "scraped." You may not know that one AI company (Anthropic) has already lost a $1.5 billion lawsuit in the US filed by an author who argued--successfully and in my view correctly--that their AI model had illegally used her copyrighted material without permission. Unfortunately, the AI model stole from so many authors that the payment to individual authors is small. I'm actually one of them: it used two books I wrote without permission. My publishers and I will split the small award later in 2026.

[Round Tuit]

Dan,

I'm with you 100% because agency is paramount for me. I create photos for the satisfaction that I get from the process of creating. I could not be proud of a photo if I had use generative AI to create any part of it because it would no longer be mine. Once finished, most of my photos live on my hard drive never to be seen again by human eyes unless they get picked for the annual calendar that I make to give away to friends and family.

However, I realize that agency means different thing to different people. For example, people often say that they renovated their kitchen/bathroom when in fact they hired someone else to design and build the renovation. I have no objection if anyone wants to consider an AI generated picture as being "art" because they did the prompting of the AI program but I would not consider it a photograph. After all, Barnett Newman hired summer students to paint his famous "Voice of Fire" using paint rollers.

To each his own but generative AI will not show up in my photos.

[Manfred M]

These are all interesting comments and they remind me of what was said at the dawn of digital photography. Things like digital images aren't real photography, modern photographers are not as good as the old masters because they rely on tools like Photoshop to make images look good, etc.

There are still people out there that believe that anything other than a straight-out-of-camera JPEG are not acceptable. Software creates pixels that never existed in the original image.

I believe that we are in a similar position with AI and we all need to draw our own personal boundaries. There are many "levels" to using AI in photography and people are more sensitive to some functionality than others.

1. Some AI functionality either speeds up and / or improves the quality of the edits that we do. As an example, sky replacement has been around for years, but it was slow and painstaking. AI speeds this up.

2. We can ask AI for advice in editing our images. The software can actually perform the edits or can give detailed instructiions on what to do. Is asking for step by step instructions any different than getting advice from someone here at Cambridge in Colour? Is it a problem to use AI to "prototype" ideas to see how they might turn out and then make our own changes to our images to achieve these results?

3. Generative AI - this is the one area where people seem to have the most problems with AI as new content is created by the AI that was never a photograph. These elements are simply created by the software and have nothing to do with using a camera and lens.

I like trying this stuff out to understand the current state of the technology and have used it from time to time in my own work to do things that would be too time consuming or achieving something that I don't have the skills to accomplish.

I have used AI in a very limited way is some of my own work. My biggest issue with the software is lack of control by the user to influence the final result in a meaningful way much of the time. I expect this to be fixed at some point as the systems become more sophisticated.

As for the GQ image here, I could have created it in its entirty is I had wanted to. Creating magazine covers was part of the course requirements in a couple of different photography courses I took over the past 15 years or so. The output (from ChatGPT) was edited quite substantially to get the image to look the way I wanted it to after the original AI rendering.

[DanK]

This has become (for me) a very thought-provoking thread.

Re Manfred's post: I don't think this is analogous to the change from film to digital per se. However, I do think it was foreshadowed in weaker form by the power of digital postprocessing, which allowed photographers to easily insert things that weren't present, etc. But even that was done by the photographer, not someone else.

Re Manfred's #1: I don't think this is a problem, and I use it freely. For example, if I decide that a power line is a distraction and remove it, the tool I use to remove it isn't relevant (IMHO) to whether the resulting image is really mine.

I think Manfred's #2 is the interesting one. Let's think about a teacher rather than an AI agent. Suppose that you are sitting with a novice who has a drab, low-contrast image that they want to make pop more. I have been in exactly that spot. Let's compare two hypotheticals. In the first, you show the person how a bunch of the tools work and then tell them to try using whichever they like, perhaps commenting along the way (e.g., like "no, that's probably too much, because if you look here, you have created areas where the highlights are clipped). That's a gray area, which in the case of written text would properly be noted as "by X, with the assistance of Y." But I think it would be reasonable for the newbie to consider it theirs.

Now take a second hypothetical, where you first show them how the tools work and then say, "OK, I'm going to tell you how to use them on this image." They follow your directions. In my view, it is not their image, any more than painting by number is really the art work of the painter.

You can replicate either of these with AI agents.

Re #3: All I would add is that we have lots of hybrids also, where an image was taken by the photographer but AI was used to create parts of the final image, like a person or a vehicle. I don't consider those to be the work of the photographer, and I entirely support my photo club's ban on them. (They actually punish violaters: if you submit an image generated in whole or in part by AI, you are banned from the next three competitions.)

Re Andre's point about agency: I think this goes beyond the question of whose work it is. For example, I suspect that no one here would say that if someone creates an image entirely on their own but sends it to a lab for printing, the print isn't the work of the photographer. I certainly consider it the work of the photographer nonetheless. After all, a fair number of the best B&W photographers used darkroom collaborators.

Nonetheless, Andre's point about agency is one of the reasons I print. Another reason is for the control: I can select the paper I want and, without bid delays, tweak and reprint if I want. However, even if a print from a lab is pretty much what I want, I don't get the satisfaction from that that I get from having it come off my own printer. I get a lot of satisfaction from seeing a good print result from my own work. Psychologists call this effectance motivation.

Virtually no one uses snail mail, but there are still some occasions (apart from holidays) that call for cards. I print some of my own. In small print on the back, it says "photo and print by..." rather than just "photo by..."

[escape]

With the help of AI, I placed my work (and I use the term loosely) on exhibit. Recently, I was at an Art Museum, and I photographed my wife looking at a blank wall. My plan was to transfer my art piece onto the wall as if it were on exhibit. Just for fun.