I can sense the tension. Many of you may be gritting your teeth and shouting, “Absolutely not!” There are certainly situations where using AI-generated images in photography is inappropriate. Initially, I’ll highlight several scenarios where it could be acceptable to incorporate AI-generated images into your photography, and then we’ll look at less favorable uses.
When AI-Generated Images Are Acceptable
I can think of three scenarios where using AI-generated images in photography is justified.
1. Creating Canvas Wraps
I had a client who wanted a canvas wrap of one of my photos in a specific size. Upon uploading my image to the printing service, I noticed it was cropping parts of my photo. I could darken the edges with black, which would leave a black margin around the photograph, but the client wanted to avoid that. So, what’s the solution?
To address this, I used the Generative Expand feature in Adobe Photoshop. This tool created edges that seamlessly blended with the photograph’s content. These extensions were only visible on the frame’s sides and not in the main picture. Eventually, the client opted for a frame, making them even less noticeable!
2. Removing Distractions from Photographs
Art can take many forms, and not every photo is historical or journalistic. If I have a distracting branch or streetlight in my picture, I’ll remove it to create art.
In the past, I’ve cropped out distractions or used tools like the clone or Content Aware Fill. While these approaches usually work, sometimes generative fill offers better results, with fewer leftover artifacts. After all, using AI to analyze the image and eliminate distractions feels no more unethical than using Content Aware or retouching a blemish in a portrait.
Ideally, I would have set the camera in a way to avoid capturing the distraction. But we all know that’s not always possible.
Most of us can agree that in photojournalistic or historical contexts, it’s best not to alter anything.
3. Correcting Perspective
The idea behind AI-generated borders is similar to the reasoning for canvas prints: we aim to enhance the image rather than deceive the viewer.
Sometimes, when I take photos of buildings, keystoning (the visual distortion causing parallel lines to converge or diverge) is unavoidable, and I can’t always back up far enough to capture a wide enough shot to correct it. Attempting to fix the distortion during editing may result in cropping out too much of the image.
A practical solution is once again to expand the edges using Generative Expand. While Content Aware Fill can be used, it typically doesn’t provide the same quality and is more time-consuming to perfect.
After expanding the edges, I can then correct the keystoning. It’s acceptable if I end up losing part of the edges, as they often vanish during the correction process. Often, it’s easier and more visually appealing than using Content Aware Fill or other techniques.
While the best way is to shoot wider, real-life challenges sometimes prevent that. Give yourself grace and utilize the technology available to you (again, for artistic purposes, not for photojournalism or historical photography).
When It’s Not Appropriate
If you disagree with the previous examples, you’ll likely feel even stronger about “deep fakes” and blatant attempts at deception. Here are a few examples that I believe should never be acceptable.
Deceit
Most would agree that using AI-generated images to deceive people undermines trust. Some viewers no longer trust my night photographs, which is troubling. More critically, misinformation has diminished trust on a broader scale. When people can’t trust what they see or read, it endangers our society.
Intellectual Theft
Many creators argue that AI learns from existing works, effectively mimicking or appropriating others’ creations without consent. Adobe tries to mitigate plagiarism concerns by training its models only on Adobe Stock images and public domain content. Some contend that we all draw inspiration from others to some extent, similar to how AI operates. Yet, many still feel uneasy about this, distinguishing between drawing inspiration and echoing others’ work without depth or intent.
Risk to Income and Job Security
The impact of AI-generated images on revenue streams is widely debated, especially regarding the microstock market.
Environmental Concerns
Another valid argument for resisting AI is its significant energy consumption. The International Energy Agency (IEA) estimates that global electricity demand from data centers will double to 945 terawatt-hours by 2030, surpassing Japan’s current total electricity consumption.
This has led me to reduce my reliance on tools like ChatGPT and AI-generated imagery. However, given that AI is becoming increasingly integrated into our lives, avoiding its use entirely is becoming difficult, whether it’s making a phone call to the dentist or browsing online.
Establishing Boundaries
A helpful guideline is that if you believe any reasonable viewer would feel misled by your explanation of how you created a photograph, then you may have crossed an ethical line.
Ultimately, it comes down to personal comfort. I personally feel at ease using AI-generated images occasionally in my work as outlined above. However, I understand that many individuals are not as comfortable with it. Given the growing concern over energy use, I intend to approach my use of AI with greater mindfulness.
