Given its potential impact on the creative industry, it’s no surprise that there is a lot of talk around the subject of Artificial Intelligence. What has taken many people by surprise, myself included, is the speed with which AI has hit the mainstream; only three or four years ago, even the term ‘AI’ didn’t mean much to me beyond a sci-fi cliché, and the occasional passive reference in the specifications of photo-editing software.

And yes, so-called ‘intelligent’ features have been present in applications like Adobe Photoshop for a while, but at nowhere near the level of sophistication we’re seeing today. Compared to apps like Firefly and ChatGPT, they’re chalk and cheese.

You may like

Since the dawn of generative AI, most of the conversation has been focused on the potential negative consequences for photographers – if it will take our jobs or render cameras redundant – but there has been relatively limited discussion about how the use of AI can enhance how we take images.

It’s important not to forget that all AI models are trained on content created by humans. In other words, using all of the web as its inspiration, AI is actually geared to offer an overview of how humanity interacts with the world and how this makes us feel: the Human Condition.

As a framework for testing how ChatGPT responds to my prompts, I’m looking for each answer to hit one of three pillars: honesty, breadth of analysis and understanding. Honesty is fairly self-descriptive – how much does it tell me what I might not want to hear?

You may like

shutter speed, metering mode, and f-stop (later, it did identify that the camera – a Sony A7 V – must have a good dynamic range), but instead it honed straight in on the aesthetics.

This is definitely a human-like reaction. When was the last time you showed someone an image and the first thing out of their mouth was “I love your choice of ISO”? It’s far more likely they’d reference the sunlight, the shadows, how autumnal the image looked, or what it made them feel. I then got my first bit of gentle criticism.

“One small consideration,” ChatGPT begins, like a friend who’s afraid to hurt my feelings. “the brightest area near the top center pulls the eye quite strongly. That’s not necessarily a flaw, but slightly reducing highlight intensity there (either by framing or post-processing) could help keep attention more firmly on the pedestrians”.

Hmm, okay, I take the point. I wanted to introduce a hazy wash of light, so it was intentional, but at least the AI seems to have guessed this was the case. Other suggestions included cropping in a little tighter using a telephoto lens, a trend we’ll discuss again soon, and lifting the shadows on the left of the frame. Overall, this response managed to hit two of our three pillars, Honesty and Breadth.

ChatGPT feedback: This is a strong, confident street photograph with a cinematic feel. The light is used deliberately, the composition feels thoughtful, and the technical execution supports the story

My impressions: ChatGPT is capable of seeing an image from the viewer’s perspective, but given a generic prompt, it’s likely to keep feedback generic.

Images of a city street at sunset

The same image with the amendments suggested by ChatGPT, including a closer crop and toned-down highlights (Image credit: Peter Fenech)

HDR cityscape” territory.”

I was kind of hoping it would notice this, as it was the image parameter I would probably have highlighted first if I were performing a portfolio review. The saturation on this shot had been turned up in Affinity Photo for a specific paper stock; in print it would look fine, but the JPEG does look overcooked. The rest of the feedback on my choice of camera settings was complimentary, but the final word surprised me.

“The composition is very centered and very balanced. While this works, it also makes the image feel a bit safe.”

I don’t actually agree with it in this case, since the placement of the clock tower was largely dictated by editorial requirements, but of course, ChatGPT didn’t know that. What’s interesting is that without too much of a detailed prompt, it was encouraging me to take risks with my framing.

This is excellent as many beginners don’t really know what to ask or that central framing might be considered a bad thing. The AI was able to provide some professional insight, the type I’d hope to receive on a one-to-one workshop with a pro. This fulfilled the second of our three pillars, Breadth of analysis.

ChatGPT feedback: This is a polished, technically accomplished cityscape, and it would sit comfortably in a curated urban portfolio. Bringing a bit more of your personal voice or risk-taking into images like this would be the next exciting step.

My impressions: Describing your intent is necessary for targeted, practical feedback, but AI can recognize common beginner mistakes and competently offer advice for improvement.

Images of a city street at sunset

The cityscape image, complete with the ChatGPT suggestions applied. This includes a closer crop – something the AI seems to prefer – with the clock tower more closely on a third, and reduced saturation (Image credit: Peter Fenech)

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!

And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.