Has Ai Ended My Fashion Photography Career?

I am a fashion photographer by profession. Is AI going to take my job immediately? Should I quit now, hang up my camera, and forget it all? I decided to put multiple AI image generators to the test to see how quickly I'm going to be out of a job. After doing so, I feel safer in my current position, and here's why.

I found the best way to approach AI image generation is to give it a reference to go off of. So, I took a campaign that I recently shot and put the images into an AI image generator to see if I could create another campaign or a lifestyle shoot with different models in a different location that otherwise a client would have to pay for, including flying out all the models, paying for the location, and flying out the photographer.

I figured if it could do it through AI, then it saves thousands upon thousands of dollars. Or does it?

First, I took the original image, put it into AI, and gave it some notes on what to make of the image. I was looking for the same outfit on a different model walking down the beach in Mexico. Its initial result, although the photo is lower res, I think in time this is something that will be improved and have higher image resolution capabilities. But for something super high res, this doesn't quite cut it.

Beyond that, the image isn't actually that bad. The outfit is looking how it's supposed to. The model looks natural with a natural pose. The lighting is clean. You have that subtle but diffused backside light coming in, and the wind is blowing in the hair. There's some motion to the photo. She is centered, which I think would be better if she were in the right third of the frame, but that's something I could always prompt for to adjust.

But let's say we want a shot of two people interacting together. So we have two different outfits on two different models, and they're walking beside each other on the beach. I'm looking for a photo that is vertical, something that would work well for social media. When prompting AI for this in Nano Banana Pro, the result I'm getting is two friends walking beside each other, although it looks like they're holding hands.

The laughs that I'm seeing from them aren't bad. They do have the same tattoos as the original model, which I don't love. It's basically just like they copied and pasted her body and put different heads on it and adjusted the skin tone. But the tattoo is still there. So then I prompted the AI to remove the tattoo, have them walking but not holding hands, and have the shot be vertical.

I tried this about five different times, and I was essentially being met with the exact same shot. I tried different ways of explaining it as technically as I possibly could, as objectively of a shot, and was consistently met with not what I was asking for. In fact, I even got to the point where I asked the image generator what I could do to get a better result out of it.

It described how to do so, and then I re-prompted it using its suggestions and still was met with the same photo. So then I tried to give it a different photo that was vertical to go off, at which point it did make the image vertical but changed the background entirely to one of the previous original photos that I used as a reference and had a different outfit on them. Slowly, I also realized that the AI is making the clothes further and further from the original. As a fashion photographer, it is important that the clothes, even as I shoot them, look the exact same as they do in person. If you're selling clothes, you want your advertising to look like the clothes that you are selling naturally.

Slowly, this was getting further and further from that. But I decided to prompt it to reference the original and have the models change where they are. Let's put them on an Italian street and have a close-up vertical shot. I said to make it about a three-quarter body, where they're cut off at the mid-thigh, having a little bit of headroom just above the head. Upon prompting it, they are still walking beside each other, holding hands. The tattoo is still there. The outfit is now further and further from the original. As I re-prompt it, now one model has become a twin of the other while wearing the wrong outfit.

So, I said let's have them wear the original outfit and be back to the other model again, cropped in of a shot, which this is not.

Now, we're actually at least cropped for the frame, but there's a little too little headroom. But at least we're starting to get there. The only thing being, the black outfit still doesn't match the original photograph. So, let's just have it match that original photograph as I'm prompting it again.

We're now down to one model who looks like a slightly warped version of the original model, wearing the outfit in a different background that is not the background that I said. So, what all of this tells me, after rounds and rounds of prompting to try to get one particular shot and then changing it to try to get something at least close to what I was asking for, is that for a one-off image, AI does a pretty good job. The further you try to push it, the results will begin to dwindle from your expectations.

The biggest thing here is consistency. Yes, you can tell the AI that consistency is of the utmost importance, but over time it still will degrade. Models will change slightly in looks. The clothing will change slightly over time, which compounds through AI revisions. As you revise the prompts, I find they start to get, in some cases, further from what you're going for rather than closer toward it.

Even if it might listen to one part of your prompt, it might disregard the other parts. It can be consistent in some ways and inconsistent in others. It is very easy to confuse AI. Now, I am still on the newer side to AI image generation, especially for fashion images. Over time, not only will AI get better, but I will get better at prompting it. But what I'm seeing is, of all of the images that I generated, if I were to send them to a client, only maybe one would actually be usable. That usable one would be rather low res.

So you can only use it for basic social media, maybe a bit of website usage. It wouldn't be anything that you could really print because the resolution would not hold up to it.

So should I hang up the camera? My thoughts are not right away. To the point of even if I wanted to add AI to a service that I could offer clients, it would be something at this point that I could not guarantee the outcome of. It would take much longer than anticipated to get the results that I would be looking for, and even then, at least for me, I can still tell that it looks like artificial intelligence.

This echoes the same thing we saw with the Coca-Cola holiday commercial, where it took thousands of prompts and over a month to make a short commercial. What this tells us is that AI is more of a tool, just like anything else, and can be built into your already existing workflow. But the chances are it will not replace it.

With all that said, it could drastically improve over the coming years, as we have seen it do over the last five years. But I do still think even in the fashion industry, there is a place for the creative professionals who have been working in it for decades. For what it's worth, at the end of the day, the AI still needs someone to prompt it and know how to prompt it to get a certain image, so that job will be created.

If you want to sharpen the craft AI still struggles to replicate, The Fundamentals of Fashion Photography is worth a look.