Here at DIYP, we’ve had a lot of fun playing with ChatGPT and asking it all sorts of stuff. Udi asked it to make a PRD for a perfect mirrorless camera, with some interesting results. I recently started testing it for crochet patterns and I have yet to make some stuff from AI-generated instructions. But worry not, the articles are still 100% human-generated. :)
Photographer Gavin Hoey wanted to know if ChatGPT could write a video script in his own style and give him instructions for a photoshoot. In this fun video from Adorama TV, Gavin shows you what the chatbot came up with.
Setting up the scene
Gavin wasn’t impressed with the style ChatGPT used to write the script. But hey, it came up with the setup, which is what I find particularly interesting about it. “We’re going to be working in a studio today, and we’ve set up a dark colored seamless background for our subject,” the scrip reads. “To add some drama to the images, we’ll be using smoke machines, colored gels, and studio strobes.”
Setting up the lights
ChatGPT proposes to use a spotlight as the main light source, positioned low and to one side of the subject to create strong shadows and highlights. So far, I’m not loving it. And what exactly is “low?” And what kind of light should Gavin use? The instructions were vague, so he asked for more details, and the chatbot suggested placing the light lower than the model’s head. Yikes.
Since it’s not specified where on the side to put the main light, Gavin gave himself some artistic freedom to get the best out of this setup:
For the fill light, the chatbot suggested using a softbox on the opposite side of the key light. Gavin’s model, Sophie, made a “mistake” of looking into the fill light – but it actually looks better than turning towards the key light AI suggested.
“For an extra layer of drama, we’ll be using color gels to change the color of our lights,” the AI-generated script reads further. It chose deep red and blue to “complement each other” and it made me grind my teeth. Gavin asked the chatbot for an alternative, and it suggested orange and blue – a combination which gave much better results.
Modeling and adding smoke
As far as modeling tips go, AI was, once again, pretty vague with explanations. Gavin asked for more details, and the bot suggested asking the model “to stand in a strong and confident stance with their shoulders back and their chin held high.” How this gets along with the idea of “dark and moody shots” the bot suggested later, I have no idea.
It also suggested using props “like a hat or an umbrella.” They tried some shots with the umbrella which made everyone chuckle: Gavin, Sophie, and me on this side of the monitor. The setup was finished off with some smoke, and Gavin shares the photo shoot and the result near the end of the video. Thanks to Gavin’s knowledge of light and photography, the results are more than decent. But if he only followed AI’s instructions, I doubt the result would be as good.
The bottom line
On the one hand, I’m quite impressed with the amount of information you can get from a chatbot, regardless of the topic. Also, it often proposed some great advice. For example, experiment with light and colors and adjust your lights throughout the shoot until you get the look you want.
However, I think AI is not there yet. Its tips are too vague, and it takes extra effort to get specific advice you can apply. It’s also pretty inconsistent: describing a “dark and moody” shoot and suggesting poses that just don’t get along with the concept. Also, it doesn’t really do a good job of suggesting the lighting setup. Types, colors and the position of lights were all pretty bad at the first attempt, and Gavin still had to make a lot of extra effort to make it all work.
I don’t think AI can replace humans in photography, at least not yet. And regardless of all the “hiccups” it made in Gavin’s script, I found this really fun to watch!