How did you go about it? There have been some horror stories about when people didn’t know how to instruct the AI and then they came up with some freaky pictures. Did you have a similar experience?
Our process was to try to understand [the bot]—how you communicate with it and how to activate the prompts. At one point, we came up with images of people with six fingers and others with octopuses everywhere, which looked like they belonged on the cover of a sci-fi book. We did some really wild experimentations just to understand how it reacts to the text prompts.

We wanted to express fashion through an AI lens, but we also wanted people to wonder if it’s real or fake. Our concept was to use Midjourney to create daily life scenarios with AI-generated models and settings, with the only thing real being the clothes we used for the shoot.

Tell us about this duality and blurring the boundaries between the mundane and surreal, with the only real element is the fashion. Why was this decision made?
In a magazine [like Tatler], we want to feature outfits from designer brands in every issue. For example, we want to feature a Chanel jacket or a Gucci dress, but Midjourney can’t generate any branded item—be it a Coca-Cola can or a Chanel jacket—because there can be huge issues around image rights and copyrights.

We were thinking of how AI can be used as a tool to carry out a shoot if budget is not a concern. Can we just combine the [generated] photos with real models and clothing [in post production]? One of the biggest challenges we encountered was how to best combine AI and reality, [because] we wanted to keep the fashion element true, we didn’t want to distort real fashion.

The legal aspect is a huge concern. When we’re doing shoots, we want to make sure that everything is above board. Tell us about the actual logistics of putting the shoot together.
We spent quite a few sessions just trying to generate the exact backgrounds we wanted, by using different text prompts and arranging them differently. For example, we would use a text prompt like, “a New York rooftop on a rainy day”. You can also adjust the lighting, the camera model and so on in AI to match what our photographer would be using, so that we can recreate the same setting. That was the first step—to narrow it down to the exact images we are recreating.

On the actual production day, we had two models wearing the real clothes in similar silhouettes and sitting in exactly the poses [generated on Midjourney], so the photographer can capture them in the exact same lighting that AI had generated for us. After that we needed to send it back to a retoucher we have been working with overseas to make sure this is doable: to merge photography and AI-generated photos, and cross-check that we can actually combine the physical and the virtual, especially when the whole storytelling is more promoted by the bot than us. So it’s quite an interesting learning process in which the whole team kept revising.

Read more: Is AI art identity theft? What creators want you to know about digitally generated artwork

Looking forward, can you see us further harnessing AI?
Despite the ongoing legal debate, generative AI is still a very strong tool for assisting creative executions. It has proven to be very useful in building visual aids, mock-ups for creative decks or for creating illustrations and generating new layouts for existing photography. And while we are waiting for upgrades in apps such as Midjourney, we really encourage everyone to spend some time with the generative AI of your choice and explore how it could be of help in any creative project you might have.

Watch the making of the fashion shoot below:

NOW READ

7 times Margot Robbie’s outfits paid tribute to the Mattel doll during her press tour

7 entrepreneurs using AI to transform the world

Opinion: How AI-Generated Art May Change the Way Designers and Architects Design