We use AI image generators to create many of our lifestyle product mockups, including our clothing models, as well as for many of our designs. Some of you are gasping and shaming us already. You may be thinking, "How could you steal jobs from plus-sized models and photographers?" or "I want reality not fantasy; how do I know when something is an accurate depiction of what I'll get if I buy it?" or "You're copying the work of other artists," or just, "Why?" Hopefully you'll still stick around and hear us out, because as usual, the issue isn't cut and dried. We here at Twisted Trail Life rarely do anything without a lot of agonizing research and deliberation (we have ADHD and autism, afterall). Trust me, we're hyper aware of the arguments, and we've taken a lot of points and counterpoints into account in our decisions. And it's personal for us too. We have close family members who are directly in the path of AI tools potentially disintegrating their jobs in not very many years--or at the very least threatening to change their jobs in ways that make them almost unrecognizable from their current state. It can be scary.
AI tools aren't new at all, but many people seem to have discovered them in the last year with the grand entrance of ChatGPT onto the scene. The integration of plain-language prompt communication and image-recognition components to these large-dataset computation programs, as well as the open source nature of the source code, made these tools accessible to a greater population, not just the code nerds, which fosters collaboration, creativity, and rapid adoption in the space. And boy, "rapid" is an understatement! All sorts of AI tools are maturing at exponential rates, and there's a big debate going on about the ethics involved in using AI in all sorts of scenarios. But while that debate rages on, mainstream businesses and governments and everyday people have raced forward and are integrating AI everywhere. It's definitely NOT going away. The discussions are important, and the tools are definitely learning and growing and being refined. We all will watch this technology grow up, and like a small child raised by many villages, the outcome could be either wholely positive, blandly neutral, or horrifyingly awful. No one knows. We think it's encouraging that great minds are contributing early on to give us the best chance of a good outcome, and hopefully blind capitalism will be only one contributor to the growth path amongst several more-positive influences.
This generation of AI image generators were trained on a problematic group of public images with a lot of built-in bias and hardly any diversity, but the beauty is that they're constantly learning from newer data sets and as more users feed them prompts and new images depicting new ideas. And yes, the image results we get are amalgamations of lots of photographers and artists' work over a long period of time, but a lot of people would agree that all art is: any creative work is influenced by other creative works. Of course we all should be striving to create new original works rather than actively copying others', but we can get lost in the whole inspiration versus imitation argument. Also, not everything that is the same is copied. With similar influences, it is demonstrably possible to have multiple people or artists come up with the same idea in the same time period (e.g., Google synchronicity or multiple discovery). It's happened to me.
But I digress. Let's talk about how and why Twisted Trail Life uses AI.
We always knew we wanted to display our products on several different body types, genders, ages, and ethnicities. Initially our research showed very few examples of inclusivity in small shops. That makes sense. For small companies, it would be prohibitively expensive to show all your products on several different models especially as your product lines grow. Photography sessions are not cheap. It's a lot of work involving lots of team members and models, takes a long time to get results, and if you release new products, you have to do it all over again. Also, you would need to produce all the products for the models to wear, which is additional expense. And you can't sell that product afterwards, so it's wasteful too. We certainly weren't in a position to head down that road! So what were our other options?
In the very early days, we tried using a print-on-demand company (POD) to produce our products (read about that disastrous journey here). One potentially helpful thing the major POD companies do is provide you "lifestyle" photos for the products you select. A "lifestyle" photo is an image of your product being used by someone or displayed in a scene with an environment that mimics real life. You select the product substrate you want to sell, upload your design image, and they superimpose the latter on their pre-selected models' images. It's called a mock up, because there's never an actual product in that process. Lifestyle images are very effective! We all want to see ideas of what a product looks like in the wild doing what it does. It's an absolute must in ecommerce if you want to sell anything at all. Another option for creating lifestyle mockups is to use one of the many popular product placement subscription services, like PlaceIt. The trouble with these services is that, unlike the POD services, they don't have images of the actual products you're selling, so you have to select the model that most closely resembles the product you're selling. A lot of people simply select a model they like, paying no attention to how accurate the product is. And with both these services, sellers end up using the same images of the same models over and over again. What little diversity they offer seems formulaic and staged.
These providers are making a killing on their services. Do the models make money from those applications? Dunno. They probably got paid for the initial shoot, but I doubt they're making residuals. Are the models real or AI? Dunno. It's getting really hard to tell anymore. My guess is that if they aren't already, the images will be AI generated very soon. And where does that leave models? Well, one of the things we discovered in looking into product modeling is that it's not as lucrative for the models as you might think. I had a friend who tried modeling for a few years, and she often ended up on sets where the bulk of her "payment" was that she got to keep the products she was modeling. WTF? Not only does that not pay rent, but she said the sets and accommodations were frequently uncomfortable, with what felt like leering folks all around. It took a tremendous amount of time for very little gain. And she was tall, slender built, white, and classically pretty. I imagine it's a lot harder for nontraditional models.
So we took a different approach. We turned to AI to help us create our own mockups. If you think that's the easy way, you haven't tried it. Trust me. I've been in the digital art space for a long time. I know my way around a computer. And it's still a struggle. Every single AI result I select needs to be heavily edited, and it's a skill just to create the prompts to get results worth selecting from! The models frequently have too many or too few fingers, multiple thumbs on the same hand, with f'd up nails. Ears are always the wrong size, too big or too small. The lighting is almost always wrong, and you can get odd or just odd-sized props showing up, like a weird fruit tree in a kitchen setting or a giant rock in a home. I've gotten results with heads on backwards and elbows facing the wrong way, eyes looking in different directions or with lifeless "dead stare," hands shoved in nonexistent pockets, a lot of pointed fang-like teeth or just a line of the same-sized teeth with no incisors or bicuspids and not split down the middle. It's creepy, y'all! I have to fix all that before I can ever start the tedious process of knocking out the clothing to create different colored shirts with different designs. It's time consuming, requires a lot of skillsets, and I pay subscription fees for the engines I use, so it's not free. But it's cheap comparatively and gives me control. I can create the images I want to display, keeping an eye on accuracy and realism, while upholding our mission of inclusivity.
A recent discussion with my daughter--who, of all the people I know, stands to have her job most affected by AI--further enlightened me. She pointed out that AI image generation has opened up creativity for a whole new group of people who might otherwise never have been able to express themselves and their ideas. Someone who doesn't have the skill to produce the image in their brain using traditional artist's tools can use an AI image generator to reveal that idea so they can share it with others, who in turn might be inspired to create other ideas, and on it goes. Isn't that a beautiful thought? Anyway, I have more thoughts, but we're all tired now.