Robots vs. Humans: Who Will Design the Best T-Shirt Graphics?

To go from the guide to the catalog, start from the personalized t-shirts page: from there you can choose men's, women's, or children's and configure the quote online.

Are robots already better than humans at designing custom t-shirts?

---

Low-cost custom t-shirts are in high demand among both companies and individuals looking to give birthday or graduation gifts or organize special events. Each customization is highly varied: from company logos to t-shirt lettering , from colorful to minimalist graphics. But who creates these graphics? Well, they're usually images taken from the web and are therefore created by humans. But lately, new opportunities are opening up… Like graphics created by robots.

Neural networks at work to create t-shirt graphics.

The T-shirts sold by Cross & Freckle, a New York fashion powerhouse, don't seem revolutionary at first glance. They're made of cotton, available in black or white, designed for a unisex fit, and sell for $25. Each shirt features a small embroidered design that references iconic New York City staples: pigeons (yes, that's right), pizza slices, and even the adorable subway rats. Sarah McBride, co-founder of the brand, compares the T-shirts to "New York City's summer uniform." But neither McBride nor anyone else actually designed the shirts. They were designed by a neural network, which analyzed data from millions of people… and returned the original works that make up the embroidery. Cross & Freckle isn't the first company to use AI to generate art; people have been doing it for years. But this project offers a glimpse into the nascent world of AI-generated fashion, where designers use machine-learning models to remix and reinterpret old designs. Cross & Freckle doesn't just use AI to create its designs. It also derived the brand's name and logo from a neural network called the Hipster Business Name Generator. It also used an AI text generator to create the text image on its website. In short: the brand is entirely powered by AI.

How did the graphics for the Cross & Freckle t-shirts come about?

The Cross & Freckle T-shirt graphics look like something a child might have scribbled with a pencil, or something an adult might have drawn with a computer mouse in Microsoft Paint. And indeed, these impressions are not far from reality: the graphics were created by a variational autoencoder trained on data from Google Creative Lab's Quick, Draw game . That game has collected over a million doodles from people around the world. Quick, Draw works a bit like Pictionary: you have to draw a drawing and the computer guesses what it is. In total, the drawings make up what Google calls "the world's largest doodle dataset." Paul Blankley, who manages the technical side of Cross & Freckle, built the autoencoder and fed it doodles of pigeons, rats, pizza, and dogs from Google data.

"There are hundreds of thousands of drawings in each of the different categories, from countries all over the world," Blankley said. "So you get this mix of what the world thinks a pigeon doodle means, or a pizza doodle means."

The autoencoder reworked those drawings, creating its own original graphics.

So could Artificial Intelligence design all the T-shirts on the market in the future?

It might sound scary, but it could be. According to McBride, the beauty of generative AI is that it churns out countless different options. It can choose what would be the best version of a puppy or a slice of pizza for a T-shirt, or which hipster name is both casual and realistic. The fashion industry already relies on AI to fine-tune various aspects of its work, from inventory management to trend forecasting to customer recommendations. "Virtual stylists" help people find their ideal measurements when shopping online, and recommendation engines help surface the pieces that best suit someone's personal style.

In conclusion

Numerous brands, both large and small, are currently trying to incorporate machine capabilities into their production and sales processes. For now, however, it remains to be seen: will a robot be able to convey the same sensations as human-made graphics? And in any case, the final decision is always human: robots create the designs, invent brand names, and even create new faces, but (for now) it's up to people to make the final decision, determining whether products are released to market. Graphics created by robots, yes, but the human hand and mind are certainly not being left aside!