Thursday, November 7, 2024

Arcade, a new AI product creation platform, designed this necklace


Mariam Naficy, the founder of cosmetics retailer Eve and design marketplace Minted, has a long history of supporting independent artists and helping them sell their products. Her latest venture, Arcade AI, is a new type of marketplace that puts creative control directly into the hands of users.

The generative AI platform, which launched its beta in September, allows customers– referred to as “Dreamers”– to input their ideas into a generator that then produces a variety of jewelry design options. They can also upload images.

Once a “Dream” is selected, it’s assigned to one of Arcade’s artisans, who brings the digital concepts to life as tangible pieces of jewelry, including bracelets, charms, earrings, necklaces, rings, and more. There are a number of materials to choose from, including gold, brass, silver, and a variety of gemstones, including diamonds, garnet, ruby, and more.

I tested the new generator myself, which leverages a mix of third-party models, such as Stable Diffusion and Midjourney. After entering my personal preferences, particularly my love for opals, the AI tool generated a delicate, gold floral necklace for me (featured in the image above).

I then placed the order with Studio Maisette, one of the design studios on the platform, and a few days later, I received a video of the finished product for approval. The entire process, from order placement to product delivery, took around two weeks.

While the necklace might seem basic at first glance, I think it perfectly captures my aesthetic. Bonus points for the trio of opals, which are of good quality, according to my gemologist friend.

Not all the results generated were up to standard, however, and I ran into some hurdles along the way. For example, the AI generator had a hard time adding multiple charms to one necklace. It also failed to create complex objects like a twenty-sided die.

Image Credits:Arcade AI/Lauren Forristal

“To make sure that everything is manufacturable, we actually have to constrain the models a little bit to what the makers are giving us in terms of data or what they can make,” Naficy told me.

Fortunately, there’s an editing tool where you can add, erase, or adjust certain elements to truly bring your vision to life. Still, it didn’t always turn out as I envisioned.

Naficy plans to improve the “magic editor,” adding things like Image Prompt (IP) Adapters, which are small models that allow pre-trained text-to-image diffusion models to create images based on image prompts. Another idea in the works is to let users be able to choose a part of the image to modify. Or drag a part of the image and move it elsewhere.



Source link

Related Articles

Latest Articles