ADIDAS OZWORLD
AI Generated Avatars for the Metaverse


The world’s first personality based AI Avatar creation platform.


 
To celebrate the arrival of Adidas Originals Ozworld’s collection Jam3 partnered with Adidas and Ready Player Me to launch the world’s first personality based Avatar creation platform.

In the midst of the Metaverse hype of 2022 many virtual worlds would take advantage of Ready Player Me’s avatars.  These human-like avatars had a diverse amount of customization options but they lacked more fantastical designs at the time. For Ozworld we saw an opportunity to experiment with style and self expression. Influenced by the upcoming Ozworld collection we created future forward digital garments that reflected peoples IRL personas. 

I was brought on near the end of development to first help optimize the WebGL experience, secondly to develop a command line tool to prepare the 500k+ avatars created by users online for Ready Player Me.

The platform

The web platform starts with a guided personality quiz which prompts the users to express their individuality through 7 different interactions. 



Once complete the AI model generates a unique avatar for the user along with a digital code they can redeem on the Ready Player Me platform. The user also has the choice to download gifs and stickers to use on their messaging services.






Ready Player Me

Every single avatar created on the web platform had to be prepared and exported as GLTFs following Ready Player Me’s avatar specification. Working closely with the Ready Player Me engineering team we developed a workflow to deliver the avatars in batches on a weekly basis. 

We tried initially to use the web platform to export avatars but soon ran into memory leaks and crashes when generating 1000s of avatars in a single session.

One advantage of using Threejs is how adaptable the library is in different execution environments. We were able to use node-canvas-webgl to emulate the WebGL rendering context in a nodejs environment. With the final command line script it took less than 10 seconds to generate and save each one to the file system. 

Closing Thoughts

I’m forever grateful to be part of the team that worked on this project. It’s a true reflection of innovation and originality in the realm of virtual fashion and web3 🖤.



Agency

Jam3
Role

WebGL Developer