RoleplAR Research

This summer I worked in Stanford's Human-Centered AI Lab on a project called RoleplAR. The idea is pretty cool: it uses AI to write little stories and then turns them into 3D scenes you can look at in AR, so learning a new language feels less like flashcards and more like actually being somewhere. My part was building the pipeline that connects the AI side to the AR side. I used OpenAI and DALL-E to generate the content, then converted it into 3D models (.glb files) that could actually render in AR. Getting a flat AI-generated image to turn into something 3D that looks decent and doesn't lag was harder than I expected. At the end of the summer I got to present what I worked on at the Stanford Summer Fellows Program, which was honestly a little nerve-wracking but a great experience.