Game design is the art of crafting the flow—where a challenge is just hard enough to be thrilling, but not so hard that it crushes you. For VR developers, finding this balance is notoriously elusive. The physical nature of VR gameplay makes testing exhausting, and yet, its experience is deeply personal. Every body is different, so the sweet spot is a moving target.
In this talk, we look beyond a bot that can simply beat the game, toward a model that simulates the diversity of human struggle. What does it take to build a player model that is not just performant, but authentic—and how might such a model reshape the design and experience of VR?
Enter Robo-Saber, our generative player model learned from the movement trajectories of over 70,000 real-world Beat Saber players. Unlike traditional gameplay agents optimized to win, Robo-Saber captures the messy, distinctive patterns of human motor control that define our gameplay diversity. We show that this "bot that plays like us" goes beyond automated playtesting: it can mimic the skill and idiosyncrasies of a specific person, or simulate an entire crowd. Leveraging Robo-Saber lets us quantify difficulty in a way that is objective yet personalized, helping identify content that lies precisely in your flow zone. Finally, we examine the feasibility of extending this generative approach toward a fully embodied, physics-based player model—one that grounds VR design in the biomechanical reality of the user.
Project page: robo-saber.github.io

Nam Hee Gordon Kim (namheegordonkim.github.io) is a PhD student at Aalto University in Finland, where he studies embodied AI and generative models of movement for interactive systems. His research asks how motor behavior and intelligence shape each other, and how breathing life into virtual agents can support human creativity and experience design. His collaborations have contributed to projects published in venues including SIGGRAPH, SCA, NeurIPS, and PMLR.
Before joining Aalto, Nam Hee earned an MSc in Computer Science from the University of British Columbia (UBC), advised by Michiel van de Panne, and explored his passion in teaching at UBC. He has also interned at Ubisoft and Amazon and is a recipient of NSERC Canada Graduate Scholarships and the Finnish Center for Artificial Intelligence's full doctoral grant. After finishing his PhD, he looks forward to continuing his curious work as an educator and a dad.