Avatar embodiment research has typically tackled two problems separately: how completely the body, hands, and face are tracked and reproduced, and how personalized the avatar looks. According to the new arXiv paper, each line of work has relied on specialized lab equipment, making the approaches difficult to translate to everyday devices.
The paper, FlowAvatar, proposes a system for real-time full-body avatars that works from sparse egocentric inputs on consumer XR hardware. The title and abstract suggest the contribution is in combining tracking completeness with personalized appearance without the need for lab setups. The available excerpt does not include implementation details, so the specific techniques remain unclear.
If the claims hold, the work could lower the barrier for realistic self-avatars in virtual reality, where users typically see only floating hands or simplified bodies. The abstract frames this as an integration problem rather than a tracking or rendering problem alone.