SARAH: Spatially Aware Real-time Agentic Humans

TL;DR

SARAH introduces a real-time, causal method for creating spatially-aware conversational agents in VR, enabling natural gestures, gaze, and orientation aligned with user movement and speech, suitable for live deployment.

Contribution

The paper presents the first fully causal, streaming VR-compatible approach for spatially-aware conversational motion using a novel transformer-based architecture.

Findings

Achieves state-of-the-art motion quality at over 300 FPS

Outperforms non-causal baselines in natural spatial dynamics

Validated on live VR system for real-time deployment

Abstract

As embodied agents become central to VR, telepresence, and digital human applications, their motion must go beyond speech-aligned gestures: agents should turn toward users, respond to their movement, and maintain natural gaze. Current methods lack this spatial awareness. We close this gap with the first real-time, fully causal method for spatially-aware conversational motion, deployable on a streaming VR headset. Given a user's position and dyadic audio, our approach produces full-body motion that aligns gestures with speech while orienting the agent according to the user. Our architecture combines a causal transformer-based VAE with interleaved latent tokens for streaming inference and a flow matching model conditioned on user trajectory and audio. To support varying gaze preferences, we introduce a gaze scoring mechanism with classifier-free guidance to decouple learning from control: the model captures natural spatial alignment from data, while users can adjust eye contact intensity at inference time. On the Embody 3D dataset, our method achieves state-of-the-art motion quality at over 300 FPS -- 3x faster than non-causal baselines -- while capturing the subtle spatial dynamics of natural conversation. We validate our approach on a live VR system, bringing spatially-aware conversational agents to real-time deployment. Please see https://evonneng.github.io/sarah/ for details.