Photorealistic telepresence is remote real-time communication in which participants appear as lifelike, volumetrically accurate representations rather than video tiles or stylised avatars. It reconstructs a person’s appearance and motion using techniques such as neural avatars and 3D Gaussian splatting, rendering them convincingly in a shared spatial scene. The goal is a sense of co-presence that closely matches being physically together.
Content
- Systems capture a subject with multi-view cameras or depth sensors, build a driveable photoreal model of their face and body, and render it in real time into a shared 3D space viewed through headsets or light-field displays. Achieving convincing co-presence requires low latency, accurate relighting, and faithful expression transfer, which remain active research challenges constraining the technology to early deployments.