Google's Starline paper describes a booth, not a headset

↗

What the record actually documents

Google's own publication listing for Project Starline: A high-fidelity telepresence system names the venue as ACM Transactions on Graphics, in the Proceedings of SIGGRAPH Asia, volume 40, issue 6, dated 2021, placing the peer-reviewed paper in December 2021 alongside that year's conference. This is worth stating plainly because a technical paper under this title is sometimes associated with 2023 in secondary summaries; Google's own listing, opened directly, places it in 2021, some months after the company's original product announcement of 18 May 2021, which describes internal use across Google's Bay Area, New York and Seattle offices and early demonstrations to healthcare and media partners.

An autostereoscopic light-field booth, not a headset

The paper's abstract, as Google presents it, describes a real-time bidirectional system built around a head-tracked autostereoscopic display: a lenticular screen that steers slightly different images to each eye without glasses, combined with multi-view three-dimensional capture, face tracking, spatial audio and compressed colour-and-depth video streamed over a network. Because the display tracks the viewer's head, correct stereo and motion parallax depend on sitting within the booth's tracked zone and largely facing the display; this is presence built from a specific hardware volume on each end of the call, not a general-purpose camera-and-screen setup. The May 2021 announcement adds that a proprietary light-field display renders a life-sized counterpart with working eye contact, a claim about apparent gaze direction rather than a claim about full free-viewpoint movement.

What the evidence supports and what it does not

Google's listing states the system was assessed as demonstrably better than ordinary two-dimensional video calling, based on participant ratings of presence and engagement, meeting recall, and observed nonverbal behaviour such as head nods. That is a company-run user study of its own product, described on the company's own publication page, not an independent replication, and the announcement itself frames the 2021 rollout as limited to internal offices and selected enterprise partners rather than general availability. Neither source states unit cost, room size requirements beyond the booth itself, or how many simultaneous participants a single unit supports.

• Is this specific booth the same generation of hardware the 2021 paper describes, or a later revision with different specifications?
• Was the cited user study run by the vendor, an independent lab, or some combination, and on what sample size?
• What happens to image and audio quality if a participant leans out of the display's tracked viewing zone?

A head-tracked light-field booth and a video call solve the same nominal problem, being face to face at a distance, through very different means, and the paper's contribution is specifically to the booth: capture, compression and a display that tracks one viewer's eyes, not a claim about telepresence in general.