John Finley

Humane video conferencing

I set a UX vision and strategy for humane and intelligent video conference features for a 60+ person team:

  • We focused on the human aspects of telepresence, creating a product development environment that emphasized empathy and collaboration.
  • We started with science and made honest and difficult choices based on user data. These choices gave us confidence we were building the right thing.
  • We built to learn. Offline rooms, dedicated lab networks, dSLRs strapped to walls… We built our own VC system to deeply understand the medium in which we work.
A meeting room with participants sized small in the distance — the bowling-alley effect.
We took over meeting rooms to install our own hardware, including multiple cameras
A close-up framing where remote participants appear life-size to those in the room.
What would a meeting be like if everyone had their own camera and shared the same virtual view?
A research prototype: a meeting room rigged with dSLR cameras strapped to the walls.
We tested novel display arrangements that tapped into human psychology
A diagram of camera angles and participant positions used to study natural-scale framing.
Test rooms were fully instrumented

The humans in the room

We identified the “bowling alley effect” as a major pain point in meetings – a blocker to how people see and are seen to others. This fundamentally affects communication and meeting outcomes.

Detecting a distant participant's face is the first challenge in framing everyone fairly

From this, we developed the concept of natural scale, where participants are shown life-size when they speak, resulting in improved communication metrics. This led us to implementing computer vision to seamlessly control the meeting room camera in a way that ensures fairness for all users.

Face position drives how the camera crops the room, framing each participant at natural scale

89% of participants reported satisfaction with seeing others at natural scale, compared to 74% in the default configuration.

A template for innovation

Our work led us to pioneer other advanced features – many of which have been launched to Google Workspace users.

Animation of a camera automatically panning and zooming to keep meeting participants centered as they move.
Automatic camera framing
Before-and-after of a participant's face under poor lighting, then re-lit by machine learning to appear evenly illuminated.
Machine-generated portrait lighting
A meeting room with a laptop on the table showing the Companion mode UI alongside the room's main video feed.
Companion mode lets meeting room participants follow along with their personal device
A dark video frame brightened in real time by computational photography, revealing a clear view of the participant.
Low-light capture enhancement