Abstract: One variation of a method for enabling autopilot functionality during a video call includes, during an operating period: receiving a sequence of frames from a camera in a first device; detecting a face, of a first user, in the sequence of frames; generating a sequence of facial landmark containers representing facial actions of the face of the first user; transmitting the sequence of facial landmark containers to a second device for combination with a look model to generate a first synthetic image feed depicting facial actions of the first user; and detecting a trigger event. The method also includes, during an autopilot operating period, entering autopilot mode; retrieving a prerecorded autopilot sequence of facial landmark containers from a memory; and transmitting the prerecorded autopilot sequence of facial landmark containers to the second device for combination with the look model to generate a second synthetic image feed depicting predefined facial actions.

Abstract: A method for publishing a synthetic video feed during a video call including, during an operating period: tracking a computational load of a first device; and receiving a sequence of frames in a video feed from a camera facing a first user. The method also includes, responsive to the computational load of the first device falling below a first computational load threshold: detecting the first user's face in the sequence of frames; generating facial landmark containers representing facial actions of the first user; inserting the facial landmark containers and a look model, into a synthetic face generator to generate a first synthetic video feed; and publishing the first synthetic video feed for access by a second device. The method further includes, responsive to the computational load of the first device exceeding the first computational load threshold, offloading generation of a second synthetic video feed to the second device.

Abstract: One variation of a method for enabling autopilot functionality during a video call includes, during an operating period: receiving a sequence of frames from a camera in a first device; detecting a face, of a first user, in the sequence of frames; generating a sequence of facial landmark containers representing facial actions of the face of the first user; transmitting the sequence of facial landmark containers to a second device for combination with a look model to generate a first synthetic image feed depicting facial actions of the first user; and detecting a trigger event. The method also includes, during an autopilot operating period, entering autopilot mode; retrieving a prerecorded autopilot sequence of facial landmark containers from a memory; and transmitting the prerecorded autopilot sequence of facial landmark containers to the second device for combination with the look model to generate a second synthetic image feed depicting predefined facial actions.

Abstract: A method for publishing a synthetic video feed during a video call including, during an operating period: tracking a computational load of a first device; and receiving a sequence of frames in a video feed from a camera facing a first user. The method also includes, responsive to the computational load of the first device falling below a first computational load threshold: detecting the first user's face in the sequence of frames; generating facial landmark containers representing facial actions of the first user; inserting the facial landmark containers and a look model, into a synthetic face generator to generate a first synthetic video feed; and publishing the first synthetic video feed for access by a second device. The method further includes, responsive to the computational load of the first device exceeding the first computational load threshold, offloading generation of a second synthetic video feed to the second device.

Abstract: One variation of a method for enabling autopilot functionality during a video call includes, during an operating period: receiving a sequence of frames from a camera in a first device; detecting a face, of a first user, in the sequence of frames; generating a sequence of facial landmark containers representing facial actions of the face of the first user; transmitting the sequence of facial landmark containers to a second device for combination with a look model to generate a first synthetic image feed depicting facial actions of the first user; and detecting a trigger event. The method also includes, during an autopilot operating period, entering autopilot mode; retrieving a prerecorded autopilot sequence of facial landmark containers from a memory; and transmitting the prerecorded autopilot sequence of facial landmark containers to the second device for combination with the look model to generate a second synthetic image feed depicting predefined facial actions.