Video conferencing (VC) typically requires a high bandwidth connection, which may not be available to users in certain situations or locations. There is also sometimes an expectation, not always met, that all participants in a VC switch on their video feed so as to participate equally. This disclosure describes the use of machine-learning techniques to reduce video-conferencing bandwidth by enabling participants to display synthesized videos of themselves to other participants. Since the bulk of the task of generating a participant's video is performed on the remote device, the videoconference can be carried out with bandwidth utilization comparable to that required for audio transmission alone. With user permission, the synthetic video obtained per the techniques can be operative at the receiver end even if the sender has turned off their camera, such that the video-conferencing anxiety felt by some users is mitigated. The described techniques are implemented with specific user permission and users are provided with options to turn off video generation features.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.