A system includes a processor and logic integrated with the processor, executable by the processor, or integrated with and executable by the processor. The logic is configured to train a participant model on video streams of a participant and train the participant model on environmental parameters. The logic is also configured to store the trained participant model. The trained participant model is usable to generate a synthetic video representation of the participant. A computer-implemented method includes generating, by the computer, a synthetic video representation of a participant based on at least a trained participant model corresponding to the participant and live audio input. In response to a trigger event, the method includes replacing, by the computer, live video transmission with a synthetic video transmission that is based at least in part on the synthetic video representation.