Text-based Editing of Talking-head Video

Supplemental materials

[back to index]

Comparison: using less data

We show results using visemes from 50, 10 and 5 percent of the full input video.