A frame this late (ms past its time in the voice's playout) is dropped rather than shown. ffmpeg holds a turn's last
frames until more video comes, so they come out when the next turn starts; showing them then would snap the face back
to the previous turn's last pose.
A frame this late (ms past its time in the voice's playout) is dropped rather than shown. ffmpeg holds a turn's last frames until more video comes, so they come out when the next turn starts; showing them then would snap the face back to the previous turn's last pose.