Here we present the video results of NED with and without CDRL on the MEAD and RAVDESS datasets.

From left to right, they are Source || Reference || NED || NED with CDRL.

Please turn on the sound to better assess the synchronization of the mouth shape and the audio.