Supplemental material:

See our paper for details:

"Visual Speech Synthesis From 3D Mesh Sequences Driven By Combined Speech Features" ICME 2017
(pdf)

Methods:
Xpost: Ours.
Ppost: Phoneme based.
Conly: MFCC based.
GT: Ground truth recordings.

GT
Conly
Ppost
Xpost
GT
Conly
Ppost
Xpost