Our method performs visual-speech aware 3D reconstruction so that speech perception from the original footage is preserved in the reconstructed talking head. On the left we include the word/phrase ...