Facial expressions convey very important information which is sometimes indispensable for people to fully understand the speech during communication. When the volume of video data is overwhelming and channel resources are limited, very low bit rate video coding becomes a solution for teleconferencing. In this paper, we propose an efficient and robust very low bit rate face coding method via visual face tracking. After the face is located in the video frame, the generic facial geometric model is adapted to the face, and facial texture for the model is extracted from the first frame of the video. The facial motion is then tracked and synthesized. The facial motion parameters, synthesized residual error and video background are transmitted at very low bit rate. Experiments show that our method can achieve better PSNR around facial area than H.26L at about the same low bit rate and have better subjective visual effects.