Your face is one of the most important factors in communicating with people. At present, the TV-phone, which can transmit face information, is becoming a part of modern life. In order to transfer more realistic face information, we propose the face shape robot WD-1 (Waseda-Docomo face robot No.1), which can express various face shapes by morphing an average mask (made of SEPTON) through 19 facial control-points. In this study, an average face mask is made in order to express a wide range of shapes with few control points. The position, number and movement range of the control-points are optimized. In addition, we introduce 3-DOFs drive units and control each control-point on a plane, not at a point. In order to evaluate the performance of WD-1, we measure the shape error between the faces of three subjects and the WD-1 equivalents. We elucidate the weaknesses of the current WD-1 implementation: a paucity of control points on the face line and wrinkling of the mask due to excessive local stress