In　order　to　solve　the　problem　of　many　to　many　mapping　between　speech　and　lip　in　visual　speech　synthesis,　this　paper　proposed　a　method　based　on　two-level　HMM　for　visual　speech　synthesis,　which　combined　the　context　of　audio　and　mouth　effectively,　solved　many　to　many　correspondence　between　audio　and　mouth,　synthesized　accurate,　coherent　and　natural　mouth　sequence.　And　this　way　is　also　roboticized　completely.