A novel sound mixing method, which can mix a voice and background music (BGM) without hiding the voice, is proposed. The proposed method can be used without having to consider the levels of input signals. We implemented this effective method by precise nonlinear processing on the time frequency plane. We use the name "smart mixer" to refer to the type of sound mixer used in this method. In this paper, we describe the basic concept of the smart mixer and its application to mixing a voice and BGM. Experiments on three sets of real audio signals were performed and the results are discussed.