A method and apparatus that generates fill frames for Voice over Internet Protocol (VoIP) applications in a communication device is disclosed. The method may include determining if there is a lost frame in a received communication, wherein if it is determined that there is a lost frame, setting a frame loss flag and storing the frame loss flag in the frame loss history register, shifting a loss history register, a line spectral frequency (LSF) history register, a voicing cutoff (VCUT) history register, a pitch history register, and a root mean squared (RMS) gain history register, wherein the loss history register, the LSF history register, the VCUT history register, the pitch history register, and the RMS history register include at least three registers, the three registers being a newest, a middle and an oldest registers, reading the frame loss flag into a newest loss history register, determining contents of the middle register of each of the LSF history register, the VCUT history register, the pitch history register, and the RMS history register, and sending the contents of the middle registers to a synthesizer to generate an output speech signal.