In this talk, based on my ICASSP-02 presentation this year, I will describe a new method for refined estimation of amplitude and frequency trajectories in sinusoidal modeling. The method consists of performing a phase-vocoder style modeling of individual peaks in the short-time Fourier transform (STFT). When a peak corresponds to a true sinusoid for the duration of the analysis window, the instantaneous amplitude and frequency reduce to constants. More generally, however, narrow-band amplitude and frequency modulations can be accurately recovered over the duration of the analysis frame. The refined estimates can replace the usual piecewise-linear amplitude and frequency trajectories in sinusoidal modeling, allowing for more accurate modeling of musical attacks and pulsed waveforms such as voice. Instantaneous phase may be optionally tracked as well. In summary, the peak-adaptive phase vocoder can be viewed as a phase vocoder having analysis channels which are constructed adaptively about time-varying peaks in the STFT.